v3.11.1

v3.11.0
v3.10.6
2025-09-17 23:44:27 +00:00 · 2020-01-06 13:55:17 -05:00 · 2020-01-02 23:54:08 -05:00 · 2019-12-25 01:26:26 -05:00 · 2019-12-21 13:19:29 -05:00 · 2019-12-09 15:59:02 -05:00
368 changed files with 51024 additions and 27442 deletions
--- a/81
+++ b/81
@@ -1,12 +1,14 @@


-Requirements:
+1. Requirements:
+---------------

 Intel Core2 or newer, or AMD Steamroller or newer CPU. ARM CPUs are not
 supported.
 64 bit Linux operating system. Apple is not supported.

-Building on linux prerequisites:
+2. Building on linux prerequisites:
+-----------------------------------

 It is assumed users know how to install packages on their system and
 be able to compile standard source packages. This is basic Linux and
@@ -20,49 +22,74 @@ http://askubuntu.com/questions/457526/how-to-install-cpuminer-in-ubuntu

 Install any additional dependencies needed by cpuminer-opt. The list below
 are some of the ones that may not be in the default install and need to
-be installed manually. There may be others, read the error messages they
-will give a clue as to the missing package.
+be installed manually. There may be others, read the compiler error messages,
+they will give a clue as to the missing package.

 The following command should install everything you need on Debian based
-distributions such as Ubuntu:
+distributions such as Ubuntu. Fedora and other distributions may have similar
+but different package names. 

-sudo apt-get install build-essential libssl-dev libcurl4-openssl-dev libjansson-dev libgmp-dev automake zlib1g-dev
-
-build-essential  (Development Tools package group on Fedora)
-automake
-libjansson-dev
-libgmp-dev
-libcurl4-openssl-dev
-libssl-dev
-lib-thread
-zlib1g-dev
+$ sudo apt-get install build-essential automake libssl-dev libcurl4-openssl-dev libjansson-dev libgmp-dev zlib1g-dev git

 SHA support on AMD Ryzen CPUs requires gcc version 5 or higher and
-openssl 1.1.0e or higher. Add one of the following, depending on the
-compiler version, to CFLAGS:
-"-march=native" or "-march=znver1" or "-msha".
+openssl 1.1.0e or higher. Add one of the following to CFLAGS for SHA
+support depending on your CPU and compiler version:
+
+"-march=native" is always the best choice
+
+"-march=znver1" for Ryzen 1000 & 2000 series, znver2 for 3000.
+
+"-msha"  Add SHA to other tuning options

 Additional instructions for static compilalation can be found here:
 https://lxadm.com/Static_compilation_of_cpuminer
 Static builds should only considered in a homogeneous HW and SW environment.
 Local builds will always have the best performance and compatibility.

-Extract cpuminer source.
+3. Download cpuminer-opt
+------------------------

-tar xvzf cpuminer-opt-x.y.z.tar.gz
-cd cpuminer-opt-x.y.z
+Download the source code for the latest realease from the official repository.

-Run ./build.sh to build on Linux or execute the following commands.
+https://github.com/JayDDee/cpuminer-opt/releases

-./autogen.sh
-CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
-make
+Extract the source code.

-Start mining.
+$ tar xvzf cpuminer-opt-x.y.z.tar.gz
+
+
+Alternatively it can be cloned from git.
+
+$ git clone https://github.com/JayDDee/cpuminer-opt.git
+ 
+4. Build cpuminer-opt
+---------------------
+
+It is recomended to Build with default options, this will usuallly
+produce the best results.
+
+$ ./build.sh to build on Linux or execute the following commands.
+
+or 
+
+$ ./autogen.sh
+$ CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
+$ make -j n
+
+n is the number of threads.
+
+5. Start mining.
+----------------
+
+$ ./cpuminer -a algo -o url -u username -p password

-./cpuminer -a algo -o url -u username -p password

 Windows
+-------
+
+See also INSTAL_WINDOWS
+
+The following procedure is obsolete and uses an old compiler.

 Precompiled Windows binaries are built on a Linux host using Mingw
 with a more recent compiler than the following Windows hosted procedure.
--- a/5
+++ b/5
@@ -22,14 +22,13 @@ Step by step...

 Refer to Linux compile instructions and install required packages.

-Additionally, install mingw-64.
+Additionally, install mingw-w64.

 sudo apt-get install mingw-w64


 2. Create a local library directory for packages to be compiled in the next
-   step. Recommended location is $HOME/usr/lib/
-
+   step. Suggested location is $HOME/usr/lib/

 3. Download and build other packages for mingw that don't have a mingw64
   version available in the repositories.
--- a/Makefile.am
+++ b/Makefile.am
@@ -18,7 +18,6 @@ dist_man_MANS	= cpuminer.1
 cpuminer_SOURCES = \
  cpu-miner.c \
  util.c \
-  uint256.cpp \
  api.c \
  sysinfos.c \
  algo-gate-api.c\
@@ -51,12 +50,15 @@ cpuminer_SOURCES = \
  algo/blake/blake.c \
  algo/blake/blake-4way.c \
  algo/blake/sph_blake2b.c \
-  algo/blake/blake2b.c \
  algo/blake/sph-blake2s.c \
  algo/blake/blake2s-hash-4way.c \
  algo/blake/blake2s.c \
  algo/blake/blake2s-gate.c \
  algo/blake/blake2s-4way.c \
+  algo/blake/blake2b-hash-4way.c \
+  algo/blake/blake2b.c \
+  algo/blake/blake2b-gate.c \
+  algo/blake/blake2b-4way.c \
  algo/blake/blakecoin-gate.c \
  algo/blake/mod_blakecoin.c \
  algo/blake/blakecoin.c \
@@ -71,6 +73,9 @@ cpuminer_SOURCES = \
  algo/bmw/bmw256-hash-4way.c \
  algo/bmw/bmw512-hash-4way.c \
  algo/bmw/bmw256.c \
+  algo/bmw/bmw512-gate.c \
+  algo/bmw/bmw512.c \
+  algo/bmw/bmw512-4way.c \
  algo/cryptonight/cryptolight.c \
  algo/cryptonight/cryptonight-common.c\
  algo/cryptonight/cryptonight-aesni.c\
@@ -79,10 +84,14 @@ cpuminer_SOURCES = \
  algo/cubehash/cubehash_sse2.c\
  algo/cubehash/cube-hash-2way.c \
  algo/echo/sph_echo.c \
+  algo/echo/echo-hash-4way.c \
  algo/echo/aes_ni/hash.c\
  algo/gost/sph_gost.c \
+  algo/groestl/groestl-gate.c \
+  algo/groestl/groestl512-hash-4way.c \
  algo/groestl/sph_groestl.c \
  algo/groestl/groestl.c \
+  algo/groestl/groestl-4way.c \
  algo/groestl/myrgr-gate.c \
  algo/groestl/myrgr-4way.c \
  algo/groestl/myr-groestl.c \
@@ -111,13 +120,15 @@ cpuminer_SOURCES = \
  algo/keccak/keccak-hash-4way.c \
  algo/keccak/keccak-4way.c\
  algo/keccak/keccak-gate.c \
-  algo/keccak/sse2/keccak.c \
+  algo/lanehash/lane.c \
  algo/luffa/sph_luffa.c \
  algo/luffa/luffa.c \
  algo/luffa/luffa_for_sse2.c \
  algo/luffa/luffa-hash-2way.c \
  algo/lyra2/lyra2.c \
  algo/lyra2/sponge.c \
+  algo/lyra2/sponge-2way.c \
+  algo/lyra2/lyra2-hash-2way.c \
  algo/lyra2/lyra2-gate.c \
  algo/lyra2/lyra2rev2.c \
  algo/lyra2/lyra2rev2-4way.c \
@@ -138,6 +149,7 @@ cpuminer_SOURCES = \
  algo/nist5/nist5-4way.c \
  algo/nist5/nist5.c \
  algo/nist5/zr5.c \
+  algo/panama/panama-hash-4way.c \
  algo/panama/sph_panama.c \
  algo/radiogatun/sph_radiogatun.c \
  algo/quark/quark-gate.c \
@@ -163,11 +175,10 @@ cpuminer_SOURCES = \
  algo/scrypt/scrypt.c \
  algo/scrypt/neoscrypt.c \
  algo/scrypt/pluck.c \
-  algo/scryptjane/scrypt-jane.c \
  algo/sha/sph_sha2.c \
  algo/sha/sph_sha2big.c \
-  algo/sha/sha2-hash-4way.c \
-  algo/sha/sha256_hash_11way.c \
+  algo/sha/sha256-hash-4way.c \
+  algo/sha/sha512-hash-4way.c \
  algo/sha/sha2.c \
  algo/sha/sha256t-gate.c \
  algo/sha/sha256t-4way.c \
@@ -179,6 +190,7 @@ cpuminer_SOURCES = \
  algo/shavite/sph_shavite.c \
  algo/shavite/sph-shavite-aesni.c \
  algo/shavite/shavite-hash-2way.c \
+  algo/shavite/shavite-hash-4way.c \
  algo/shavite/shavite.c \
  algo/simd/sph_simd.c \
  algo/simd/nist.c \
@@ -191,9 +203,9 @@ cpuminer_SOURCES = \
  algo/skein/skein-gate.c \
  algo/skein/skein2.c \
  algo/skein/skein2-4way.c \
-  algo/skein/skein2-gate.c \
  algo/sm3/sm3.c \
  algo/sm3/sm3-hash-4way.c \
+  algo/swifftx/swifftx.c \
  algo/tiger/sph_tiger.c \
  algo/whirlpool/sph_whirlpool.c \
  algo/whirlpool/whirlpool-hash-4way.c \
@@ -238,6 +250,8 @@ cpuminer_SOURCES = \
  algo/x13/skunk-4way.c \
  algo/x13/skunk.c \
  algo/x13/drop.c \
+  algo/x13/x13bcd-4way.c \
+  algo/x13/x13bcd.c \
  algo/x14/x14-gate.c \
  algo/x14/x14.c \
  algo/x14/x14-4way.c \
@@ -254,6 +268,13 @@ cpuminer_SOURCES = \
  algo/x16/x16r-gate.c \
  algo/x16/x16r.c \
  algo/x16/x16r-4way.c \
+  algo/x16/x16rv2.c \
+  algo/x16/x16rv2-4way.c \
+  algo/x16/x16rt.c \
+  algo/x16/x16rt-4way.c \
+  algo/x16/hex.c \
+  algo/x16/x21s-4way.c \
+  algo/x16/x21s.c \
  algo/x17/x17-gate.c \
  algo/x17/x17.c \
  algo/x17/x17-4way.c \
@@ -264,10 +285,17 @@ cpuminer_SOURCES = \
  algo/x17/sonoa-4way.c \
  algo/x17/sonoa.c \
  algo/x20/x20r.c \
+  algo/x22/x22i-4way.c \
+  algo/x22/x22i.c \
+  algo/x22/x22i-gate.c \
+  algo/x22/x25x.c \
+  algo/x22/x25x-4way.c \
  algo/yescrypt/yescrypt.c \
  algo/yescrypt/sha256_Y.c \
  algo/yescrypt/yescrypt-best.c \
-  algo/yespower/yespower.c \
+  algo/yespower/yespower-gate.c \
+  algo/yespower/yespower-blake2b.c \
+  algo/yespower/crypto/blake2b-yp.c \
  algo/yespower/sha256_p.c \
  algo/yespower/yespower-opt.c

--- a/README.md
+++ b/README.md
@@ -24,7 +24,7 @@ Requirements

 1. A x86_64 architecture CPU with a minimum of SSE2 support. This includes
 Intel Core2 and newer and AMD equivalents. In order to take advantage of AES_NI
-optimizations a CPU with AES_NI is required. This includes Intel Westbridge
+optimizations a CPU with AES_NI is required. This includes Intel Westmere
 and newer and AMD equivalents. Further optimizations are available on some
 algoritms for CPUs with AVX and AVX2, Sandybridge and Haswell respectively.

@@ -55,9 +55,11 @@ Supported Algorithms
                          axiom         Shabal-256 MemoHash
                          bastion
                          blake         Blake-256 (SFR)
-                          blakecoin     blake256r8
+                          blake2b       Blake2b 256
                          blake2s       Blake-2 S
+                          blakecoin     blake256r8
                          bmw           BMW 256
+                          bmw512        BMW 512
                          c11           Chaincoin
                          decred
                          deep          Deepcoin (DCN)
@@ -66,6 +68,7 @@ Supported Algorithms
                          fresh         Fresh
                          groestl       Groestl coin
                          heavy         Heavy
+                          hex           x16r-hex
                          hmq1725       Espers
                          hodl          Hodlcoin
                          jha           Jackpotcoin
@@ -84,10 +87,12 @@ Supported Algorithms
                          neoscrypt     NeoScrypt(128, 2, 1)
                          nist5         Nist5
                          pentablake    Pentablake
-                          phi1612       phi, LUX coin (original algo)
-                          phi2          LUX coin (new algo)
+                          phi1612       phi
+                          phi2          Luxcoin (LUX)
+                          phi2-lux      identical to phi2
                          pluck         Pluck:128 (Supcoin)
                          polytimos     Ninja
+                          power2b       MicroBitcoin (MBC)
                          quark         Quark
                          qubit         Qubit
                          scrypt        scrypt(1024, 1, 1) (default)
@@ -113,12 +118,19 @@ Supported Algorithms
                          x11gost       sib (SibCoin)
                          x12           Galaxie Cash (GCH)
                          x13           X13
+                          x13bcd        bcd
                          x13sm3        hsr (Hshare)
                          x14           X14
                          x15           X15
-                          x16r          Ravencoin (RVN)
-                          x16s          pigeoncoin (PGN)
+                          x16r          
+                          x16rv2        Ravencoin (RVN)
+                          x16rt         Gincoin (GIN)
+                          x16rt-veil    Veil (VEIL)
+                          x16s          Pigeoncoin (PGN)
                          x17
+                          x21s
+                          x22i
+                          x25x
                          xevan         Bitsend (BSD)
                          yescrypt      Globalboost-Y (BSTY)
                          yescryptr8    BitZeny (ZNY)
@@ -126,11 +138,15 @@ Supported Algorithms
                          yescryptr32   WAVI
                          yespower      Cryply
                          yespowerr16   Yenten (YTN)
+                          yespower-b2b  generic yespower + blake2b
                          zr5           Ziftr

 Errata
 ------

+Old algorithms that are no longer used frequently will not have the latest
+optimizations.
+
 Cryptonight and variants are no longer supported, use another miner.

 Neoscrypt crashes on Windows, use legacy version.
@@ -148,14 +164,17 @@ Benchmark testing does not work for x11evo.
 Bugs
 ----

-Users are encouraged to post their bug reports on the Bitcoin Talk
-forum at:
+Users are encouraged to post their bug reports using git issues or on the
+Bitcoin Talk forum or opening an issue in git:

 https://bitcointalk.org/index.php?topic=1326803.0

-All problem reports must be accompanied by a proper definition.
+https://github.com/JayDDee/cpuminer-opt/issues
+
+All problem reports must be accompanied by a proper problem definition.
 This should include how the problem occurred, the command line and
-output from the miner showing the startup and any errors.
+output from the miner showing the startup messages and any errors.
+A history is also useful, ie did it work before.

 Donations
 ---------
@@ -163,10 +182,6 @@ Donations
 cpuminer-opt has no fees of any kind but donations are accepted.

 BTC: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT
- ETH: 0x72122edabcae9d3f57eab0729305a425f6fef6d0
- LTC: LdUwoHJnux9r9EKqFWNvAi45kQompHk6e8
- BCH: 1QKYkB6atn4P7RFozyziAXLEnurwnUM1cQ
- BTG: GVUyECtRHeC5D58z9F3nGGfVQndwnsPnHQ

 Happy mining!

--- a/README.txt
+++ b/README.txt
@@ -15,20 +15,29 @@ the features listed at cpuminer startup to ensure you are mining at
 optimum speed using the best available features.

 Architecture names and compile options used are only provided for Intel
-Core series. Even the newest Pentium and Celeron CPUs are often missing
-features.
+Core series. Budget CPUs like Pentium and Celeron are often missing the
+latest features.

 AMD CPUs older than Piledriver, including Athlon x2 and Phenom II x4, are not
 supported by cpuminer-opt due to an incompatible implementation of SSE2 on
 these CPUs. Some algos may crash the miner with an invalid instruction.
 Users are recommended to use an unoptimized miner such as cpuminer-multi.

+More information for Intel and AMD CPU architectures and their features
+can be found on Wikipedia.
+
+https://en.wikipedia.org/wiki/List_of_Intel_CPU_microarchitectures
+
+https://en.wikipedia.org/wiki/List_of_AMD_CPU_microarchitectures
+
+
 Exe name                Compile flags            Arch name

 cpuminer-sse2.exe      "-msse2"                  Core2, Nehalem   
 cpuminer-aes-sse42.exe "-march=westmere"         Westmere
-cpuminer-avx.exe       "-march=corei7-avx"       Sandy-Ivybridge
-cpuminer-avx2.exe      "-march=core-avx2"        Haswell, Sky-Kaby-Coffeelake
+cpuminer-avx.exe       "-march=corei7-avx"       Sandybridge
+cpuminer-avx2.exe      "-march=core-avx2 -maes"  Haswell, Skylake, Coffeelake
+cpuminer-avx512.exe    "-march=skylake-avx512"   Skylake-X, Cascadelake-X
 cpuminer-zen           "-march=znver1"           AMD Ryzen, Threadripper

 If you like this software feel free to donate:
--- a/206
+++ b/206
@@ -1,21 +1,17 @@
 cpuminer-opt is a console program run from the command line using the
 keyboard, not the mouse.

-cpuminer-opt now supports HW SHA acceleration available on AMD Ryzen CPUs.
-This feature requires recent SW including GCC version 5 or higher and
-openssl version 1.1 or higher. It may also require using "-march=znver1"
-compile flag.
-
-cpuminer-opt is a console program, if you're using a mouse you're doing it
-wrong.
+See also README.md for list of supported algorithms,

 Security warning
 ----------------

 Miner programs are often flagged as malware by antivirus programs. This is
-a false positive, they are flagged simply because they are cryptocurrency 
-miners. The source code is open for anyone to inspect. If you don't trust 
-the software, don't use it.
+usually a false positive, they are flagged simply because they are
+cryptocurrency miners. However, some malware has been spread using the
+cover that miners are known to be subject to false positives. Always be on
+alert. The source code of cpuminer-opt is open for anyone to inspect.
+If you don't trust the software don't download it.

 The cryptographic hashing code has been taken from trusted sources but has been
 modified for speed at the expense of accepted security practices. This
@@ -25,7 +21,7 @@ required.
 Compile Instructions
 --------------------

-See INSTALL_LINUX or INSTALL_WINDOWS fror compile instruuctions
+See INSTALL_LINUX or INSTALL_WINDOWS for compile instruuctions

 Requirements
 ------------
@@ -33,11 +29,197 @@ Requirements
 Intel Core2 or newer, or AMD Steamroller or newer CPU. ARM CPUs are not
 supported.

-64 bit Linux or Windows operating system. Apple and Android are not supported.
+64 bit Linux or Windows operating system. Apple, Android and Rpi are
+not supported. FreeBSD YMMV.

 Change Log
 ----------

+v3.11.1
+
+Faster panama for x25x AVX2 & AVX512.
+
+Fixed echo VAES for Xevan.
+
+Removed support for scryptjane algo.
+
+Reverted macro implemtations of hash functions to SPH reference code
+for SSE2 versions of algos.
+
+v3.11.0
+
+Fixed x25x AVX512 lane 4 invalid shares.
+
+AVX512 for hex, phi2.
+
+VAES optimzation for Intel Icelake CPUs for most algos recently optimized
+with AVX512, source code only.
+
+v3.10.7
+
+AVX512 for x25x, lbry, x13bcd (bcd).
+
+v3.10.6
+
+Added support for SSL stratum: stratum+tcps://
+
+Added job id reporting again, but leaner, suppressed with --quiet.
+
+AVX512 for x21s, x22i, lyra2z, allium.
+
+Fixed share overflow warnings mining lbry with Ryzen (SHA).
+
+v3.10.5
+
+AVX512 for x17, sonoa, xevan, hmq1725, lyra2rev3, lyra2rev2. 
+Faster hmq1725 AVX2.
+
+v3.10.4
+
+AVX512 for x16r, x16rv2, x16rt, x16s, x16rt-veil (veil).
+
+v3.10.3
+
+AVX512 for x12, x13, x14, x15.
+Fixed x12 AVX2 invalid shares.
+
+v.10.2
+
+AVX512 added for bmw512, c11, phi1612 (phi), qubit, skunk, x11, x11gost (sib).
+Fixed c11 AVX2 invalid shares.
+
+v3.10.1
+
+AVX512 for blake2b, nist5, quark, tribus.
+
+More broken lane fixes, fixed buffer overflow in skein AVX512, fixed
+quark invalid shares AVX2.
+
+Only the highest ranking feature in a class is listed at startup, lower ranking
+features are available but no longer listed.
+
+v3.10.0
+
+AVX512 is now supported on selected algos, Windows binary is now available.
+AVX512 optimizations are available for argon2d, blake2s, keccak, keccakc,
+skein & skein2.
+
+Fixed CPU temperature for some CPU models (Linux only).
+
+Fixed a bug that caused some lanes not to submit shares.
+
+Fixed some previously undetected buffer overflows.
+
+Lyra2rev2 3% faster SSE2 and AVX2.
+
+Added "-fno-asynchronous-unwind-tables" to AVX512 build script for Windows
+to fix known mingw issue.
+
+Changed AVX2 build script to explicitly add AES to address change in
+behaviour in GCC 9. 
+
+v3.9.11
+
+Added x22i & x25x algos.
+Blake2s 2% faster AVX2 with Intel CPU, slower with Ryzen v1, v2 ?
+
+v3.9.10
+
+Faster X* algos with AVX2.
+Small improvements to summary stats report.
+
+v3.9.9.1
+
+Fixed a day1 bug that could cause the miner to idle for up to 2 minutes
+under certain circumstances.
+
+Redesigned summary stats report now includes session statistics.
+
+More robust handling of statistics to reduce corruption.
+
+Removed --hide-diff option.
+
+Better handling of cpu-affinity with more than 64 CPUs.
+
+v3.9.9
+
+Added power2b algo for MicroBitcoin.
+Added generic yespower-b2b (yespower + blake2b) algo to be used with
+the parameters introduced in v3.9.7 for yespower & yescrypt.
+Display additional info when a share is rejected.
+Some low level enhancements and minor tweaking of log output.
+RELEASE_NOTES (this file) and README.md  added to Windows release package.
+
+v3.9.8.1
+
+Summary log report will be generated on stratum diff change or after 5 minutes,
+whichever comes first, to prevent incorrect data in the report.
+
+Removed phi2-lux alias (introduced in v3.9.8) due to Luxcoin's planned fork
+to a new algo. The new Luxcoin algo is not supported by cpuminer-opt.
+Until the fork Luxcoin can be mined using phi2 algo.
+
+--hide-diff option is deprecated and has no effect. It will be removed in a
+future release.
+
+v3.9.8
+
+Changes to log output to provide data more relevant to actual mining
+performance.
+phi2 can now handle pools with a mix of coins that use and don't use roots.
+phi2-lux added as an alias for phi2 as they are identical except for roots.
+Add x16rv2 algo for Ravencoin fork.
+
+v3.9.7
+
+Command line option changes:
+
+"-R" is no longer used as a shortcut for "--retry-pause", users must
+use the long option.
+
+New options:
+
+-N, --param-n: set the N parameter for yescrypt, yespower or scrypt algos
+-R, --param-r: set the R parameter for yescrypt or yespower algos, scrypt is
+     hardcoded with R=1
+-K, --param-key: set the client key/pers parameter for yescrypt/yespower algos.
+
+These options can be used to mine yescrypt or yespower variations using
+the generic yescrypt or yespower algo name and specifying the parameters
+manually. They can even be used to mine variations that aren't formally
+supported by a unique algo name. Existing algos can continue to to be mined
+using their original name without parameters.
+
+v3.9.6.2
+
+New algo blake2b.
+Faster myr-gr on Ryzen using SHA.
+Faster blake2s SSE2.
+Small speedup of around 1% for several other algos.
+
+v3.9.6.1
+
+New algos: x21s, hex (alias x16r-hex).
+
+v3.9.6
+
+New algos: bmw512, x16rt, x16rt-veil (alias veil), x13bcd (alias bcd).
+
+v3.9.5.4
+
+Fixed sha256q AVX2 poor performance.
+Fixed skein2 buffer overflow and restored bswap-interleave optimization.
+More restructuring.
+
+v3.9.5.3
+
+Fix crash mining hodl with aes-sse42.
+More restructuring and share report tweaks.
+
+v3.9.5.2
+
+Revert bswap-interleave optimization for causing crashes on Windows.
+
 v3.9.5.1

 Fixed skein2 crash on Windows.
--- a/algo-gate-api.c
+++ b/algo-gate-api.c
@@ -116,13 +116,10 @@ void init_algo_gate( algo_gate_t* gate )
   gate->get_nonceptr            = (void*)&std_get_nonceptr;
   gate->work_decode             = (void*)&std_le_work_decode;
   gate->decode_extra_data       = (void*)&do_nothing;
-   gate->wait_for_diff           = (void*)&std_wait_for_diff;
-   gate->get_max64               = (void*)&get_max64_0x1fffffLL;
   gate->gen_merkle_root         = (void*)&sha256d_gen_merkle_root;
   gate->stratum_gen_work        = (void*)&std_stratum_gen_work;
   gate->build_stratum_request   = (void*)&std_le_build_stratum_request;
   gate->malloc_txs_request      = (void*)&std_malloc_txs_request;
-   gate->set_target              = (void*)&std_set_target;
   gate->submit_getwork_result   = (void*)&std_le_submit_getwork_result;
   gate->build_block_header      = (void*)&std_build_block_header;
   gate->build_extraheader       = (void*)&std_build_extraheader;
@@ -167,9 +164,10 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
    case ALGO_AXIOM:         register_axiom_algo         ( gate ); break;
    case ALGO_BASTION:       register_bastion_algo       ( gate ); break;
    case ALGO_BLAKE:         register_blake_algo         ( gate ); break;
-    case ALGO_BLAKECOIN:     register_blakecoin_algo     ( gate ); break;
-//    case ALGO_BLAKE2B:      register_blake2b_algo     ( gate ); break;
+    case ALGO_BLAKE2B:       register_blake2b_algo       ( gate ); break;
    case ALGO_BLAKE2S:       register_blake2s_algo       ( gate ); break;
+    case ALGO_BLAKECOIN:     register_blakecoin_algo     ( gate ); break;
+    case ALGO_BMW512:        register_bmw512_algo        ( gate ); break;
    case ALGO_C11:           register_c11_algo           ( gate ); break;
    case ALGO_CRYPTOLIGHT:   register_cryptolight_algo   ( gate ); break;
    case ALGO_CRYPTONIGHT:   register_cryptonight_algo   ( gate ); break;
@@ -181,6 +179,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
    case ALGO_FRESH:         register_fresh_algo         ( gate ); break;
    case ALGO_GROESTL:       register_groestl_algo       ( gate ); break;
    case ALGO_HEAVY:         register_heavy_algo         ( gate ); break;
+    case ALGO_HEX:           register_hex_algo           ( gate ); break;
    case ALGO_HMQ1725:       register_hmq1725_algo       ( gate ); break;
    case ALGO_HODL:          register_hodl_algo          ( gate ); break;
    case ALGO_JHA:           register_jha_algo           ( gate ); break;
@@ -203,10 +202,10 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
    case ALGO_PHI2:          register_phi2_algo          ( gate ); break;
    case ALGO_PLUCK:         register_pluck_algo         ( gate ); break;
    case ALGO_POLYTIMOS:     register_polytimos_algo     ( gate ); break;
+    case ALGO_POWER2B:       register_power2b_algo       ( gate ); break;
    case ALGO_QUARK:         register_quark_algo         ( gate ); break;
    case ALGO_QUBIT:         register_qubit_algo         ( gate ); break;
    case ALGO_SCRYPT:        register_scrypt_algo        ( gate ); break;
-    case ALGO_SCRYPTJANE:    register_scryptjane_algo    ( gate ); break;
    case ALGO_SHA256D:       register_sha256d_algo       ( gate ); break;
    case ALGO_SHA256Q:       register_sha256q_algo       ( gate ); break;
    case ALGO_SHA256T:       register_sha256t_algo       ( gate ); break;
@@ -227,12 +226,19 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
    case ALGO_X11GOST:       register_x11gost_algo       ( gate ); break;
    case ALGO_X12:           register_x12_algo           ( gate ); break;
    case ALGO_X13:           register_x13_algo           ( gate ); break;
+    case ALGO_X13BCD:        register_x13bcd_algo        ( gate ); break;
    case ALGO_X13SM3:        register_x13sm3_algo        ( gate ); break;
    case ALGO_X14:           register_x14_algo           ( gate ); break;
    case ALGO_X15:           register_x15_algo           ( gate ); break;
    case ALGO_X16R:          register_x16r_algo          ( gate ); break;
+    case ALGO_X16RV2:        register_x16rv2_algo        ( gate ); break;
+    case ALGO_X16RT:         register_x16rt_algo         ( gate ); break;
+    case ALGO_X16RT_VEIL:    register_x16rt_veil_algo    ( gate ); break;
    case ALGO_X16S:          register_x16s_algo          ( gate ); break;
    case ALGO_X17:           register_x17_algo           ( gate ); break;
+    case ALGO_X21S:          register_x21s_algo          ( gate ); break;
+    case ALGO_X22I:          register_x22i_algo          ( gate ); break;
+    case ALGO_X25X:          register_x25x_algo          ( gate ); break;
    case ALGO_XEVAN:         register_xevan_algo         ( gate ); break;
 /*    case ALGO_YESCRYPT:     register_yescrypt_05_algo     ( gate ); break;
     case ALGO_YESCRYPTR8:   register_yescryptr8_05_algo   ( gate ); break;
@@ -245,6 +251,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
    case ALGO_YESCRYPTR32:   register_yescryptr32_algo   ( gate ); break;
    case ALGO_YESPOWER:      register_yespower_algo      ( gate ); break;
    case ALGO_YESPOWERR16:   register_yespowerr16_algo   ( gate ); break;
+    case ALGO_YESPOWER_B2B:  register_yespower_b2b_algo  ( gate ); break;
    case ALGO_ZR5:           register_zr5_algo           ( gate ); break;
   default:
      applog(LOG_ERR,"FAIL: algo_gate registration failed, unknown algo %s.\n", algo_names[opt_algo] );
@@ -270,7 +277,7 @@ bool register_json_rpc2( algo_gate_t *gate )
  applog(LOG_WARNING,"supported by cpuminer-opt. Shares submitted will");
  applog(LOG_WARNING,"likely be rejected. Proceed at your own risk.\n");

-  gate->wait_for_diff           = (void*)&do_nothing;
+//  gate->wait_for_diff           = (void*)&do_nothing;
  gate->get_new_work            = (void*)&jr2_get_new_work;
  gate->get_nonceptr            = (void*)&jr2_get_nonceptr;
  gate->stratum_gen_work        = (void*)&jr2_stratum_gen_work;
@@ -309,6 +316,7 @@ const char* const algo_alias_map[][2] =
  { "argon2d-crds",      "argon2d250"   },
  { "argon2d-dyn",       "argon2d500"   },
  { "argon2d-uis",       "argon2d4096"  },
+  { "bcd",               "x13bcd"       },
  { "bitcore",           "timetravel10" },
  { "bitzeny",           "yescryptr8"   },
  { "blake256r8",        "blakecoin"    },
@@ -327,19 +335,16 @@ const char* const algo_alias_map[][2] =
  { "lyra2",             "lyra2re"      },
  { "lyra2v2",           "lyra2rev2"    },
  { "lyra2v3",           "lyra2rev3"    },
-  { "lyra2zoin",         "lyra2z330"    },
  { "myrgr",             "myr-gr"       },
  { "myriad",            "myr-gr"       },
  { "neo",               "neoscrypt"    },
  { "phi",               "phi1612"      },
-//  { "sia",               "blake2b"      },
  { "sib",               "x11gost"      },
  { "timetravel8",       "timetravel"   },
-  { "ziftr",             "zr5"          },
+  { "veil",              "x16rt-veil"   },
+  { "x16r-hex",          "hex"          },
  { "yenten",            "yescryptr16"  },
-  { "yescryptr8k",       "yescrypt"     },
-  { "zcoin",             "lyra2z"       },
-  { "zoin",              "lyra2z330"    },
+  { "ziftr",             "zr5"          },
  { NULL,                NULL           }   
 };

@@ -361,40 +366,3 @@ void get_algo_alias( char** algo_or_alias )
 #undef ALIAS
 #undef PROPER

-bool submit_solution( struct work *work, void *hash,
-                      struct thr_info *thr )
-{
-     work_set_target_ratio( work, hash );
-     if ( submit_work( thr, work ) )
-     {
-         if ( !opt_quiet )
-            applog( LOG_BLUE, "Share %d submitted by thread %d, job %s.",
-                    accepted_share_count + rejected_share_count + 1,
-                    thr->id, work->job_id );
-         return true;
-     }
-     else
-          applog( LOG_WARNING, "Failed to submit share." );
-     return false;
-}
-
-bool submit_lane_solution( struct work *work, void *hash,
-                           struct thr_info *thr, int lane )
-{
-     work_set_target_ratio( work, hash );
-     if ( submit_work( thr, work ) )
-     {
-         if ( !opt_quiet )
-//            applog( LOG_BLUE, "Share %d submitted by thread %d, lane %d.",
-//                    accepted_share_count + rejected_share_count + 1,
-//                    thr->id, lane );
-            applog( LOG_BLUE, "Share %d submitted by thread %d, lane %d, job %s.",
-                    accepted_share_count + rejected_share_count + 1, thr->id,
-                    lane, work->job_id );
-         return true;
-     }
-     else
-          applog( LOG_WARNING, "Failed to submit share." );
-     return false;
-}
-
--- a/algo-gate-api.h
+++ b/algo-gate-api.h
@@ -35,7 +35,7 @@
 //    6. Determine if other non existant functions are required.
 //    That is determined by the need to add code in cpu-miner.c
 //    that applies only to the new algo. That is forbidden. All
-//    algo specific code must be in theh algo's file.
+//    algo specific code must be in the algo's file.
 //
 //    7. If new functions need to be added to the gate add the type
 //    to the structure, declare a null instance in this file and define
@@ -48,10 +48,10 @@
 //    instances as they are defined by default, or unsafe functions that
 //    are not needed by the algo.
 //
-//    9. Add an case entry to the switch/case in function register_gate
+//    9. Add a case entry to the switch/case in function register_gate
 //    in file algo-gate-api.c for the new algo.
 //
-//    10 If a new function type was defined add an entry to ini talgo_gate
+//    10 If a new function type was defined add an entry to init algo_gate
 //    to initialize the new function to its null instance described in step 7.
 //
 //    11. If the new algo has aliases add them to the alias array in
@@ -85,14 +85,16 @@

 typedef  uint32_t set_t;

-#define EMPTY_SET       0
-#define SSE2_OPT        1
-#define AES_OPT         2  
-#define SSE42_OPT       4
-#define AVX_OPT         8
-#define AVX2_OPT     0x10
-#define SHA_OPT      0x20
-#define AVX512_OPT   0x40
+#define EMPTY_SET        0
+#define SSE2_OPT         1
+#define AES_OPT          2  
+#define SSE42_OPT        4
+#define AVX_OPT          8   // Sandybridge
+#define AVX2_OPT      0x10   // Haswell
+#define SHA_OPT       0x20   // sha256 (Ryzen, Ice Lake)
+#define AVX512_OPT    0x40   // AVX512- F, VL, DQ, BW (Skylake-X)
+#define VAES_OPT      0x80   // VAES (Ice Lake)
+

 // return set containing all elements from sets a & b
 inline set_t set_union ( set_t a, set_t b ) { return a | b; }
@@ -108,14 +110,7 @@ inline bool set_excl ( set_t a, set_t b ) { return (a & b) == 0; }

 typedef struct
 {
-// special case, only one target, provides a callback for scanhash to
-// submit work with less overhead.
-// bool (*submit_work )             ( struct thr_info*, const struct work* );
-
 // mandatory functions, must be overwritten
-// Added a 5th arg for the thread_info structure to replace the int thr id
-// in the first arg. Both will co-exist during the trasition.
-//int ( *scanhash ) ( int, struct work*, uint32_t, uint64_t* );
 int ( *scanhash ) ( struct work*, uint32_t, uint64_t*, struct thr_info* );

 // optional unsafe, must be overwritten if algo uses function
@@ -123,28 +118,55 @@ void ( *hash )     ( void*, const void*, uint32_t ) ;
 void ( *hash_suw ) ( void*, const void* );

 //optional, safe to use default in most cases
+
+// Allocate thread local buffers and other initialization specific to miner
+// threads.
 bool ( *miner_thread_init )      ( int );
+
+// Generate global blockheader from stratum data.
 void ( *stratum_gen_work )       ( struct stratum_ctx*, struct work* );
+
+// Get thread local copy of blockheader with unique nonce.
 void ( *get_new_work )           ( struct work*, struct work*, int, uint32_t*,
                                   bool );
+
+// Return pointer to nonce in blockheader.
 uint32_t *( *get_nonceptr )      ( uint32_t* );
-void ( *decode_extra_data )      ( struct work*, uint64_t* );
-void ( *wait_for_diff )          ( struct stratum_ctx* );
-int64_t ( *get_max64 )           ();
+
+// Decode getwork blockheader
 bool ( *work_decode )            ( const json_t*, struct work* );
-void ( *set_target)              ( struct work*, double );
+
+// Extra getwork data
+void ( *decode_extra_data )      ( struct work*, uint64_t* );
+
 bool ( *submit_getwork_result )  ( CURL*, struct work* );
+
 void ( *gen_merkle_root )        ( char*, struct stratum_ctx* );
+
+// Increment extranonce
 void ( *build_extraheader )      ( struct work*, struct stratum_ctx* );
+
 void ( *build_block_header )     ( struct work*, uint32_t, uint32_t*,
-	                           uint32_t*, uint32_t, uint32_t );
+	                                uint32_t*, uint32_t, uint32_t );
+// Build mining.submit message
 void ( *build_stratum_request )  ( char*, struct work*, struct stratum_ctx* );
+
 char* ( *malloc_txs_request )    ( struct work* );
+
+// Big or little
 void ( *set_work_data_endian )   ( struct work* );
+
 double ( *calc_network_diff )    ( struct work* );
+
+// Wait for first work
 bool ( *ready_to_mine )          ( struct work*, struct stratum_ctx*, int );
-void ( *resync_threads )         ( struct work* );
+
+// Diverge mining threads
 bool ( *do_this_thread )         ( int );
+
+// After do_this_thread
+void ( *resync_threads )         ( struct work* );
+
 json_t* (*longpoll_rpc_call)     ( CURL*, int*, char* );
 bool ( *stratum_handle_response )( json_t* );
 set_t optimizations;
@@ -193,23 +215,12 @@ void four_way_not_tested();
 // allways returns failure
 int null_scanhash();

-// Allow algos to submit from scanhash loop.
-bool submit_solution( struct work *work, void *hash,
-                      struct thr_info *thr );
-bool submit_lane_solution( struct work *work, void *hash,
-                          struct thr_info *thr, int lane );
-
- 
-bool submit_work( struct thr_info *thr, const struct work *work_in );
-
 // displays warning
 void null_hash    ();
 void null_hash_suw();

 // optional safe targets, default listed first unless noted.

-void std_wait_for_diff();
-
 uint32_t *std_get_nonceptr( uint32_t *work_data );
 uint32_t *jr2_get_nonceptr( uint32_t *work_data );

@@ -224,25 +235,13 @@ void jr2_stratum_gen_work( struct stratum_ctx *sctx, struct work *work );
 void sha256d_gen_merkle_root( char *merkle_root, struct stratum_ctx *sctx );
 void SHA256_gen_merkle_root ( char *merkle_root, struct stratum_ctx *sctx );

-// pick your favorite or define your own
-int64_t get_max64_0x1fffffLL(); // default
-int64_t get_max64_0x40LL();
-int64_t get_max64_0x3ffff();
-int64_t get_max64_0x3fffffLL();
-int64_t get_max64_0x1ffff();
-int64_t get_max64_0xffffLL();
-
-void std_set_target(    struct work *work, double job_diff );
-void alt_set_target(    struct work* work, double job_diff );
-void scrypt_set_target( struct work *work, double job_diff );
-
 bool std_le_work_decode( const json_t *val, struct work *work );
 bool std_be_work_decode( const json_t *val, struct work *work );
-bool jr2_work_decode( const json_t *val, struct work *work );
+bool jr2_work_decode(    const json_t *val, struct work *work );

 bool std_le_submit_getwork_result( CURL *curl, struct work *work );
 bool std_be_submit_getwork_result( CURL *curl, struct work *work );
-bool jr2_submit_getwork_result( CURL *curl, struct work *work );
+bool jr2_submit_getwork_result(    CURL *curl, struct work *work );

 void std_le_build_stratum_request( char *req, struct work *work );
 void std_be_build_stratum_request( char *req, struct work *work );
@@ -256,8 +255,8 @@ void set_work_data_big_endian( struct work *work );
 double std_calc_network_diff( struct work *work );

 void std_build_block_header( struct work* g_work, uint32_t version,
-	                     uint32_t *prevhash,  uint32_t *merkle_root,
-   	                     uint32_t ntime, uint32_t nbits );
+	                          uint32_t *prevhash,  uint32_t *merkle_root,
+   	                       uint32_t ntime,      uint32_t nbits );

 void std_build_extraheader( struct work *work, struct stratum_ctx *sctx );

@@ -278,8 +277,8 @@ int std_get_work_data_size();
 // by calling the algo's register function.
 bool register_algo_gate( int algo, algo_gate_t *gate );

-// Override any default gate functions that are applicable and do any other
-// algo-specific initialization.
+// Called by algos toverride any default gate functions that are applicable
+// and do any other algo-specific initialization.
 // The register functions for all the algos can be declared here to reduce
 // compiler warnings but that's just more work for devs adding new algos.
 bool register_algo( algo_gate_t *gate );
@@ -292,5 +291,7 @@ bool register_json_rpc2( algo_gate_t *gate );
 // use this to call the hash function of an algo directly, ie util.c test.
 void exec_hash_function( int algo, void *output, const void *pdata );

-void get_algo_alias( char** algo_or_alias );
+// Validate a string as a known algo and alias, updates arg to proper
+// algo name if valid alias, NULL if invalid alias or algo.
+void get_algo_alias( char **algo_or_alias );

--- a/algo/argon2/argon2a/argon2a.c
+++ b/algo/argon2/argon2a/argon2a.c
@@ -74,19 +74,14 @@ int scanhash_argon2( struct work* work, uint32_t max_nonce,
 	return 0;
 }

-int64_t argon2_get_max64 ()
-{
-  return 0x1ffLL;
-}
-
 bool register_argon2_algo( algo_gate_t* gate )
 {
  gate->optimizations = SSE2_OPT | AVX_OPT | AVX2_OPT;
  gate->scanhash        = (void*)&scanhash_argon2;
  gate->hash            = (void*)&argon2hash;
  gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
-  gate->set_target      = (void*)&scrypt_set_target;
-  gate->get_max64       = (void*)&argon2_get_max64;
+  opt_target_factor = 65536.0;
+
  return true;
 };

--- a/algo/argon2/argon2d/argon2d-gate.c
+++ b/algo/argon2/argon2d/argon2d-gate.c
@@ -36,43 +36,39 @@ void argon2d_crds_hash( void *output, const void *input )
 int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
                      uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t _ALIGN(64) endiandata[20];
-        uint32_t _ALIGN(64) hash[8];
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
-        int thr_id = mythr->id;  // thr_id arg is deprecated
+   uint32_t _ALIGN(64) endiandata[20];
+   uint32_t _ALIGN(64) hash[8];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   int thr_id = mythr->id;  // thr_id arg is deprecated
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t Htarg = ptarget[7];
+   uint32_t nonce = first_nonce;

-        const uint32_t first_nonce = pdata[19];
-        const uint32_t Htarg = ptarget[7];
+   swab32_array( endiandata, pdata, 20 );

-        uint32_t nonce = first_nonce;
+   do {
+      be32enc(&endiandata[19], nonce);
+      argon2d_crds_hash( hash, endiandata );
+      if ( hash[7] <= Htarg && fulltest( hash, ptarget ) && !opt_benchmark )
+      {
+          pdata[19] = nonce;
+          submit_solution( work, hash, mythr );
+      }
+      nonce++;
+   } while (nonce < max_nonce && !work_restart[thr_id].restart);

-        swab32_array( endiandata, pdata, 20 );
-
-        do {
-                be32enc(&endiandata[19], nonce);
-                argon2d_crds_hash( hash, endiandata );
-                if ( hash[7] <= Htarg && fulltest( hash, ptarget ) )
-                {
-                        pdata[19] = nonce;
-                        *hashes_done = pdata[19] - first_nonce;
-                        work_set_target_ratio(work, hash);
-                        return 1;
-                }
-                nonce++;
-        } while (nonce < max_nonce && !work_restart[thr_id].restart);
-
-        pdata[19] = nonce;
-        *hashes_done = pdata[19] - first_nonce + 1;
-        return 0;
+   pdata[19] = nonce;
+   *hashes_done = pdata[19] - first_nonce + 1;
+   return 0;
 }

 bool register_argon2d_crds_algo( algo_gate_t* gate )
 {
        gate->scanhash = (void*)&scanhash_argon2d_crds;
        gate->hash = (void*)&argon2d_crds_hash;
-        gate->set_target = (void*)&scrypt_set_target;
        gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
+        opt_target_factor = 65536.0;
        return true;
 }

@@ -107,43 +103,40 @@ void argon2d_dyn_hash( void *output, const void *input )
 int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
                      uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t _ALIGN(64) endiandata[20];
-        uint32_t _ALIGN(64) hash[8];
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
-        int thr_id = mythr->id;  // thr_id arg is deprecated
+   uint32_t _ALIGN(64) endiandata[20];
+   uint32_t _ALIGN(64) hash[8];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   int thr_id = mythr->id;  // thr_id arg is deprecated
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t Htarg = ptarget[7];
+   uint32_t nonce = first_nonce;

-        const uint32_t first_nonce = pdata[19];
-        const uint32_t Htarg = ptarget[7];
+   swab32_array( endiandata, pdata, 20 );

-        uint32_t nonce = first_nonce;
+   do
+   {
+      be32enc(&endiandata[19], nonce);
+      argon2d_dyn_hash( hash, endiandata );
+      if ( hash[7] <= Htarg && fulltest( hash, ptarget ) && !opt_benchmark )
+      {
+          pdata[19] = nonce;
+          submit_solution( work, hash, mythr );
+      }
+      nonce++;
+  } while (nonce < max_nonce && !work_restart[thr_id].restart);

-        swab32_array( endiandata, pdata, 20 );
-
-        do {
-                be32enc(&endiandata[19], nonce);
-                argon2d_dyn_hash( hash, endiandata );
-                if ( hash[7] <= Htarg && fulltest( hash, ptarget ) )
-                {
-                        pdata[19] = nonce;
-                        *hashes_done = pdata[19] - first_nonce;
-                        work_set_target_ratio(work, hash);
-                        return 1;
-                }
-                nonce++;
-        } while (nonce < max_nonce && !work_restart[thr_id].restart);
-
-        pdata[19] = nonce;
-        *hashes_done = pdata[19] - first_nonce + 1;
-        return 0;
+   pdata[19] = nonce;
+   *hashes_done = pdata[19] - first_nonce + 1;
+   return 0;
 }

 bool register_argon2d_dyn_algo( algo_gate_t* gate )
 {
        gate->scanhash = (void*)&scanhash_argon2d_dyn;
        gate->hash = (void*)&argon2d_dyn_hash;
-        gate->set_target = (void*)&scrypt_set_target;
        gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
+        opt_target_factor = 65536.0;
        return true;
 }

@@ -171,11 +164,10 @@ int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
      be32enc( &endiandata[19], n );
      argon2d_hash_raw( t_cost, m_cost, parallelism, (char*) endiandata, 80,
                 (char*) endiandata, 80, (char*) vhash, 32, ARGON2_VERSION_13 );
-      if ( vhash[7] < Htarg && fulltest( vhash, ptarget ) )
+      if ( vhash[7] < Htarg && fulltest( vhash, ptarget ) && !opt_benchmark )
      {
-         *hashes_done = n - first_nonce + 1;
         pdata[19] = n;
-         return true;
+         submit_solution( work, vhash, mythr );
      }
      n++;

@@ -187,14 +179,11 @@ int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
   return 0;
 }

-int64_t get_max64_0x1ff() { return 0x1ff; }
-
 bool register_argon2d4096_algo( algo_gate_t* gate )
 {
        gate->scanhash = (void*)&scanhash_argon2d4096;
-        gate->set_target = (void*)&scrypt_set_target;
-        gate->get_max64  = (void*)&get_max64_0x1ff;
        gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
+        opt_target_factor = 65536.0;
        return true;
 }

--- a/algo/argon2/argon2d/argon2d/core.c
+++ b/algo/argon2/argon2d/argon2d/core.c
@@ -28,6 +28,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <mm_malloc.h>

 #include "core.h"
 #include "argon2d_thread.h"
@@ -99,7 +100,8 @@ int allocate_memory(const argon2_context *context, uint8_t **memory,
    if (context->allocate_cbk) {
        (context->allocate_cbk)(memory, memory_size);
    } else {
-        *memory = malloc(memory_size);
+        *memory = _mm_malloc( memory_size, 64 );
+//        *memory = malloc(memory_size);
    }

    if (*memory == NULL) {
@@ -116,7 +118,8 @@ void free_memory(const argon2_context *context, uint8_t *memory,
    if (context->free_cbk) {
        (context->free_cbk)(memory, memory_size);
    } else {
-        free(memory);
+//        free(memory);
+        _mm_free( memory );
    }
 }

--- a/algo/argon2/argon2d/argon2d/opt.c
+++ b/algo/argon2/argon2d/argon2d/opt.c
@@ -21,7 +21,7 @@

 #include "argon2.h"
 #include "core.h"
-
+#include "simd-utils.h"
 #include "../blake2/blake2.h"
 #include "../blake2/blamka-round-opt.h"

@@ -37,24 +37,28 @@

 #if defined(__AVX512F__)

-static void fill_block(__m512i *state, const block *ref_block,
-                       block *next_block, int with_xor) {
+static void fill_block( __m512i *state, const block *ref_block,
+                       block *next_block, int with_xor )
+{
    __m512i block_XY[ARGON2_512BIT_WORDS_IN_BLOCK];
    unsigned int i;

-    if (with_xor) {
-        for (i = 0; i < ARGON2_512BIT_WORDS_IN_BLOCK; i++) {
-            state[i] = _mm512_xor_si512(
-                state[i], _mm512_loadu_si512((const __m512i *)ref_block->v + i));
-            block_XY[i] = _mm512_xor_si512(
-                state[i], _mm512_loadu_si512((const __m512i *)next_block->v + i));
-        }
-    } else {
-        for (i = 0; i < ARGON2_512BIT_WORDS_IN_BLOCK; i++) {
-            block_XY[i] = state[i] = _mm512_xor_si512(
-                state[i], _mm512_loadu_si512((const __m512i *)ref_block->v + i));
+    if ( with_xor )
+    {
+        for ( i = 0; i < ARGON2_512BIT_WORDS_IN_BLOCK; i++ )
+        {
+            state[i] = _mm512_xor_si512( state[i],
+                      _mm512_load_si512( (const __m512i*)ref_block->v + i ) );
+            block_XY[i] = _mm512_xor_si512( state[i],
+                      _mm512_load_si512( (const __m512i*)next_block->v + i ) );
        }
    }
+    else
+    {
+        for ( i = 0; i < ARGON2_512BIT_WORDS_IN_BLOCK; i++ )
+            block_XY[i] = state[i] = _mm512_xor_si512( state[i],
+                      _mm512_load_si512( (const __m512i*)ref_block->v + i ) );
+    }

    BLAKE2_ROUND_1( state[ 0], state[ 1], state[ 2], state[ 3],
                    state[ 4], state[ 5], state[ 6], state[ 7] );
@@ -66,23 +70,10 @@ static void fill_block(__m512i *state, const block *ref_block,
    BLAKE2_ROUND_2( state[ 1], state[ 3], state[ 5], state[ 7],
                    state[ 9], state[11], state[13], state[15] );

-/*
-    for (i = 0; i < 2; ++i) {
-        BLAKE2_ROUND_1(
-            state[8 * i + 0], state[8 * i + 1], state[8 * i + 2], state[8 * i + 3],
-            state[8 * i + 4], state[8 * i + 5], state[8 * i + 6], state[8 * i + 7]);
-    }
-
-    for (i = 0; i < 2; ++i) {
-        BLAKE2_ROUND_2(
-            state[2 * 0 + i], state[2 * 1 + i], state[2 * 2 + i], state[2 * 3 + i],
-            state[2 * 4 + i], state[2 * 5 + i], state[2 * 6 + i], state[2 * 7 + i]);
-    }
-*/
-
-    for (i = 0; i < ARGON2_512BIT_WORDS_IN_BLOCK; i++) {
-        state[i] = _mm512_xor_si512(state[i], block_XY[i]);
-        _mm512_storeu_si512((__m512i *)next_block->v + i, state[i]);
+    for ( i = 0; i < ARGON2_512BIT_WORDS_IN_BLOCK; i++ )
+    {
+        state[i] = _mm512_xor_si512( state[i], block_XY[i] );
+        _mm512_store_si512( (__m512i*)next_block->v + i, state[i] );
    }
 }

@@ -96,14 +87,14 @@ static void fill_block(__m256i *state, const block *ref_block,
    if (with_xor) {
        for (i = 0; i < ARGON2_HWORDS_IN_BLOCK; i++) {
            state[i] = _mm256_xor_si256(
-                state[i], _mm256_loadu_si256((const __m256i *)ref_block->v + i));
+                state[i], _mm256_load_si256((const __m256i *)ref_block->v + i));
            block_XY[i] = _mm256_xor_si256(
-                state[i], _mm256_loadu_si256((const __m256i *)next_block->v + i));
+                state[i], _mm256_load_si256((const __m256i *)next_block->v + i));
        }
    } else {
        for (i = 0; i < ARGON2_HWORDS_IN_BLOCK; i++) {
            block_XY[i] = state[i] = _mm256_xor_si256(
-                state[i], _mm256_loadu_si256((const __m256i *)ref_block->v + i));
+                state[i], _mm256_load_si256((const __m256i *)ref_block->v + i));
        }
    }

@@ -125,21 +116,9 @@ static void fill_block(__m256i *state, const block *ref_block,
    BLAKE2_ROUND_2( state[ 3], state[ 7], state[11], state[15],
                    state[19], state[23], state[27], state[31] );

-/*
-    for (i = 0; i < 4; ++i) {
-        BLAKE2_ROUND_1(state[8 * i + 0], state[8 * i + 4], state[8 * i + 1], state[8 * i + 5],
-                       state[8 * i + 2], state[8 * i + 6], state[8 * i + 3], state[8 * i + 7]);
-    }
-
-    for (i = 0; i < 4; ++i) {
-        BLAKE2_ROUND_2(state[ 0 + i], state[ 4 + i], state[ 8 + i], state[12 + i],
-                       state[16 + i], state[20 + i], state[24 + i], state[28 + i]);
-    }
-*/
-
    for (i = 0; i < ARGON2_HWORDS_IN_BLOCK; i++) {
        state[i] = _mm256_xor_si256(state[i], block_XY[i]);
-        _mm256_storeu_si256((__m256i *)next_block->v + i, state[i]);
+        _mm256_store_si256((__m256i *)next_block->v + i, state[i]);
    }
 }

@@ -153,14 +132,14 @@ static void fill_block(__m128i *state, const block *ref_block,
    if (with_xor) {
        for (i = 0; i < ARGON2_OWORDS_IN_BLOCK; i++) {
            state[i] = _mm_xor_si128(
-                state[i], _mm_loadu_si128((const __m128i *)ref_block->v + i));
+                state[i], _mm_load_si128((const __m128i *)ref_block->v + i));
            block_XY[i] = _mm_xor_si128(
-                state[i], _mm_loadu_si128((const __m128i *)next_block->v + i));
+                state[i], _mm_load_si128((const __m128i *)next_block->v + i));
        }
    } else {
        for (i = 0; i < ARGON2_OWORDS_IN_BLOCK; i++) {
            block_XY[i] = state[i] = _mm_xor_si128(
-                state[i], _mm_loadu_si128((const __m128i *)ref_block->v + i));
+                state[i], _mm_load_si128((const __m128i *)ref_block->v + i));
        }
    }

@@ -198,22 +177,9 @@ static void fill_block(__m128i *state, const block *ref_block,
    BLAKE2_ROUND( state[ 7], state[15], state[23], state[31],  
                  state[39], state[47], state[55], state[63] );

-/*
-    for (i = 0; i < 8; ++i) {
-        BLAKE2_ROUND(state[8 * i + 0], state[8 * i + 1], state[8 * i + 2],
-            state[8 * i + 3], state[8 * i + 4], state[8 * i + 5],
-            state[8 * i + 6], state[8 * i + 7]);
-    }
-
-    for (i = 0; i < 8; ++i) {
-        BLAKE2_ROUND(state[8 * 0 + i], state[8 * 1 + i], state[8 * 2 + i],
-            state[8 * 3 + i], state[8 * 4 + i], state[8 * 5 + i],
-            state[8 * 6 + i], state[8 * 7 + i]);
-    }
-*/
    for (i = 0; i < ARGON2_OWORDS_IN_BLOCK; i++) {
        state[i] = _mm_xor_si128(state[i], block_XY[i]);
-        _mm_storeu_si128((__m128i *)next_block->v + i, state[i]);
+        _mm_store_si128((__m128i *)next_block->v + i, state[i]);
    }
 }

--- a/algo/argon2/argon2d/blake2/blamka-round-opt.h
+++ b/algo/argon2/argon2d/blake2/blamka-round-opt.h
@@ -29,6 +29,8 @@
 #include <x86intrin.h>
 #endif

+#include "simd-utils.h"
+
 #if !defined(__AVX512F__)
 #if !defined(__AVX2__)
 #if !defined(__XOP__)
@@ -182,64 +184,63 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {

 #include <immintrin.h>

-#define rotr32(x)   _mm256_shuffle_epi32(x, _MM_SHUFFLE(2, 3, 0, 1))
-#define rotr24(x)   _mm256_shuffle_epi8(x, _mm256_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10, 3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10))
-#define rotr16(x)   _mm256_shuffle_epi8(x, _mm256_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9, 2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9))
-#define rotr63(x)   _mm256_xor_si256(_mm256_srli_epi64((x), 63), _mm256_add_epi64((x), (x)))
+#define  rotr32( x )  mm256_ror_64( x, 32 )
+#define  rotr24( x )  mm256_ror_64( x, 24 )
+#define  rotr16( x )  mm256_ror_64( x, 16 )
+#define  rotr63( x )  mm256_rol_64( x,  1 )
+
+//#define rotr32(x)   _mm256_shuffle_epi32(x, _MM_SHUFFLE(2, 3, 0, 1))
+//#define rotr24(x)   _mm256_shuffle_epi8(x, _mm256_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10, 3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10))
+//#define rotr16(x)   _mm256_shuffle_epi8(x, _mm256_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9, 2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9))
+//#define rotr63(x)   _mm256_xor_si256(_mm256_srli_epi64((x), 63), _mm256_add_epi64((x), (x)))

 #define G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
    do { \
-        __m256i ml = _mm256_mul_epu32(A0, B0); \
-        ml = _mm256_add_epi64(ml, ml); \
-        A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml)); \
+        __m256i ml0, ml1; \
+        ml0 = _mm256_mul_epu32(A0, B0); \
+        ml1 = _mm256_mul_epu32(A1, B1); \
+        ml0 = _mm256_add_epi64(ml0, ml0); \
+        ml1 = _mm256_add_epi64(ml1, ml1); \
+        A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml0)); \
+        A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml1)); \
        D0 = _mm256_xor_si256(D0, A0); \
-        D0 = rotr32(D0); \
-        \
-        ml = _mm256_mul_epu32(C0, D0); \
-        ml = _mm256_add_epi64(ml, ml); \
-        C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml)); \
-        \
-        B0 = _mm256_xor_si256(B0, C0); \
-        B0 = rotr24(B0); \
-        \
-        ml = _mm256_mul_epu32(A1, B1); \
-        ml = _mm256_add_epi64(ml, ml); \
-        A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml)); \
        D1 = _mm256_xor_si256(D1, A1); \
+        D0 = rotr32(D0); \
        D1 = rotr32(D1); \
-        \
-        ml = _mm256_mul_epu32(C1, D1); \
-        ml = _mm256_add_epi64(ml, ml); \
-        C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml)); \
-        \
+        ml0 = _mm256_mul_epu32(C0, D0); \
+        ml1 = _mm256_mul_epu32(C1, D1); \
+        ml0 = _mm256_add_epi64(ml0, ml0); \
+        ml1 = _mm256_add_epi64(ml1, ml1); \
+        C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml0)); \
+        C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml1)); \
+        B0 = _mm256_xor_si256(B0, C0); \
        B1 = _mm256_xor_si256(B1, C1); \
+        B0 = rotr24(B0); \
        B1 = rotr24(B1); \
    } while((void)0, 0);

 #define G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
    do { \
-        __m256i ml = _mm256_mul_epu32(A0, B0); \
-        ml = _mm256_add_epi64(ml, ml); \
-        A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml)); \
+        __m256i ml0, ml1; \
+        ml0 = _mm256_mul_epu32(A0, B0); \
+        ml1 = _mm256_mul_epu32(A1, B1); \
+        ml0 = _mm256_add_epi64(ml0, ml0); \
+        ml1 = _mm256_add_epi64(ml1, ml1); \
+        A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml0)); \
+        A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml1)); \
        D0 = _mm256_xor_si256(D0, A0); \
-        D0 = rotr16(D0); \
-        \
-        ml = _mm256_mul_epu32(C0, D0); \
-        ml = _mm256_add_epi64(ml, ml); \
-        C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml)); \
-        B0 = _mm256_xor_si256(B0, C0); \
-        B0 = rotr63(B0); \
-        \
-        ml = _mm256_mul_epu32(A1, B1); \
-        ml = _mm256_add_epi64(ml, ml); \
-        A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml)); \
        D1 = _mm256_xor_si256(D1, A1); \
+        D0 = rotr16(D0); \
        D1 = rotr16(D1); \
-        \
-        ml = _mm256_mul_epu32(C1, D1); \
-        ml = _mm256_add_epi64(ml, ml); \
-        C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml)); \
+        ml0 = _mm256_mul_epu32(C0, D0); \
+        ml1 = _mm256_mul_epu32(C1, D1); \
+        ml0 = _mm256_add_epi64(ml0, ml0); \
+        ml1 = _mm256_add_epi64(ml1, ml1); \
+        C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml0)); \
+        C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml1)); \
+        B0 = _mm256_xor_si256(B0, C0); \
        B1 = _mm256_xor_si256(B1, C1); \
+        B0 = rotr63(B0); \
        B1 = rotr63(B1); \
    } while((void)0, 0);

@@ -259,16 +260,14 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
        __m256i tmp1 = _mm256_blend_epi32(B0, B1, 0xCC); \
        __m256i tmp2 = _mm256_blend_epi32(B0, B1, 0x33); \
        B1 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
-        B0 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
-        \
        tmp1 = C0; \
+        B0 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
        C0 = C1; \
-        C1 = tmp1; \
-        \
-        tmp1 = _mm256_blend_epi32(D0, D1, 0xCC); \
        tmp2 = _mm256_blend_epi32(D0, D1, 0x33); \
-        D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
+        C1 = tmp1; \
+        tmp1 = _mm256_blend_epi32(D0, D1, 0xCC); \
        D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
+        D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
    } while(0);

 #define UNDIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
@@ -287,16 +286,14 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
        __m256i tmp1 = _mm256_blend_epi32(B0, B1, 0xCC); \
        __m256i tmp2 = _mm256_blend_epi32(B0, B1, 0x33); \
        B0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
-        B1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
-        \
        tmp1 = C0; \
+        B1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
        C0 = C1; \
-        C1 = tmp1; \
-        \
-        tmp1 = _mm256_blend_epi32(D0, D1, 0x33); \
        tmp2 = _mm256_blend_epi32(D0, D1, 0xCC); \
-        D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
+        C1 = tmp1; \
+        tmp1 = _mm256_blend_epi32(D0, D1, 0x33); \
        D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
+        D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
    } while((void)0, 0);

 #define BLAKE2_ROUND_1(A0, A1, B0, B1, C0, C1, D0, D1) \
@@ -430,14 +427,14 @@ static __m512i muladd(__m512i x, __m512i y)
 #define SWAP_QUARTERS(A0, A1) \
    do { \
        SWAP_HALVES(A0, A1); \
-        A0 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A0); \
-        A1 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A1); \
+        A0 = _mm512_shuffle_i64x2( A0, A0, 0xd8 ); \
+        A1 = _mm512_shuffle_i64x2( A1, A1, 0xd8 ); \
    } while((void)0, 0)

 #define UNSWAP_QUARTERS(A0, A1) \
    do { \
-        A0 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A0); \
-        A1 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A1); \
+        A0 = _mm512_shuffle_i64x2( A0, A0, 0xd8 ); \
+        A1 = _mm512_shuffle_i64x2( A1, A1, 0xd8 ); \
        SWAP_HALVES(A0, A1); \
    } while((void)0, 0)

--- a/algo/blake/blake-4way.c
+++ b/algo/blake/blake-4way.c
@@ -27,25 +27,19 @@ int scanhash_blake_4way( struct work *work, uint32_t max_nonce,
   uint32_t *ptarget = work->target;
   const uint32_t first_nonce = pdata[19];
   uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
   uint32_t n = first_nonce;
   int thr_id = mythr->id;  // thr_id arg is deprecated

   if (opt_benchmark)
      HTarget = 0x7f;

-   // we need big endian data...
-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
   blake256r14_4way_init( &blake_4w_ctx );
   blake256r14_4way( &blake_4w_ctx, vdata, 64 );

-   uint32_t *noncep = vdata + 76;   // 19*4
   do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );

      blakehash_4way( hash, vdata );

@@ -76,9 +70,9 @@ void blakehash_8way( void *state, const void *input )
     memcpy( &ctx, &blake_8w_ctx, sizeof ctx );
     blake256r14_8way( &ctx, input + (64<<3), 16 );
     blake256r14_8way_close( &ctx, vhash );
-     mm256_dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
-                              state+128, state+160, state+192, state+224,
-                              vhash, 256 );
+     _dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
+                    state+128, state+160, state+192, state+224,
+                    vhash, 256 );
 }

 int scanhash_blake_8way( struct work *work, uint32_t max_nonce,
@@ -90,32 +84,21 @@ int scanhash_blake_8way( struct work *work, uint32_t max_nonce,
   uint32_t *ptarget = work->target;
   const uint32_t first_nonce = pdata[19];
   uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
   uint32_t n = first_nonce;
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
   int thr_id = mythr->id;  // thr_id arg is deprecated

   if (opt_benchmark)
      HTarget = 0x7f;

-   // we need big endian data...
-   swab32_array( edata, pdata, 20 );
-
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );

   blake256r14_8way_init( &blake_8w_ctx );
   blake256r14_8way( &blake_8w_ctx, vdata, 64 );

-   uint32_t *noncep = vdata + 152;   // 19*8
   do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
-      be32enc( noncep +4, n+4 );
-      be32enc( noncep +5, n+5 );
-      be32enc( noncep +6, n+6 );
-      be32enc( noncep +7, n+7 );
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
      pdata[19] = n;

      blakehash_8way( hash, vdata );
--- a/algo/blake/blake-gate.c
+++ b/algo/blake/blake-gate.c
@@ -1,18 +1,8 @@
 #include "blake-gate.h"

-int64_t blake_get_max64 ()
-{
-  return 0x7ffffLL;
-}
-
 bool register_blake_algo( algo_gate_t* gate )
 {
  gate->optimizations = AVX2_OPT;
-  gate->get_max64 = (void*)&blake_get_max64;
-//#if defined (__AVX2__) && defined (FOUR_WAY)
-//   gate->optimizations = SSE2_OPT | AVX2_OPT;
-//  gate->scanhash  = (void*)&scanhash_blake_8way;
-//  gate->hash      = (void*)&blakehash_8way;
 #if defined(BLAKE_4WAY)
  four_way_not_tested();
  gate->scanhash  = (void*)&scanhash_blake_4way;
--- a/algo/blake/blake-hash-4way.h
+++ b/algo/blake/blake-hash-4way.h
@@ -59,7 +59,6 @@ extern "C"{
 typedef struct {
   unsigned char buf[64<<2];
   uint32_t H[8<<2];
-   uint32_t S[4<<2];
 //   __m128i buf[16] __attribute__ ((aligned (64)));
 //   __m128i H[8];
 //   __m128i S[4];    
@@ -71,19 +70,22 @@ typedef struct {
 // Default 14 rounds
 typedef blake_4way_small_context blake256_4way_context;
 void blake256_4way_init(void *ctx);
-void blake256_4way(void *ctx, const void *data, size_t len);
+void blake256_4way_update(void *ctx, const void *data, size_t len);
+#define blake256_4way blake256_4way_update
 void blake256_4way_close(void *ctx, void *dst);

 // 14 rounds, blake, decred
 typedef blake_4way_small_context blake256r14_4way_context;
 void blake256r14_4way_init(void *cc);
-void blake256r14_4way(void *cc, const void *data, size_t len);
+void blake256r14_4way_update(void *cc, const void *data, size_t len);
+#define blake256r14_4way blake256r14_4way_update
 void blake256r14_4way_close(void *cc, void *dst);

 // 8 rounds, blakecoin, vanilla
 typedef blake_4way_small_context blake256r8_4way_context;
 void blake256r8_4way_init(void *cc);
-void blake256r8_4way(void *cc, const void *data, size_t len);
+void blake256r8_4way_update(void *cc, const void *data, size_t len);
+#define blake256r8_4way blake256r8_4way_update
 void blake256r8_4way_close(void *cc, void *dst);

 #ifdef __AVX2__
@@ -93,7 +95,6 @@ void blake256r8_4way_close(void *cc, void *dst);
 typedef struct {
   __m256i buf[16] __attribute__ ((aligned (64)));
   __m256i H[8];
-   __m256i S[4];
   size_t ptr;
   sph_u32 T0, T1;
   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
@@ -102,38 +103,93 @@ typedef struct {
 // Default 14 rounds
 typedef blake_8way_small_context blake256_8way_context;
 void blake256_8way_init(void *cc);
-void blake256_8way(void *cc, const void *data, size_t len);
+void blake256_8way_update(void *cc, const void *data, size_t len);
+//#define blake256_8way blake256_8way_update
 void blake256_8way_close(void *cc, void *dst);

 // 14 rounds, blake, decred
 typedef blake_8way_small_context blake256r14_8way_context;
 void blake256r14_8way_init(void *cc);
-void blake256r14_8way(void *cc, const void *data, size_t len);
+void blake256r14_8way_update(void *cc, const void *data, size_t len);
 void blake256r14_8way_close(void *cc, void *dst);

 // 8 rounds, blakecoin, vanilla
 typedef blake_8way_small_context blake256r8_8way_context;
 void blake256r8_8way_init(void *cc);
-void blake256r8_8way(void *cc, const void *data, size_t len);
+void blake256r8_8way_update(void *cc, const void *data, size_t len);
+#define blake256r8_8way blake256r8_8way_update
 void blake256r8_8way_close(void *cc, void *dst);

 // Blake-512 4 way

 typedef struct {
-   __m256i buf[16] __attribute__ ((aligned (64)));
+   __m256i buf[16];
   __m256i H[8];
   __m256i S[4];   
   size_t ptr;
   sph_u64 T0, T1;
-} blake_4way_big_context;
+} blake_4way_big_context __attribute__ ((aligned (128)));

 typedef blake_4way_big_context blake512_4way_context;

-void blake512_4way_init(void *cc);
-void blake512_4way(void *cc, const void *data, size_t len);
-void blake512_4way_close(void *cc, void *dst);
-void blake512_4way_addbits_and_close(
-	void *cc, unsigned ub, unsigned n, void *dst);
+void blake512_4way_init( blake_4way_big_context *sc );
+void blake512_4way_update( void *cc, const void *data, size_t len );
+#define blake512_4way blake512_4way_update
+void blake512_4way_close( void *cc, void *dst );
+void blake512_4way_addbits_and_close( void *cc, unsigned ub, unsigned n,
+                                      void *dst );
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+//Blake-256 16 way
+
+typedef struct {
+   __m512i buf[16];
+   __m512i H[8];
+   size_t ptr;
+   uint32_t T0, T1;
+   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
+} blake_16way_small_context __attribute__ ((aligned (128)));
+
+// Default 14 rounds
+typedef blake_16way_small_context blake256_16way_context;
+void blake256_16way_init(void *cc);
+void blake256_16way_update(void *cc, const void *data, size_t len);
+void blake256_16way_close(void *cc, void *dst);
+
+// 14 rounds, blake, decred
+typedef blake_16way_small_context blake256r14_16way_context;
+void blake256r14_16way_init(void *cc);
+void blake256r14_16way_update(void *cc, const void *data, size_t len);
+void blake256r14_16way_close(void *cc, void *dst);
+
+// 8 rounds, blakecoin, vanilla
+typedef blake_16way_small_context blake256r8_16way_context;
+void blake256r8_16way_init(void *cc);
+void blake256r8_16way_update(void *cc, const void *data, size_t len);
+void blake256r8_16way_close(void *cc, void *dst);
+
+
+// Blake-512 8 way
+
+typedef struct {
+   __m512i buf[16];
+   __m512i H[8];
+   __m512i S[4];
+   size_t ptr;
+   sph_u64 T0, T1;
+} blake_8way_big_context __attribute__ ((aligned (128)));
+
+typedef blake_8way_big_context blake512_8way_context;
+
+void blake512_8way_init( blake_8way_big_context *sc );
+void blake512_8way_update( void *cc, const void *data, size_t len );
+void blake512_8way_close( void *cc, void *dst );
+void blake512_8way_addbits_and_close( void *cc, unsigned ub, unsigned n,
+                                      void *dst );
+
+#endif  // AVX512
+

 #endif  // AVX2

--- a/algo/blake/blake256-hash-4way.c
+++ b/algo/blake/blake256-hash-4way.c
@@ -304,16 +304,17 @@ static const sph_u32 CS[16] = {

 #endif

+// Blake-256 4 way

 #define GS_4WAY( m0, m1, c0, c1, a, b, c, d ) \
 do { \
-   a = _mm_add_epi32( _mm_add_epi32( _mm_xor_si128( \
-                 _mm_set_epi32( c1, c1, c1, c1 ), m0 ), b ), a ); \
+   a = _mm_add_epi32( _mm_add_epi32( a, b ), \
+                      _mm_xor_si128( _mm_set1_epi32( c1 ), m0 ) ); \
   d = mm128_ror_32( _mm_xor_si128( d, a ), 16 ); \
   c = _mm_add_epi32( c, d ); \
   b = mm128_ror_32( _mm_xor_si128( b, c ), 12 ); \
-   a = _mm_add_epi32( _mm_add_epi32( _mm_xor_si128( \
-                 _mm_set_epi32( c0, c0, c0, c0 ), m1 ), b ), a ); \
+   a = _mm_add_epi32( _mm_add_epi32( a, b ), \
+                      _mm_xor_si128( _mm_set1_epi32( c0 ), m1 ) ); \
   d = mm128_ror_32( _mm_xor_si128( d, a ), 8 ); \
   c = _mm_add_epi32( c, d ); \
   b = mm128_ror_32( _mm_xor_si128( b, c ), 7 ); \
@@ -321,7 +322,8 @@ do { \

 #if SPH_COMPACT_BLAKE_32

-// Blake-256 4 way
+// Not used
+#if 0

 #define ROUND_S_4WAY(r)   do { \
 	GS_4WAY(M[sigma[r][0x0]], M[sigma[r][0x1]], \
@@ -342,6 +344,8 @@ do { \
 		CS[sigma[r][0xE]], CS[sigma[r][0xF]], V3, V4, V9, VE); \
 } while (0)

+#endif
+
 #else

 #define ROUND_S_4WAY(r)   do { \
@@ -359,7 +363,6 @@ do { \

 #define DECL_STATE32_4WAY \
 	__m128i H0, H1, H2, H3, H4, H5, H6, H7; \
-	__m128i S0, S1, S2, S3; \
        uint32_t T0, T1;

 #define READ_STATE32_4WAY(state)   do { \
@@ -371,10 +374,6 @@ do { \
 		H5 = casti_m128i( state->H, 5 ); \
 		H6 = casti_m128i( state->H, 6 ); \
 		H7 = casti_m128i( state->H, 7 ); \
-		S0 = casti_m128i( state->S, 0 ); \
-		S1 = casti_m128i( state->S, 1 ); \
-		S2 = casti_m128i( state->S, 2 ); \
-		S3 = casti_m128i( state->S, 3 ); \
 		T0 = (state)->T0; \
 		T1 = (state)->T1; \
 	} while (0)
@@ -388,17 +387,13 @@ do { \
 		casti_m128i( state->H, 5 ) = H5; \
 		casti_m128i( state->H, 6 ) = H6; \
 		casti_m128i( state->H, 7 ) = H7; \
-		casti_m128i( state->S, 0 ) = S0; \
-		casti_m128i( state->S, 1 ) = S1; \
-		casti_m128i( state->S, 2 ) = S2; \
-		casti_m128i( state->S, 3 ) = S3; \
 		(state)->T0 = T0; \
 		(state)->T1 = T1; \
 	} while (0)

 #if SPH_COMPACT_BLAKE_32
 // not used
-
+#if 0
 #define COMPRESS32_4WAY( rounds )   do { \
 	__m128i M[16]; \
 	__m128i V0, V1, V2, V3, V4, V5, V6, V7; \
@@ -412,34 +407,16 @@ do { \
 	V5 = H5; \
 	V6 = H6; \
 	V7 = H7; \
-        V8 = _mm_xor_si128( S0, _mm_set_epi32( CS0, CS0, CS0, CS0 ) ); \
-        V9 = _mm_xor_si128( S1, _mm_set_epi32( CS1, CS1, CS1, CS1 ) ); \
-        VA = _mm_xor_si128( S2, _mm_set_epi32( CS2, CS2, CS2, CS2 ) ); \
-        VB = _mm_xor_si128( S3, _mm_set_epi32( CS3, CS3, CS3, CS3 ) ); \
-        VC = _mm_xor_si128( _mm_set_epi32( T0, T0, T0, T0 ), \
-                            _mm_set_epi32( CS4, CS4, CS4, CS4 ) ); \
-        VD = _mm_xor_si128( _mm_set_epi32( T0, T0, T0, T0 ), \
-                            _mm_set_epi32( CS5, CS5, CS5, CS5 ) ); \
-        VE = _mm_xor_si128( _mm_set_epi32( T1, T1, T1, T1 ) \
-                          , _mm_set_epi32( CS6, CS6, CS6, CS6 ) ); \
-        VF = _mm_xor_si128( _mm_set_epi32( T1, T1, T1, T1 ), \
-                            _mm_set_epi32( CS7, CS7, CS7, CS7 ) ); \
-	M[0x0] = mm128_bswap_32( *(buf +  0) ); \
-	M[0x1] = mm128_bswap_32( *(buf +  1) ); \
-	M[0x2] = mm128_bswap_32( *(buf +  2) ); \
-	M[0x3] = mm128_bswap_32( *(buf +  3) ); \
-	M[0x4] = mm128_bswap_32( *(buf +  4) ); \
-	M[0x5] = mm128_bswap_32( *(buf +  5) ); \
-	M[0x6] = mm128_bswap_32( *(buf +  6) ); \
-	M[0x7] = mm128_bswap_32( *(buf +  7) ); \
-	M[0x8] = mm128_bswap_32( *(buf +  8) ); \
-	M[0x9] = mm128_bswap_32( *(buf +  9) ); \
-	M[0xA] = mm128_bswap_32( *(buf + 10) ); \
-	M[0xB] = mm128_bswap_32( *(buf + 11) ); \
-	M[0xC] = mm128_bswap_32( *(buf + 12) ); \
-	M[0xD] = mm128_bswap_32( *(buf + 13) ); \
-	M[0xE] = mm128_bswap_32( *(buf + 14) ); \
-	M[0xF] = mm128_bswap_32( *(buf + 15) ); \
+   V8 = _mm_xor_si128( S0, _mm_set1_epi32( CS0 ) ); \
+   V9 = _mm_xor_si128( S1, _mm_set1_epi32( CS1 ) ); \
+   VA = _mm_xor_si128( S2, _mm_set1_epi32( CS2 ) ); \
+   VB = _mm_xor_si128( S3, _mm_set1_epi32( CS3 ) ); \
+   VC = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS4 ) ); \
+   VD = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS5 ) ); \
+   VE = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS6 ) ); \
+   VF = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS7 ) ); \
+   mm128_block_bswap_32( M, buf ); \
+   mm128_block_bswap_32( M+8, buf+8 ); \
 	for (r = 0; r < rounds; r ++) \
 		ROUND_S_4WAY(r); \
        H0 = _mm_xor_si128( _mm_xor_si128( \
@@ -459,11 +436,60 @@ do { \
        H7 = _mm_xor_si128( _mm_xor_si128( \
                                   _mm_xor_si128( S3, V7 ), VF ), H7 ); \
 	} while (0)
+#endif

 #else

 // current impl

+#if defined(__SSSE3__)
+
+#define BLAKE256_4WAY_BLOCK_BSWAP32 do \
+{ \
+   __m128i shuf_bswap32 = _mm_set_epi64x( 0x0c0d0e0f08090a0b, \
+                                          0x0405060700010203 ); \
+   M0 = _mm_shuffle_epi8( buf[ 0], shuf_bswap32 ); \
+   M1 = _mm_shuffle_epi8( buf[ 1], shuf_bswap32 ); \
+   M2 = _mm_shuffle_epi8( buf[ 2], shuf_bswap32 ); \
+   M3 = _mm_shuffle_epi8( buf[ 3], shuf_bswap32 ); \
+   M4 = _mm_shuffle_epi8( buf[ 4], shuf_bswap32 ); \
+   M5 = _mm_shuffle_epi8( buf[ 5], shuf_bswap32 ); \
+   M6 = _mm_shuffle_epi8( buf[ 6], shuf_bswap32 ); \
+   M7 = _mm_shuffle_epi8( buf[ 7], shuf_bswap32 ); \
+   M8 = _mm_shuffle_epi8( buf[ 8], shuf_bswap32 ); \
+   M9 = _mm_shuffle_epi8( buf[ 9], shuf_bswap32 ); \
+   MA = _mm_shuffle_epi8( buf[10], shuf_bswap32 ); \
+   MB = _mm_shuffle_epi8( buf[11], shuf_bswap32 ); \
+   MC = _mm_shuffle_epi8( buf[12], shuf_bswap32 ); \
+   MD = _mm_shuffle_epi8( buf[13], shuf_bswap32 ); \
+   ME = _mm_shuffle_epi8( buf[14], shuf_bswap32 ); \
+   MF = _mm_shuffle_epi8( buf[15], shuf_bswap32 ); \
+} while(0)
+
+#else  // SSE2
+
+#define BLAKE256_4WAY_BLOCK_BSWAP32 do \
+{ \
+   M0 = mm128_bswap_32( buf[0] ); \
+   M1 = mm128_bswap_32( buf[1] ); \
+   M2 = mm128_bswap_32( buf[2] ); \
+   M3 = mm128_bswap_32( buf[3] ); \
+   M4 = mm128_bswap_32( buf[4] ); \
+   M5 = mm128_bswap_32( buf[5] ); \
+   M6 = mm128_bswap_32( buf[6] ); \
+   M7 = mm128_bswap_32( buf[7] ); \
+   M8 = mm128_bswap_32( buf[8] ); \
+   M9 = mm128_bswap_32( buf[9] ); \
+   MA = mm128_bswap_32( buf[10] ); \
+   MB = mm128_bswap_32( buf[11] ); \
+   MC = mm128_bswap_32( buf[12] ); \
+   MD = mm128_bswap_32( buf[13] ); \
+   ME = mm128_bswap_32( buf[14] ); \
+   MF = mm128_bswap_32( buf[15] ); \
+} while(0)
+
+#endif  // SSSE3 else SSE2
+
 #define COMPRESS32_4WAY( rounds ) \
 do { \
   __m128i M0, M1, M2, M3, M4, M5, M6, M7; \
@@ -478,30 +504,19 @@ do { \
   V5 = H5; \
   V6 = H6; \
   V7 = H7; \
-   V8 = _mm_xor_si128( S0, _mm_set1_epi32( CS0 ) ); \
-   V9 = _mm_xor_si128( S1, _mm_set1_epi32( CS1 ) ); \
-   VA = _mm_xor_si128( S2, _mm_set1_epi32( CS2 ) ); \
-   VB = _mm_xor_si128( S3, _mm_set1_epi32( CS3 ) ); \
-   VC = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS4 ) ); \
-   VD = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS5 ) ); \
-   VE = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS6 ) ); \
-   VF = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS7 ) ); \
-   M0 = mm128_bswap_32( buf[ 0] ); \
-   M1 = mm128_bswap_32( buf[ 1] ); \
-   M2 = mm128_bswap_32( buf[ 2] ); \
-   M3 = mm128_bswap_32( buf[ 3] ); \
-   M4 = mm128_bswap_32( buf[ 4] ); \
-   M5 = mm128_bswap_32( buf[ 5] ); \
-   M6 = mm128_bswap_32( buf[ 6] ); \
-   M7 = mm128_bswap_32( buf[ 7] ); \
-   M8 = mm128_bswap_32( buf[ 8] ); \
-   M9 = mm128_bswap_32( buf[ 9] ); \
-   MA = mm128_bswap_32( buf[10] ); \
-   MB = mm128_bswap_32( buf[11] ); \
-   MC = mm128_bswap_32( buf[12] ); \
-   MD = mm128_bswap_32( buf[13] ); \
-   ME = mm128_bswap_32( buf[14] ); \
-   MF = mm128_bswap_32( buf[15] ); \
+   V8 = m128_const1_64( 0x243F6A88243F6A88 ); \
+   V9 = m128_const1_64( 0x85A308D385A308D3 ); \
+   VA = m128_const1_64( 0x13198A2E13198A2E ); \
+   VB = m128_const1_64( 0x0370734403707344 ); \
+   VC = _mm_xor_si128( _mm_set1_epi32( T0 ), \
+                           m128_const1_64( 0xA4093822A4093822 ) ); \
+   VD = _mm_xor_si128( _mm_set1_epi32( T0 ), \
+                           m128_const1_64( 0x299F31D0299F31D0 ) ); \
+   VE = _mm_xor_si128( _mm_set1_epi32( T1 ), \
+                           m128_const1_64( 0x082EFA98082EFA98 ) ); \
+   VF = _mm_xor_si128( _mm_set1_epi32( T1 ), \
+                           m128_const1_64( 0xEC4E6C89EC4E6C89 ) ); \
+   BLAKE256_4WAY_BLOCK_BSWAP32; \
   ROUND_S_4WAY(0); \
   ROUND_S_4WAY(1); \
   ROUND_S_4WAY(2); \
@@ -519,14 +534,14 @@ do { \
      ROUND_S_4WAY(2); \
      ROUND_S_4WAY(3); \
   } \
-   H0 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( V8, V0 ), S0 ), H0 ); \
-   H1 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( V9, V1 ), S1 ), H1 ); \
-   H2 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( VA, V2 ), S2 ), H2 ); \
-   H3 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( VB, V3 ), S3 ), H3 ); \
-   H4 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( VC, V4 ), S0 ), H4 ); \
-   H5 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( VD, V5 ), S1 ), H5 ); \
-   H6 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( VE, V6 ), S2 ), H6 ); \
-   H7 = _mm_xor_si128( _mm_xor_si128( _mm_xor_si128( VF, V7 ), S3 ), H7 ); \
+   H0 = _mm_xor_si128( _mm_xor_si128( V8, V0 ), H0 ); \
+   H1 = _mm_xor_si128( _mm_xor_si128( V9, V1 ), H1 ); \
+   H2 = _mm_xor_si128( _mm_xor_si128( VA, V2 ), H2 ); \
+   H3 = _mm_xor_si128( _mm_xor_si128( VB, V3 ), H3 ); \
+   H4 = _mm_xor_si128( _mm_xor_si128( VC, V4 ), H4 ); \
+   H5 = _mm_xor_si128( _mm_xor_si128( VD, V5 ), H5 ); \
+   H6 = _mm_xor_si128( _mm_xor_si128( VE, V6 ), H6 ); \
+   H7 = _mm_xor_si128( _mm_xor_si128( VF, V7 ), H7 ); \
 } while (0)

 #endif
@@ -537,13 +552,13 @@ do { \

 #define GS_8WAY( m0, m1, c0, c1, a, b, c, d ) \
 do { \
-   a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
-                 _mm256_set1_epi32( c1 ), m0 ), b ), a ); \
+   a = _mm256_add_epi32( _mm256_add_epi32( a, b ), \
+                         _mm256_xor_si256( _mm256_set1_epi32( c1 ), m0 ) ); \
   d = mm256_ror_32( _mm256_xor_si256( d, a ), 16 ); \
   c = _mm256_add_epi32( c, d ); \
   b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
-   a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
-                 _mm256_set1_epi32( c0 ), m1 ), b ), a ); \
+   a = _mm256_add_epi32( _mm256_add_epi32( a, b ), \
+                         _mm256_xor_si256( _mm256_set1_epi32( c0 ), m1 ) ); \
   d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
   c = _mm256_add_epi32( c, d ); \
   b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
@@ -562,7 +577,6 @@ do { \

 #define DECL_STATE32_8WAY \
   __m256i H0, H1, H2, H3, H4, H5, H6, H7; \
-   __m256i S0, S1, S2, S3; \
   sph_u32 T0, T1;

 #define READ_STATE32_8WAY(state) \
@@ -575,10 +589,6 @@ do { \
   H5 = (state)->H[5]; \
   H6 = (state)->H[6]; \
   H7 = (state)->H[7]; \
-   S0 = (state)->S[0]; \
-   S1 = (state)->S[1]; \
-   S2 = (state)->S[2]; \
-   S3 = (state)->S[3]; \
   T0 = (state)->T0; \
   T1 = (state)->T1; \
 } while (0)
@@ -593,10 +603,6 @@ do { \
   (state)->H[5] = H5; \
   (state)->H[6] = H6; \
   (state)->H[7] = H7; \
-   (state)->S[0] = S0; \
-   (state)->S[1] = S1; \
-   (state)->S[2] = S2; \
-   (state)->S[3] = S3; \
   (state)->T0 = T0; \
   (state)->T1 = T1; \
 } while (0)
@@ -607,6 +613,7 @@ do { \
   __m256i M8, M9, MA, MB, MC, MD, ME, MF; \
   __m256i V0, V1, V2, V3, V4, V5, V6, V7; \
   __m256i V8, V9, VA, VB, VC, VD, VE, VF; \
+   __m256i shuf_bswap32; \
   V0 = H0; \
   V1 = H1; \
   V2 = H2; \
@@ -615,30 +622,36 @@ do { \
   V5 = H5; \
   V6 = H6; \
   V7 = H7; \
-   V8 = _mm256_xor_si256( S0, _mm256_set1_epi32( CS0 ) ); \
-   V9 = _mm256_xor_si256( S1, _mm256_set1_epi32( CS1 ) ); \
-   VA = _mm256_xor_si256( S2, _mm256_set1_epi32( CS2 ) ); \
-   VB = _mm256_xor_si256( S3, _mm256_set1_epi32( CS3 ) ); \
-   VC = _mm256_xor_si256( _mm256_set1_epi32( T0 ), _mm256_set1_epi32( CS4 ) ); \
-   VD = _mm256_xor_si256( _mm256_set1_epi32( T0 ), _mm256_set1_epi32( CS5 ) ); \
-   VE = _mm256_xor_si256( _mm256_set1_epi32( T1 ), _mm256_set1_epi32( CS6 ) ); \
-   VF = _mm256_xor_si256( _mm256_set1_epi32( T1 ), _mm256_set1_epi32( CS7 ) ); \
-   M0 = mm256_bswap_32( * buf ); \
-   M1 = mm256_bswap_32( *(buf+1) ); \
-   M2 = mm256_bswap_32( *(buf+2) ); \
-   M3 = mm256_bswap_32( *(buf+3) ); \
-   M4 = mm256_bswap_32( *(buf+4) ); \
-   M5 = mm256_bswap_32( *(buf+5) ); \
-   M6 = mm256_bswap_32( *(buf+6) ); \
-   M7 = mm256_bswap_32( *(buf+7) ); \
-   M8 = mm256_bswap_32( *(buf+8) ); \
-   M9 = mm256_bswap_32( *(buf+9) ); \
-   MA = mm256_bswap_32( *(buf+10) ); \
-   MB = mm256_bswap_32( *(buf+11) ); \
-   MC = mm256_bswap_32( *(buf+12) ); \
-   MD = mm256_bswap_32( *(buf+13) ); \
-   ME = mm256_bswap_32( *(buf+14) ); \
-   MF = mm256_bswap_32( *(buf+15) ); \
+   V8 = m256_const1_64( 0x243F6A88243F6A88 ); \
+   V9 = m256_const1_64( 0x85A308D385A308D3 ); \
+   VA = m256_const1_64( 0x13198A2E13198A2E ); \
+   VB = m256_const1_64( 0x0370734403707344 ); \
+   VC = _mm256_xor_si256( _mm256_set1_epi32( T0 ),\
+                              m256_const1_64( 0xA4093822A4093822 ) ); \
+   VD = _mm256_xor_si256( _mm256_set1_epi32( T0 ),\
+                              m256_const1_64( 0x299F31D0299F31D0 ) ); \
+   VE = _mm256_xor_si256( _mm256_set1_epi32( T1 ), \
+                              m256_const1_64( 0x082EFA98082EFA98 ) ); \
+   VF = _mm256_xor_si256( _mm256_set1_epi32( T1 ), \
+                              m256_const1_64( 0xEC4E6C89EC4E6C89 ) ); \
+   shuf_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b, 0x1415161710111213, \
+                                 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
+   M0 = _mm256_shuffle_epi8( * buf    , shuf_bswap32 ); \
+   M1 = _mm256_shuffle_epi8( *(buf+ 1), shuf_bswap32 ); \
+   M2 = _mm256_shuffle_epi8( *(buf+ 2), shuf_bswap32 ); \
+   M3 = _mm256_shuffle_epi8( *(buf+ 3), shuf_bswap32 ); \
+   M4 = _mm256_shuffle_epi8( *(buf+ 4), shuf_bswap32 ); \
+   M5 = _mm256_shuffle_epi8( *(buf+ 5), shuf_bswap32 ); \
+   M6 = _mm256_shuffle_epi8( *(buf+ 6), shuf_bswap32 ); \
+   M7 = _mm256_shuffle_epi8( *(buf+ 7), shuf_bswap32 ); \
+   M8 = _mm256_shuffle_epi8( *(buf+ 8), shuf_bswap32 ); \
+   M9 = _mm256_shuffle_epi8( *(buf+ 9), shuf_bswap32 ); \
+   MA = _mm256_shuffle_epi8( *(buf+10), shuf_bswap32 ); \
+   MB = _mm256_shuffle_epi8( *(buf+11), shuf_bswap32 ); \
+   MC = _mm256_shuffle_epi8( *(buf+12), shuf_bswap32 ); \
+   MD = _mm256_shuffle_epi8( *(buf+13), shuf_bswap32 ); \
+   ME = _mm256_shuffle_epi8( *(buf+14), shuf_bswap32 ); \
+   MF = _mm256_shuffle_epi8( *(buf+15), shuf_bswap32 ); \
   ROUND_S_8WAY(0); \
   ROUND_S_8WAY(1); \
   ROUND_S_8WAY(2); \
@@ -656,25 +669,155 @@ do { \
      ROUND_S_8WAY(2); \
      ROUND_S_8WAY(3); \
   } \
-   H0 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( V8, V0 ), \
-                                                              S0 ), H0 ); \
-   H1 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( V9, V1 ), \
-                                                              S1 ), H1 ); \
-   H2 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( VA, V2 ), \
-                                                              S2 ), H2 ); \
-   H3 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( VB, V3 ), \
-                                                              S3 ), H3 ); \
-   H4 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( VC, V4 ), \
-                                                              S0 ), H4 ); \
-   H5 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( VD, V5 ), \
-                                                              S1 ), H5 ); \
-   H6 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( VE, V6 ), \
-                                                              S2 ), H6 ); \
-   H7 = _mm256_xor_si256( _mm256_xor_si256( _mm256_xor_si256( VF, V7 ), \
-                                                              S3 ), H7 ); \
+   H0 = _mm256_xor_si256( _mm256_xor_si256( V8, V0 ), H0 ); \
+   H1 = _mm256_xor_si256( _mm256_xor_si256( V9, V1 ), H1 ); \
+   H2 = _mm256_xor_si256( _mm256_xor_si256( VA, V2 ), H2 ); \
+   H3 = _mm256_xor_si256( _mm256_xor_si256( VB, V3 ), H3 ); \
+   H4 = _mm256_xor_si256( _mm256_xor_si256( VC, V4 ), H4 ); \
+   H5 = _mm256_xor_si256( _mm256_xor_si256( VD, V5 ), H5 ); \
+   H6 = _mm256_xor_si256( _mm256_xor_si256( VE, V6 ), H6 ); \
+   H7 = _mm256_xor_si256( _mm256_xor_si256( VF, V7 ), H7 ); \
 } while (0)


+#endif
+
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// Blaske-256 16 way AVX512
+
+#define GS_16WAY( m0, m1, c0, c1, a, b, c, d ) \
+do { \
+   a = _mm512_add_epi32( _mm512_add_epi32( a, b ), \
+                         _mm512_xor_si512( _mm512_set1_epi32( c1 ), m0 ) ); \
+   d = mm512_ror_32( _mm512_xor_si512( d, a ), 16 ); \
+   c = _mm512_add_epi32( c, d ); \
+   b = mm512_ror_32( _mm512_xor_si512( b, c ), 12 ); \
+   a = _mm512_add_epi32( _mm512_add_epi32( a, b ), \
+                         _mm512_xor_si512( _mm512_set1_epi32( c0 ), m1 ) ); \
+   d = mm512_ror_32( _mm512_xor_si512( d, a ), 8 ); \
+   c = _mm512_add_epi32( c, d ); \
+   b = mm512_ror_32( _mm512_xor_si512( b, c ), 7 ); \
+} while (0)
+
+#define ROUND_S_16WAY(r)   do { \
+        GS_16WAY(Mx(r, 0), Mx(r, 1), CSx(r, 0), CSx(r, 1), V0, V4, V8, VC); \
+        GS_16WAY(Mx(r, 2), Mx(r, 3), CSx(r, 2), CSx(r, 3), V1, V5, V9, VD); \
+        GS_16WAY(Mx(r, 4), Mx(r, 5), CSx(r, 4), CSx(r, 5), V2, V6, VA, VE); \
+        GS_16WAY(Mx(r, 6), Mx(r, 7), CSx(r, 6), CSx(r, 7), V3, V7, VB, VF); \
+        GS_16WAY(Mx(r, 8), Mx(r, 9), CSx(r, 8), CSx(r, 9), V0, V5, VA, VF); \
+        GS_16WAY(Mx(r, A), Mx(r, B), CSx(r, A), CSx(r, B), V1, V6, VB, VC); \
+        GS_16WAY(Mx(r, C), Mx(r, D), CSx(r, C), CSx(r, D), V2, V7, V8, VD); \
+        GS_16WAY(Mx(r, E), Mx(r, F), CSx(r, E), CSx(r, F), V3, V4, V9, VE); \
+} while (0)
+
+#define DECL_STATE32_16WAY \
+   __m512i H0, H1, H2, H3, H4, H5, H6, H7; \
+   sph_u32 T0, T1;
+
+#define READ_STATE32_16WAY(state) \
+do { \
+   H0 = (state)->H[0]; \
+   H1 = (state)->H[1]; \
+   H2 = (state)->H[2]; \
+   H3 = (state)->H[3]; \
+   H4 = (state)->H[4]; \
+   H5 = (state)->H[5]; \
+   H6 = (state)->H[6]; \
+   H7 = (state)->H[7]; \
+   T0 = (state)->T0; \
+   T1 = (state)->T1; \
+} while (0)
+
+#define WRITE_STATE32_16WAY(state) \
+do { \
+   (state)->H[0] = H0; \
+   (state)->H[1] = H1; \
+   (state)->H[2] = H2; \
+   (state)->H[3] = H3; \
+   (state)->H[4] = H4; \
+   (state)->H[5] = H5; \
+   (state)->H[6] = H6; \
+   (state)->H[7] = H7; \
+   (state)->T0 = T0; \
+   (state)->T1 = T1; \
+} while (0)
+
+#define COMPRESS32_16WAY( rounds ) \
+do { \
+   __m512i M0, M1, M2, M3, M4, M5, M6, M7; \
+   __m512i M8, M9, MA, MB, MC, MD, ME, MF; \
+   __m512i V0, V1, V2, V3, V4, V5, V6, V7; \
+   __m512i V8, V9, VA, VB, VC, VD, VE, VF; \
+   __m512i shuf_bswap32; \
+   V0 = H0; \
+   V1 = H1; \
+   V2 = H2; \
+   V3 = H3; \
+   V4 = H4; \
+   V5 = H5; \
+   V6 = H6; \
+   V7 = H7; \
+   V8 = m512_const1_64( 0x243F6A88243F6A88 ); \
+   V9 = m512_const1_64( 0x85A308D385A308D3 ); \
+   VA = m512_const1_64( 0x13198A2E13198A2E ); \
+   VB = m512_const1_64( 0x0370734403707344 ); \
+   VC = _mm512_xor_si512( _mm512_set1_epi32( T0 ),\
+                              m512_const1_64( 0xA4093822A4093822 ) ); \
+   VD = _mm512_xor_si512( _mm512_set1_epi32( T0 ),\
+                              m512_const1_64( 0x299F31D0299F31D0 ) ); \
+   VE = _mm512_xor_si512( _mm512_set1_epi32( T1 ), \
+                              m512_const1_64( 0x082EFA98082EFA98 ) ); \
+   VF = _mm512_xor_si512( _mm512_set1_epi32( T1 ), \
+                              m512_const1_64( 0xEC4E6C89EC4E6C89 ) ); \
+   shuf_bswap32 = m512_const_64( 0x3c3d3e3f38393a3b, 0x3435363730313233, \
+                                 0x2c2d2e2f28292a2b, 0x2425262720212223, \
+                                 0x1c1d1e1f18191a1b, 0x1415161710111213, \
+                                 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
+   M0 = _mm512_shuffle_epi8( * buf    , shuf_bswap32 ); \
+   M1 = _mm512_shuffle_epi8( *(buf+ 1), shuf_bswap32 ); \
+   M2 = _mm512_shuffle_epi8( *(buf+ 2), shuf_bswap32 ); \
+   M3 = _mm512_shuffle_epi8( *(buf+ 3), shuf_bswap32 ); \
+   M4 = _mm512_shuffle_epi8( *(buf+ 4), shuf_bswap32 ); \
+   M5 = _mm512_shuffle_epi8( *(buf+ 5), shuf_bswap32 ); \
+   M6 = _mm512_shuffle_epi8( *(buf+ 6), shuf_bswap32 ); \
+   M7 = _mm512_shuffle_epi8( *(buf+ 7), shuf_bswap32 ); \
+   M8 = _mm512_shuffle_epi8( *(buf+ 8), shuf_bswap32 ); \
+   M9 = _mm512_shuffle_epi8( *(buf+ 9), shuf_bswap32 ); \
+   MA = _mm512_shuffle_epi8( *(buf+10), shuf_bswap32 ); \
+   MB = _mm512_shuffle_epi8( *(buf+11), shuf_bswap32 ); \
+   MC = _mm512_shuffle_epi8( *(buf+12), shuf_bswap32 ); \
+   MD = _mm512_shuffle_epi8( *(buf+13), shuf_bswap32 ); \
+   ME = _mm512_shuffle_epi8( *(buf+14), shuf_bswap32 ); \
+   MF = _mm512_shuffle_epi8( *(buf+15), shuf_bswap32 ); \
+   ROUND_S_16WAY(0); \
+   ROUND_S_16WAY(1); \
+   ROUND_S_16WAY(2); \
+   ROUND_S_16WAY(3); \
+   ROUND_S_16WAY(4); \
+   ROUND_S_16WAY(5); \
+   ROUND_S_16WAY(6); \
+   ROUND_S_16WAY(7); \
+   if (rounds == 14) \
+   { \
+      ROUND_S_16WAY(8); \
+      ROUND_S_16WAY(9); \
+      ROUND_S_16WAY(0); \
+      ROUND_S_16WAY(1); \
+      ROUND_S_16WAY(2); \
+      ROUND_S_16WAY(3); \
+   } \
+   H0 = _mm512_xor_si512( _mm512_xor_si512( V8, V0 ), H0 ); \
+   H1 = _mm512_xor_si512( _mm512_xor_si512( V9, V1 ), H1 ); \
+   H2 = _mm512_xor_si512( _mm512_xor_si512( VA, V2 ), H2 ); \
+   H3 = _mm512_xor_si512( _mm512_xor_si512( VB, V3 ), H3 ); \
+   H4 = _mm512_xor_si512( _mm512_xor_si512( VC, V4 ), H4 ); \
+   H5 = _mm512_xor_si512( _mm512_xor_si512( VD, V5 ), H5 ); \
+   H6 = _mm512_xor_si512( _mm512_xor_si512( VE, V6 ), H6 ); \
+   H7 = _mm512_xor_si512( _mm512_xor_si512( VF, V7 ), H7 ); \
+} while (0)
+
 #endif

 // Blake-256 4 way
@@ -685,32 +828,22 @@ static void
 blake32_4way_init( blake_4way_small_context *ctx, const uint32_t *iv,
                   const uint32_t *salt, int rounds )
 {
-   casti_m128i( ctx->H, 0 ) = _mm_set1_epi32( iv[0] );
-   casti_m128i( ctx->H, 1 ) = _mm_set1_epi32( iv[1] );
-   casti_m128i( ctx->H, 2 ) = _mm_set1_epi32( iv[2] );
-   casti_m128i( ctx->H, 3 ) = _mm_set1_epi32( iv[3] );
-   casti_m128i( ctx->H, 4 ) = _mm_set1_epi32( iv[4] );
-   casti_m128i( ctx->H, 5 ) = _mm_set1_epi32( iv[5] );
-   casti_m128i( ctx->H, 6 ) = _mm_set1_epi32( iv[6] );
-   casti_m128i( ctx->H, 7 ) = _mm_set1_epi32( iv[7] );
-
-   casti_m128i( ctx->S, 0 ) = m128_zero;
-   casti_m128i( ctx->S, 1 ) = m128_zero;
-   casti_m128i( ctx->S, 2 ) = m128_zero;
-   casti_m128i( ctx->S, 3 ) = m128_zero;
-/*
-   sc->S[0] = _mm_set1_epi32( salt[0] );
-   sc->S[1] = _mm_set1_epi32( salt[1] );
-   sc->S[2] = _mm_set1_epi32( salt[2] );
-   sc->S[3] = _mm_set1_epi32( salt[3] );
-*/
+   casti_m128i( ctx->H, 0 ) = m128_const1_64( 0x6A09E6676A09E667 );
+   casti_m128i( ctx->H, 1 ) = m128_const1_64( 0xBB67AE85BB67AE85 );
+   casti_m128i( ctx->H, 2 ) = m128_const1_64( 0x3C6EF3723C6EF372 );
+   casti_m128i( ctx->H, 3 ) = m128_const1_64( 0xA54FF53AA54FF53A );
+   casti_m128i( ctx->H, 4 ) = m128_const1_64( 0x510E527F510E527F );
+   casti_m128i( ctx->H, 5 ) = m128_const1_64( 0x9B05688C9B05688C );
+   casti_m128i( ctx->H, 6 ) = m128_const1_64( 0x1F83D9AB1F83D9AB );
+   casti_m128i( ctx->H, 7 ) = m128_const1_64( 0x5BE0CD195BE0CD19 );
   ctx->T0 = ctx->T1 = 0;
   ctx->ptr = 0;
   ctx->rounds = rounds;
 }

 static void
-blake32_4way( blake_4way_small_context *ctx, const void *data, size_t len )
+blake32_4way( blake_4way_small_context *ctx, const void *data,
+              size_t len )
 {
   __m128i *buf = (__m128i*)ctx->buf;
   size_t  bptr = ctx->ptr<<2;
@@ -773,12 +906,13 @@ blake32_4way_close( blake_4way_small_context *ctx, unsigned ub, unsigned n,
   else
      ctx->T0 -= 512 - bit_len;

-   buf[vptr] = _mm_set1_epi32( 0x80 );
+   buf[vptr] = m128_const1_64( 0x0000008000000080 );

   if ( vptr < 12 )
   {
      memset_zero_128( buf + vptr + 1, 13 - vptr  );
-      buf[ 13 ] = _mm_or_si128( buf[ 13 ], _mm_set1_epi32( 0x01000000UL ) );
+      buf[ 13 ] = _mm_or_si128( buf[ 13 ],
+                                m128_const1_64( 0x0100000001000000ULL ) );
      buf[ 14 ] = mm128_bswap_32( _mm_set1_epi32( th ) );
      buf[ 15 ] = mm128_bswap_32( _mm_set1_epi32( tl ) );
      blake32_4way( ctx, buf + vptr, 64 - ptr );
@@ -790,20 +924,14 @@ blake32_4way_close( blake_4way_small_context *ctx, unsigned ub, unsigned n,
      ctx->T0 = 0xFFFFFE00UL;
      ctx->T1 = 0xFFFFFFFFUL;
      memset_zero_128( buf, 56>>2 );
-      buf[ 13 ] = _mm_or_si128( buf[ 13 ], _mm_set1_epi32( 0x01000000UL ) );
+      buf[ 13 ] = _mm_or_si128( buf[ 13 ],
+                                m128_const1_64( 0x0100000001000000ULL ) );
      buf[ 14 ] = mm128_bswap_32( _mm_set1_epi32( th ) );
      buf[ 15 ] = mm128_bswap_32( _mm_set1_epi32( tl ) );
      blake32_4way( ctx, buf, 64 );
   }

-   casti_m128i( dst, 0 ) = mm128_bswap_32( casti_m128i( ctx->H, 0 ) );
-   casti_m128i( dst, 1 ) = mm128_bswap_32( casti_m128i( ctx->H, 1 ) );
-   casti_m128i( dst, 2 ) = mm128_bswap_32( casti_m128i( ctx->H, 2 ) );
-   casti_m128i( dst, 3 ) = mm128_bswap_32( casti_m128i( ctx->H, 3 ) );
-   casti_m128i( dst, 4 ) = mm128_bswap_32( casti_m128i( ctx->H, 4 ) );
-   casti_m128i( dst, 5 ) = mm128_bswap_32( casti_m128i( ctx->H, 5 ) );
-   casti_m128i( dst, 6 ) = mm128_bswap_32( casti_m128i( ctx->H, 6 ) );
-   casti_m128i( dst, 7 ) = mm128_bswap_32( casti_m128i( ctx->H, 7 ) );
+   mm128_block_bswap_32( (__m128i*)dst, (__m128i*)ctx->H );
 }

 #if defined (__AVX2__)
@@ -816,11 +944,14 @@ static void
 blake32_8way_init( blake_8way_small_context *sc, const sph_u32 *iv,
                   const sph_u32 *salt, int rounds )
 {
-   int i;
-   for ( i = 0; i < 8; i++ )
-      sc->H[i] = _mm256_set1_epi32( iv[i] );
-   for ( i = 0; i < 4; i++ )
-      sc->S[i] = _mm256_set1_epi32( salt[i] );
+   casti_m256i( sc->H, 0 ) = m256_const1_64( 0x6A09E6676A09E667 );
+   casti_m256i( sc->H, 1 ) = m256_const1_64( 0xBB67AE85BB67AE85 );
+   casti_m256i( sc->H, 2 ) = m256_const1_64( 0x3C6EF3723C6EF372 );
+   casti_m256i( sc->H, 3 ) = m256_const1_64( 0xA54FF53AA54FF53A );
+   casti_m256i( sc->H, 4 ) = m256_const1_64( 0x510E527F510E527F );
+   casti_m256i( sc->H, 5 ) = m256_const1_64( 0x9B05688C9B05688C );
+   casti_m256i( sc->H, 6 ) = m256_const1_64( 0x1F83D9AB1F83D9AB );
+   casti_m256i( sc->H, 7 ) = m256_const1_64( 0x5BE0CD195BE0CD19 );
   sc->T0 = sc->T1 = 0;
   sc->ptr = 0;
   sc->rounds = rounds;
@@ -872,18 +1003,14 @@ static void
 blake32_8way_close( blake_8way_small_context *sc, unsigned ub, unsigned n,
                    void *dst, size_t out_size_w32 )
 {
-//   union {
-        __m256i buf[16];
-//        sph_u32 dummy;
-//   } u;
-   size_t ptr, k;
+   __m256i buf[16];
+   size_t ptr;
   unsigned bit_len;
   sph_u32 th, tl;
-   __m256i *out;

   ptr = sc->ptr;
   bit_len = ((unsigned)ptr << 3);
-   buf[ptr>>2] = _mm256_set1_epi32( 0x80 );
+   buf[ptr>>2] = m256_const1_64( 0x0000008000000080ULL );
   tl = sc->T0 + bit_len;
   th = sc->T1;

@@ -905,7 +1032,7 @@ blake32_8way_close( blake_8way_small_context *sc, unsigned ub, unsigned n,
       memset_zero_256( buf + (ptr>>2) + 1, (52 - ptr) >> 2 );
       if ( out_size_w32 == 8 )
           buf[52>>2] = _mm256_or_si256( buf[52>>2],
-                                           _mm256_set1_epi32( 0x01000000UL ) );
+                                m256_const1_64( 0x0100000001000000ULL ) );
       *(buf+(56>>2)) = mm256_bswap_32( _mm256_set1_epi32( th ) );
       *(buf+(60>>2)) = mm256_bswap_32( _mm256_set1_epi32( tl ) );
       blake32_8way( sc, buf + (ptr>>2), 64 - ptr );
@@ -918,18 +1045,189 @@ blake32_8way_close( blake_8way_small_context *sc, unsigned ub, unsigned n,
        sc->T1 = SPH_C32(0xFFFFFFFFUL);
        memset_zero_256( buf, 56>>2 );
       if ( out_size_w32 == 8 )
-           buf[52>>2] = _mm256_set1_epi32( 0x01000000UL );
+           buf[52>>2] = m256_const1_64( 0x0100000001000000ULL );
        *(buf+(56>>2)) = mm256_bswap_32( _mm256_set1_epi32( th ) );
        *(buf+(60>>2)) = mm256_bswap_32( _mm256_set1_epi32( tl ) );
        blake32_8way( sc, buf, 64 );
   }
-   out = (__m256i*)dst;
-   for ( k = 0; k < out_size_w32; k++ )
-        out[k] = mm256_bswap_32( sc->H[k] );
+   mm256_block_bswap_32( (__m256i*)dst, (__m256i*)sc->H );
 }

 #endif

+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+//Blake-256 16 way AVX512
+
+static void
+blake32_16way_init( blake_16way_small_context *sc, const sph_u32 *iv,
+                   const sph_u32 *salt, int rounds )
+{
+   casti_m512i( sc->H, 0 ) = m512_const1_64( 0x6A09E6676A09E667 );
+   casti_m512i( sc->H, 1 ) = m512_const1_64( 0xBB67AE85BB67AE85 );
+   casti_m512i( sc->H, 2 ) = m512_const1_64( 0x3C6EF3723C6EF372 );
+   casti_m512i( sc->H, 3 ) = m512_const1_64( 0xA54FF53AA54FF53A );
+   casti_m512i( sc->H, 4 ) = m512_const1_64( 0x510E527F510E527F );
+   casti_m512i( sc->H, 5 ) = m512_const1_64( 0x9B05688C9B05688C );
+   casti_m512i( sc->H, 6 ) = m512_const1_64( 0x1F83D9AB1F83D9AB );
+   casti_m512i( sc->H, 7 ) = m512_const1_64( 0x5BE0CD195BE0CD19 );
+   sc->T0 = sc->T1 = 0;
+   sc->ptr = 0;
+   sc->rounds = rounds;
+}
+
+static void
+blake32_16way( blake_16way_small_context *sc, const void *data, size_t len )
+{
+   __m512i *vdata = (__m512i*)data;
+   __m512i *buf;
+   size_t ptr;
+   const int buf_size = 64;   // number of elements, sizeof/4
+   DECL_STATE32_16WAY
+   buf = sc->buf;
+   ptr = sc->ptr;
+   if ( len < buf_size - ptr )
+   {
+        memcpy_512( buf + (ptr>>2), vdata, len>>2 );
+        ptr += len;
+        sc->ptr = ptr;
+        return;
+   }
+   READ_STATE32_16WAY(sc);
+   while ( len > 0 )
+   {
+      size_t clen;
+
+      clen = buf_size - ptr;
+      if (clen > len)
+           clen = len;
+      memcpy_512( buf + (ptr>>2), vdata, clen>>2 );
+      ptr += clen;
+      vdata += (clen>>2);
+      len -= clen;
+      if ( ptr == buf_size )
+      {
+          if ( ( T0 = T0 + 512 ) < 512 )
+                T1 = T1 + 1;
+          COMPRESS32_16WAY( sc->rounds );
+          ptr = 0;
+      }
+   }
+   WRITE_STATE32_16WAY(sc);
+   sc->ptr = ptr;
+}
+
+static void
+blake32_16way_close( blake_16way_small_context *sc, unsigned ub, unsigned n,
+                    void *dst, size_t out_size_w32 )
+{
+   __m512i buf[16];
+   size_t ptr;
+   unsigned bit_len;
+   sph_u32 th, tl;
+
+   ptr = sc->ptr;
+   bit_len = ((unsigned)ptr << 3);
+   buf[ptr>>2] = m512_const1_64( 0x0000008000000080ULL );
+   tl = sc->T0 + bit_len;
+   th = sc->T1;
+
+   if ( ptr == 0 )
+   {
+        sc->T0 = 0xFFFFFE00UL;
+        sc->T1 = 0xFFFFFFFFUL;
+   }
+   else if ( sc->T0 == 0 )
+   {
+        sc->T0 = 0xFFFFFE00UL + bit_len;
+        sc->T1 = sc->T1 - 1;
+   }
+   else
+        sc->T0 -= 512 - bit_len;
+
+   if ( ptr <= 52 )
+   {
+       memset_zero_512( buf + (ptr>>2) + 1, (52 - ptr) >> 2 );
+       if ( out_size_w32 == 8 )
+           buf[52>>2] = _mm512_or_si512( buf[52>>2],
+                                m512_const1_64( 0x0100000001000000ULL ) );
+       buf[+56>>2] = mm512_bswap_32( _mm512_set1_epi32( th ) );
+       buf[+60>>2] = mm512_bswap_32( _mm512_set1_epi32( tl ) );
+       blake32_16way( sc, buf + (ptr>>2), 64 - ptr );
+   }
+   else
+   {
+        memset_zero_512( buf + (ptr>>2) + 1, (60-ptr) >> 2 );
+        blake32_16way( sc, buf + (ptr>>2), 64 - ptr );
+        sc->T0 = 0xFFFFFE00UL;
+        sc->T1 = 0xFFFFFFFFUL;
+        memset_zero_512( buf, 56>>2 );
+       if ( out_size_w32 == 8 )
+           buf[52>>2] = m512_const1_64( 0x0100000001000000ULL );
+        buf[56>>2] = mm512_bswap_32( _mm512_set1_epi32( th ) );
+        buf[60>>2] = mm512_bswap_32( _mm512_set1_epi32( tl ) );
+        blake32_16way( sc, buf, 64 );
+   }
+   mm512_block_bswap_32( (__m512i*)dst, (__m512i*)sc->H );
+}
+
+void
+blake256_16way_init(void *cc)
+{
+   blake32_16way_init( cc, IV256, salt_zero_8way_small, 14 );
+}
+
+void
+blake256_16way_update(void *cc, const void *data, size_t len)
+{
+        blake32_16way(cc, data, len);
+}
+
+void
+blake256_16way_close(void *cc, void *dst)
+{
+        blake32_16way_close(cc, 0, 0, dst, 8);
+}
+
+void blake256r14_16way_init(void *cc)
+{
+   blake32_16way_init( cc, IV256, salt_zero_8way_small, 14 );
+}
+
+void
+blake256r14_16way_update(void *cc, const void *data, size_t len)
+{
+   blake32_16way(cc, data, len);
+}
+
+void
+blake256r14_16way_close(void *cc, void *dst)
+{
+   blake32_16way_close(cc, 0, 0, dst, 8);
+}
+
+void blake256r8_16way_init(void *cc)
+{
+   blake32_16way_init( cc, IV256, salt_zero_8way_small, 8 );
+}
+
+void
+blake256r8_16way_update(void *cc, const void *data, size_t len)
+{
+   blake32_16way(cc, data, len);
+}
+
+void
+blake256r8_16way_close(void *cc, void *dst)
+{
+   blake32_16way_close(cc, 0, 0, dst, 8);
+}
+
+#endif // AVX512
+
+
+
 // Blake-256 4 way

 // default 14 rounds, backward copatibility
@@ -940,7 +1238,7 @@ blake256_4way_init(void *ctx)
 }

 void
-blake256_4way(void *ctx, const void *data, size_t len)
+blake256_4way_update(void *ctx, const void *data, size_t len)
 {
 	blake32_4way(ctx, data, len);
 }
@@ -962,7 +1260,7 @@ blake256_8way_init(void *cc)
 }

 void
-blake256_8way(void *cc, const void *data, size_t len)
+blake256_8way_update(void *cc, const void *data, size_t len)
 {
        blake32_8way(cc, data, len);
 }
@@ -982,7 +1280,7 @@ void blake256r14_4way_init(void *cc)
 }

 void
-blake256r14_4way(void *cc, const void *data, size_t len)
+blake256r14_4way_update(void *cc, const void *data, size_t len)
 {
   blake32_4way(cc, data, len);
 }
@@ -1001,7 +1299,7 @@ void blake256r14_8way_init(void *cc)
 }

 void
-blake256r14_8way(void *cc, const void *data, size_t len)
+blake256r14_8way_update(void *cc, const void *data, size_t len)
 {
   blake32_8way(cc, data, len);
 }
@@ -1021,7 +1319,7 @@ void blake256r8_4way_init(void *cc)
 }

 void
-blake256r8_4way(void *cc, const void *data, size_t len)
+blake256r8_4way_update(void *cc, const void *data, size_t len)
 {
   blake32_4way(cc, data, len);
 }
@@ -1040,7 +1338,7 @@ void blake256r8_8way_init(void *cc)
 }

 void
-blake256r8_8way(void *cc, const void *data, size_t len)
+blake256r8_8way_update(void *cc, const void *data, size_t len)
 {
   blake32_8way(cc, data, len);
 }
--- a/algo/blake/blake256-hash-4way.c.new
+++ b/algo/blake/blake256-hash-4way.c.new
@@ -1,322 +0,0 @@
-// convert blake256 32 bit to use 64 bit with serial vectoring
-//
-//  cut calls to GS in half
-//
-// combine V
-// v0 = {V0,V1}
-// v1 = {V2,V3}
-// v2 = {V4,V5}
-// v3 = {V6,V7}
-// v4 = {V8,V9}
-// v5 = {VA,VB}
-// v6 = {VC,VD}
-// v7 = {CE,VF}
-//
-// v6x = {VD,VC}      swap(VC,VD)   swap(v6)
-// v7x = {VF,VE}      swap(VE,VF)   swap(v7)
-//
-// V0 = v1v0
-// V1 = v3v2
-// V2 = v5v4
-// V3 = v7v6
-// V4 = v9v8
-// V5 = vbva
-// V6 = vdvc
-// V7 = vfve
-//
-// The rotate in ROUND is to effect straddle and unstraddle for the third
-// and 4th iteration of GS.
-// It concatenates 2 contiguous 256 bit vectors and extracts the middle
-// 256 bits. After the transform they must be restored with only the
-// chosen bits modified in the original 2 vectors.
-// ror1x128 achieves this by putting the chosen bits in arg1, the "low"
-// 256 bit vector and saves the untouched bits temporailly in arg0, the
-// "high" 256 bit vector. Simply reverse the process to restore data back
-// to original positions.
-
-// Use standard 4way when AVX2 is not available use x2 mode with AVX2.
-//
-// Data is organised the same as 32 bit 4 way, in effect serial vectoring
-// on top of parallel vectoring. Same data in the same place just taking
-// two chunks at a time.
-//
-// Transparent to user, x2 mode used when AVX2 detected.
-// Use existing 4way context but revert to scalar types.
-// Same interleave function (128 bit) or x2 with 256 bit?
-// User trsnaparency would have to apply to interleave as well.
-//
-// Use common 4way update and close
-
-/*
-typedef struct {
-   unsigned char buf[64<<2];
-   uint32_t H[8<<2];
-   uint32_t S[4<<2];
-   size_t ptr;
-   uint32_t T0, T1;
-   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
-} blakex2_4way_small_context __attribute__ ((aligned (64)));
-*/
-
-static void
-blake32x2_4way_init( blake_4way_small_context *ctx, const uint32_t *iv,
-                   const uint32_t *salt, int rounds )
-{
-   casti_m128i( ctx->H, 0 ) = _mm_set1_epi32( iv[0] );
-   casti_m128i( ctx->H, 1 ) = _mm_set1_epi32( iv[1] );
-   casti_m128i( ctx->H, 2 ) = _mm_set1_epi32( iv[2] );
-   casti_m128i( ctx->H, 3 ) = _mm_set1_epi32( iv[3] );
-   casti_m128i( ctx->H, 4 ) = _mm_set1_epi32( iv[4] );
-   casti_m128i( ctx->H, 5 ) = _mm_set1_epi32( iv[5] );
-   casti_m128i( ctx->H, 6 ) = _mm_set1_epi32( iv[6] );
-   casti_m128i( ctx->H, 7 ) = _mm_set1_epi32( iv[7] );
-
-   casti_m128i( ctx->S, 0 ) = m128_zero;
-   casti_m128i( ctx->S, 1 ) = m128_zero;
-   casti_m128i( ctx->S, 2 ) = m128_zero;
-   casti_m128i( ctx->S, 3 ) = m128_zero;
-/*
-   sc->S[0] = _mm_set1_epi32( salt[0] );
-   sc->S[1] = _mm_set1_epi32( salt[1] );
-   sc->S[2] = _mm_set1_epi32( salt[2] );
-   sc->S[3] = _mm_set1_epi32( salt[3] );
-*/
-   ctx->T0 = ctx->T1 = 0;
-   ctx->ptr = 0;
-   ctx->rounds = rounds;
-}
-
-static void
-blake32x2( blake_4way_small_context *ctx, const void *data, size_t len )
-{
-   __m128i *buf = (__m256i*)ctx->buf;
-   size_t  bptr = ctx->ptr << 2;
-   size_t  vptr = ctx->ptr >> 3;
-   size_t  blen = len << 2;
-//    unsigned char *buf = ctx->buf;
-//    size_t ptr         = ctx->ptr<<4;  // repurposed
-    DECL_STATE32x2
-
-//    buf = sc->buf;
-//    ptr = sc->ptr;
-
-// adjust len for use with ptr, clen, all absolute bytes.
-//    int blen = len<<2;
-
-    if ( blen < (sizeof ctx->buf) - bptr )
-    {
-        memcpy( buf + vptr, data, blen );
-        ptr += blen;
-        ctx->ptr = bptr >> 2;;
-        return;
-    }
-
-    READ_STATE32( ctx );
-    while ( blen > 0 )
-    {
-        size_t clen;
-
-        clen = ( sizeof sc->buf ) - ptr;
-        if ( clen > blen )
-            clen = blen;
-        memcpy( buf + vptr, data, clen );
-        bptr += clen;
-        vptr = bptr >> 5;
-	data = (const unsigned char *)data + clen;
-        blen -= clen;
-        if ( bptr == sizeof ctx->buf )
-       	{
-           if ( ( T0 = T0 + 512 ) < 512 ) // not needed, will never rollover
-               T1 += 1;
-           COMPRESS32x2_4WAY( ctx->rounds );
-           ptr = 0;
-        }
-    }
-    WRITE_STATE32x2( ctx );
-    ctx->ptr = bptr >> 2;
-}
-
-static void
-blake32x2_4way_close( blake_4way_small_context *ctx, void *dst )
-{
-   __m256i buf[8] __attribute__ ((aligned (64)));
-   size_t   ptr     = ctx->ptr;
-   size_t   vptr    = ctx->ptr>>2;
-   unsigned bit_len = ( (unsigned)ptr << 3 );  // one lane
-   uint32_t th      = ctx->T1;
-   uint32_t tl      = ctx->T0 + bit_len;
-
-   if ( ptr == 0 )
-   {
-        ctx->T0 = 0xFFFFFE00UL;
-        ctx->T1 = 0xFFFFFFFFUL;
-   }
-   else if ( ctx->T0 == 0 )
-   {
-      ctx->T0 = 0xFFFFFE00UL + bit_len;
-      ctx->T1 -= 1;
-   }
-   else
-      ctx->T0 -= 512 - bit_len;
-
-   // memset doesn't do ints
-   buf[ vptr ] = _mm256_set_epi32( 0,0,0,0, 0x80, 0x80, 0x80, 0x80 );
-
-   if ( vptr < 5 )
-   {
-       memset_zero_256( buf + vptr + 1, 6 - vptr  );
-       buf[ 6 ] = _mm256_or_si256( vbuf[ 6 ], _mm256_set_epi32(
-             0x01000000UL,0x01000000UL,0x01000000UL,0x01000000UL, 0,0,0,0 ) ); 
-       buf[ 7 ] = mm256_bswap_32( _mm256_set_epi32( tl,tl,tl,tl,
-			                            th,th,th,th ) );
-       blake32x2_4way( ctx, buf + vptr, 64 - ptr );
-   }
-   else
-   {
-       memset_zero_256( vbuf + vptr + 1, 7 - vptr );
-       blake32x2_4way( ctx,  vbuf + ptr, 64 - ptr );
-       ctx->T0 = 0xFFFFFE00UL;
-       ctx->T1 = 0xFFFFFFFFUL;
-       buf[ 6 ] = mm256_zero;
-       buf[ 6 ] = _mm256_set_epi32( 0,0,0,0,
-		         0x01000000UL,0x01000000UL,0x01000000UL,0x01000000UL );
-       buf[ 7 ] = mm256_bswap_32( _mm256_set_epi32( tl, tl, tl, tl,
-                                                    th, th, th, th );
-       blake32x2_4way( ctx, buf, 64 );
-   }
-
-   casti_m256i( dst, 0 ) = mm256_bswap_32( casti_m256i( ctx->H, 0 ) );
-   casti_m256i( dst, 1 ) = mm256_bswap_32( casti_m256i( ctx->H, 1 ) );
-   casti_m256i( dst, 2 ) = mm256_bswap_32( casti_m256i( ctx->H, 2 ) );
-   casti_m256i( dst, 3 ) = mm256_bswap_32( casti_m256i( ctx->H, 3 ) );
-}
-
-
-
-
-#define DECL_STATE32x2_4WAY \
-   __m256i H0, H1, H2, H3; \
-   __m256i S0, S1; \
-   uint32_t T0, T1;
-
-#define READ_STATE32x2_4WAY(state)  do \
-{ \
-   H0 = casti_m256i( state->H, 0 ); \
-   H1 = casti_m256i( state->H, 1 ); \
-   H2 = casti_m256i( state->H, 2 ); \
-   H3 = casti_m256i( state->H, 3 ); \
-   S0 = casti_m256i( state->S, 0 ); \
-   S1 = casti_m256i( state->S, 1 ); \
-   T0 = state->T0; \
-   T1 = state->T1; \
-
-#define WRITE_STATE32x2_4WAY(state)   do { \
-   casti_m256i( state->H, 0 ) = H0; \
-   casti_m256i( state->H, 1 ) = H1; \
-   casti_m256i( state->H, 2 ) = H2; \
-   casti_m256i( state->H, 3 ) = H3; \
-   casti_m256i( state->S, 0 ) = S0; \
-   casti_m256i( state->S, 1 ) = S1; \
-   state->T0 = T0; \
-   state->T1 = T1; \
-} while (0)
-
-
-#define GSx2_4WAY( m0m2, m1m3, c0c2, c1c3, a, b, c, d ) do \
-{ \
-   a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
-          _mm256_set_epi32( c1,c3, c1,c3, c1,c3, c1,c3 ), \
-	  _mm256_set_epi32( m0,m2, m0,m2, m0,m2, m0,m2 ) ), b ), a ); \
-   d = mm256_ror_32( _mm_xor_si128( d, a ), 16 ); \
-   c = _mm256_add_epi32( c, d ); \
-   b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
-   a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
-          _mm256_set_epi32( c0,c2, c0,c2, c0,c2, c0,c2 ), \
-	  _mm256_set_epi32( m1,m3, m1,m3, m1,m3, m1,m3 ) ), b ), a ); \
-   d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
-   c = _mm256_add_epi32( c, d ); \
-   b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
-} while (0)
-
-#define ROUND_Sx2_4WAY(r)   do \
-{ \
-  GS2_4WAY( Mx(r, 0),  Mx(r, 1),  Mx(r, 2),  Mx(r, 3), \
-           CSx(r, 0), CSx(r, 1), CSx(r, 2), CSx(r, 3), V0, V2, V4, V6 ); \
-  GS2_4WAY( Mx(r, 4),  Mx(r, 5),  Mx(r, 6),  Mx(r, 7), \
-           CSx(r, 4), CSx(r, 5), CSx(r, 6), CSx(r, 7), V1, V3, V5, V7 ); \
-  mm256_ror1x128_512( V3, V2 ); \
-  mm256_ror1x128_512( V6, V7 ); \
-  GS2_4WAY( Mx(r, 8),  Mx(r, 9),  Mx(r, A),  Mx(r, B), \
-           CSx(r, 8), CSx(r, 9), CSx(r, A), CSx(r, B), V0, V2, V5, V7 ); \
-  GS2_4WAY( Mx(r, C),  Mx(r, D),  Mx(r, C),  Mx(r, D), \
-           CSx(r, C), CSx(r, D), CSx(r, C), CSx(r, D), V1, V3, V4, V6 ); \
-  mm256_rol1x128_512( V2, V3 ); \
-  mm256_rol1x128_512( V7, V6 ); 
-
-#define COMPRESS32x2_4WAY( rounds ) do \
-{ \
-   __m256i M0, M1, M2, M3, M4, M5, M6, M7; \
-   __m256i V0, V1, V2, V3, V4, V5, V6, V7; \
-   unsigned r; \
-   V0 = H0; \
-   V1 = H1; \
-   V2 = H2; \
-   V3 = H3; \
-   V4 = _mm256_xor_si256( S0, _mm256_set_epi32( CS1, CS1, CS1, CS1, \
-			                        CS0, CS0, CS0, CS0 ) ); \
-   V5 = _mm256_xor_si256( S1, _mm256_set_epi32( CS3, CS3, CS3, CS3, \
-                                                CS2, CS2, CS2, CS2 ) ); \
-   V6 = _mm256_xor_si256( _mm256_set1_epi32( T0 ), \
-                              _mm256_set_epi32( CS5, CS5, CS5, CS5, \
-		                                CS4, CS4, CS4, CS4 ) ); \
-   V7 = _mm256_xor_si256( _mm256_set1_epi32( T1 ), \
-                              _mm256_set_epi32( CS7, CS7, CS7, CS7, \
-                                                CS6, CS6, CS6, CS6 ) ); \
-   M0 = mm256_bswap_32( buf[ 0] ); \
-   M1 = mm256_bswap_32( buf[ 1] ); \
-   M2 = mm256_bswap_32( buf[ 2] ); \
-   M3 = mm256_bswap_32( buf[ 3] ); \
-   M4 = mm256_bswap_32( buf[ 4] ); \
-   M5 = mm256_bswap_32( buf[ 5] ); \
-   M6 = mm256_bswap_32( buf[ 6] ); \
-   M7 = mm256_bswap_32( buf[ 7] ); \
-   ROUND_Sx2_4WAY(0); \
-   ROUND_Sx2_4WAY(1); \
-   ROUND_Sx2_4WAY(2); \
-   ROUND_Sx2_4WAY(3); \
-   ROUND_Sx2_4WAY(4); \
-   ROUND_Sx2_4WAY(5); \
-   ROUND_Sx2_4WAY(6); \
-   ROUND_Sx2_4WAY(7); \
-   if (rounds == 14) \
-   { \
-      ROUND_Sx2_4WAY(8); \
-      ROUND_Sx2_4WAY(9); \
-      ROUND_Sx2_4WAY(0); \
-      ROUND_Sx2_4WAY(1); \
-      ROUND_Sx2_4WAY(2); \
-      ROUND_Sx2_4WAY(3); \
-   } \
-   H0 = _mm256_xor_si256( _mm256_xor_si256( \
-			           _mm256_xor_si256( V8, V0 ), S0 ), H0 ); \
-   H1 = _mm256_xor_si256( _mm256_xor_si256( \
-			           _mm256_xor_si256( V9, V1 ), S1 ), H1 ); \
-   H2 = _mm256_xor_si256( _mm256_xor_si256( \
-			           _mm256_xor_si256( VA, V2 ), S2 ), H2 ); \
-   H3 = _mm256_xor_si256( _mm256_xor_si256( \
-			           _mm256_xor_si256( VB, V3 ), S3 ), H3 ); \
-} while (0)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
--- a/algo/blake/blake2b-4way.c
+++ b/algo/blake/blake2b-4way.c
@@ -0,0 +1,113 @@
+/**
+ * Blake2-B Implementation
+ * tpruvot@github 2015-2016
+ */
+
+#include "blake2b-gate.h"
+#include <string.h>
+#include <stdint.h>
+#include "blake2b-hash-4way.h"
+
+#if defined(BLAKE2B_8WAY)
+
+int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
+                           uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t hash[8*8] __attribute__ ((aligned (128)));;
+   uint32_t vdata[20*8] __attribute__ ((aligned (64)));;
+   uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+   blake2b_8way_ctx ctx __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[49]);   // 3*16+1
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   int thr_id = mythr->id;
+   __m512i  *noncev = (__m512i*)vdata + 9;   // aligned
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+
+   uint32_t n = first_nonce;
+
+   mm512_bswap32_intrlv80_8x64( vdata, pdata );
+
+   do {
+      *noncev = mm512_intrlv_blend_32( mm512_bswap_32(
+                _mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
+                                  n+3, 0, n+2, 0, n+1, 0, n  , 0 ) ), *noncev );
+
+      blake2b_8way_init( &ctx );
+      blake2b_8way_update( &ctx, vdata, 80 );
+      blake2b_8way_final( &ctx, hash );
+
+      for ( int lane = 0; lane < 8; lane++ )
+      if ( hash7[ lane<<1 ] < Htarg )
+      {
+          extr_lane_8x64( lane_hash, hash, lane, 256 );
+          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+          }
+      }
+      n += 8;
+   } while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#elif defined(BLAKE2B_4WAY)
+
+// Function not used, code inlined.
+void blake2b_4way_hash(void *output, const void *input)
+{
+    blake2b_4way_ctx ctx;
+    blake2b_4way_init( &ctx );
+    blake2b_4way_update( &ctx, input, 80 );
+    blake2b_4way_final( &ctx, output );
+}
+
+int scanhash_blake2b_4way( struct work *work, uint32_t max_nonce,
+                           uint64_t *hashes_done, struct thr_info *mythr )
+{
+	uint32_t hash[8*4] __attribute__ ((aligned (64)));;
+   uint32_t vdata[20*4] __attribute__ ((aligned (32)));;
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   blake2b_4way_ctx ctx __attribute__ ((aligned (32)));
+   uint32_t *hash7 = &(hash[25]);   // 3*8+1
+	uint32_t *pdata = work->data;
+	uint32_t *ptarget = work->target;
+   int thr_id = mythr->id;
+   __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+	const uint32_t Htarg = ptarget[7];
+	const uint32_t first_nonce = pdata[19];
+
+	uint32_t n = first_nonce;
+
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
+
+	do {
+      *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
+
+      blake2b_4way_init( &ctx ); 
+      blake2b_4way_update( &ctx, vdata, 80 );
+      blake2b_4way_final( &ctx, hash );
+
+      for ( int lane = 0; lane < 4; lane++ )
+      if ( hash7[ lane<<1 ] < Htarg )
+      {
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
+          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+          }
+      }
+      n += 4;
+   } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#endif
--- a/algo/blake/blake2b-gate.c
+++ b/algo/blake/blake2b-gate.c
@@ -0,0 +1,20 @@
+#include "blake2b-gate.h"
+
+
+bool register_blake2b_algo( algo_gate_t* gate )
+{
+#if defined(BLAKE2B_8WAY)
+  gate->scanhash  = (void*)&scanhash_blake2b_8way;
+//  gate->hash      = (void*)&blake2b_8way_hash;
+#elif defined(BLAKE2B_4WAY)
+  gate->scanhash  = (void*)&scanhash_blake2b_4way;
+  gate->hash      = (void*)&blake2b_4way_hash;
+#else
+  gate->scanhash  = (void*)&scanhash_blake2b;
+  gate->hash      = (void*)&blake2b_hash;
+#endif
+  gate->optimizations =  AVX2_OPT | AVX512_OPT;
+  return true;
+};
+
+
--- a/algo/blake/blake2b-gate.h
+++ b/algo/blake/blake2b-gate.h
@@ -0,0 +1,34 @@
+#ifndef __BLAKE2B_GATE_H__
+#define __BLAKE2B_GATE_H__ 1
+
+#include <stdint.h>
+#include "algo-gate-api.h"
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define BLAKE2B_8WAY
+#elif defined(__AVX2__)
+  #define BLAKE2B_4WAY
+#endif
+
+bool register_blake2b_algo( algo_gate_t* gate );
+
+#if defined(BLAKE2B_8WAY)
+
+//void blake2b_8way_hash( void *state, const void *input );
+int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+#elif defined(BLAKE2B_4WAY)
+
+void blake2b_4way_hash( void *state, const void *input );
+int scanhash_blake2b_4way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+#else
+
+void blake2b_hash( void *state, const void *input );
+int scanhash_blake2b( struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr );
+
+#endif
+
+#endif
--- a/algo/blake/blake2b-hash-4way.c
+++ b/algo/blake/blake2b-hash-4way.c
@@ -0,0 +1,372 @@
+/*
+ * Copyright 2009 Colin Percival, 2014 savale
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * This file was originally written by Colin Percival as part of the Tarsnap
+ * online backup system.
+ */
+
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+
+#include "blake2b-hash-4way.h"
+
+static const uint8_t sigma[12][16] =
+{
+      { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
+      { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
+      { 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
+      { 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
+      { 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
+      { 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
+      { 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
+      { 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
+      { 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
+      { 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0 },
+      { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
+      { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 }
+};
+
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define B2B8W_G(a, b, c, d, x, y) \
+{ \
+   v[a] = _mm512_add_epi64( _mm512_add_epi64( v[a], v[b] ), x ); \
+   v[d] = mm512_ror_64( _mm512_xor_si512( v[d], v[a] ), 32 ); \
+   v[c] = _mm512_add_epi64( v[c], v[d] ); \
+   v[b] = mm512_ror_64( _mm512_xor_si512( v[b], v[c] ), 24 ); \
+   v[a] = _mm512_add_epi64( _mm512_add_epi64( v[a], v[b] ), y ); \
+   v[d] = mm512_ror_64( _mm512_xor_si512( v[d], v[a] ), 16 ); \
+   v[c] = _mm512_add_epi64( v[c], v[d] ); \
+   v[b] = mm512_ror_64( _mm512_xor_si512( v[b], v[c] ), 63 ); \
+}
+
+static void blake2b_8way_compress( blake2b_8way_ctx *ctx, int last )
+{  
+   __m512i v[16], m[16];
+
+   v[ 0] = ctx->h[0];
+   v[ 1] = ctx->h[1];
+   v[ 2] = ctx->h[2];
+   v[ 3] = ctx->h[3];
+   v[ 4] = ctx->h[4];
+   v[ 5] = ctx->h[5];
+   v[ 6] = ctx->h[6];
+   v[ 7] = ctx->h[7];
+   v[ 8] = m512_const1_64( 0x6A09E667F3BCC908 );
+   v[ 9] = m512_const1_64( 0xBB67AE8584CAA73B );
+   v[10] = m512_const1_64( 0x3C6EF372FE94F82B );
+   v[11] = m512_const1_64( 0xA54FF53A5F1D36F1 );
+   v[12] = m512_const1_64( 0x510E527FADE682D1 );
+   v[13] = m512_const1_64( 0x9B05688C2B3E6C1F );
+   v[14] = m512_const1_64( 0x1F83D9ABFB41BD6B );
+   v[15] = m512_const1_64( 0x5BE0CD19137E2179 );
+
+   v[12] = _mm512_xor_si512( v[12], _mm512_set1_epi64( ctx->t[0] ) );
+   v[13] = _mm512_xor_si512( v[13], _mm512_set1_epi64( ctx->t[1] ) );
+
+   if ( last )
+      v[14] = mm512_not( v[14] );
+
+   m[ 0] = ctx->b[ 0];
+   m[ 1] = ctx->b[ 1];
+   m[ 2] = ctx->b[ 2];
+   m[ 3] = ctx->b[ 3];
+   m[ 4] = ctx->b[ 4];
+   m[ 5] = ctx->b[ 5];
+   m[ 6] = ctx->b[ 6];
+   m[ 7] = ctx->b[ 7];
+   m[ 8] = ctx->b[ 8];
+   m[ 9] = ctx->b[ 9];
+   m[10] = ctx->b[10];
+   m[11] = ctx->b[11];
+   m[12] = ctx->b[12];
+   m[13] = ctx->b[13];
+   m[14] = ctx->b[14];
+   m[15] = ctx->b[15];
+
+   for ( int i = 0; i < 12; i++ )
+   {
+      B2B8W_G( 0, 4,  8, 12, m[ sigma[i][ 0] ], m[ sigma[i][ 1] ] );
+      B2B8W_G( 1, 5,  9, 13, m[ sigma[i][ 2] ], m[ sigma[i][ 3] ] );
+      B2B8W_G( 2, 6, 10, 14, m[ sigma[i][ 4] ], m[ sigma[i][ 5] ] );
+      B2B8W_G( 3, 7, 11, 15, m[ sigma[i][ 6] ], m[ sigma[i][ 7] ] );
+      B2B8W_G( 0, 5, 10, 15, m[ sigma[i][ 8] ], m[ sigma[i][ 9] ] );
+      B2B8W_G( 1, 6, 11, 12, m[ sigma[i][10] ], m[ sigma[i][11] ] );
+      B2B8W_G( 2, 7,  8, 13, m[ sigma[i][12] ], m[ sigma[i][13] ] );
+      B2B8W_G( 3, 4,  9, 14, m[ sigma[i][14] ], m[ sigma[i][15] ] );
+   }
+
+   ctx->h[0] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[0], v[0] ), v[ 8] );
+   ctx->h[1] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[1], v[1] ), v[ 9] );
+   ctx->h[2] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[2], v[2] ), v[10] );
+   ctx->h[3] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[3], v[3] ), v[11] );
+   ctx->h[4] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[4], v[4] ), v[12] );
+   ctx->h[5] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[5], v[5] ), v[13] );
+   ctx->h[6] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[6], v[6] ), v[14] );
+   ctx->h[7] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[7], v[7] ), v[15] );
+}
+
+int blake2b_8way_init( blake2b_8way_ctx *ctx )
+{
+   size_t i;
+
+   ctx->h[0] = m512_const1_64( 0x6A09E667F3BCC908 );
+   ctx->h[1] = m512_const1_64( 0xBB67AE8584CAA73B );
+   ctx->h[2] = m512_const1_64( 0x3C6EF372FE94F82B );
+   ctx->h[3] = m512_const1_64( 0xA54FF53A5F1D36F1 );
+   ctx->h[4] = m512_const1_64( 0x510E527FADE682D1 );
+   ctx->h[5] = m512_const1_64( 0x9B05688C2B3E6C1F );
+   ctx->h[6] = m512_const1_64( 0x1F83D9ABFB41BD6B );
+   ctx->h[7] = m512_const1_64( 0x5BE0CD19137E2179 );
+
+   ctx->h[0] = _mm512_xor_si512( ctx->h[0], m512_const1_64( 0x01010020 ) );
+
+   ctx->t[0] = 0;
+   ctx->t[1] = 0;
+   ctx->c = 0;
+   ctx->outlen = 32;
+
+   for ( i = 0; i < 16; i++ )
+     ctx->b[i] = m512_zero;
+
+   return 0;
+}
+
+
+void blake2b_8way_update( blake2b_8way_ctx *ctx, const void *input,
+                          size_t inlen )
+{
+   __m512i* in =(__m512i*)input;
+
+   size_t i, c;
+   c = ctx->c >> 3;
+
+   for ( i = 0; i < (inlen >> 3); i++ )
+   {
+      if ( ctx->c == 128 )
+      {
+         ctx->t[0] += ctx->c;
+         if ( ctx->t[0] < ctx->c )
+            ctx->t[1]++;
+         blake2b_8way_compress( ctx, 0 );
+         ctx->c = 0;
+      }
+      ctx->b[ c++ ] = in[i];
+      ctx->c += 8;
+   }
+}
+
+void blake2b_8way_final( blake2b_8way_ctx *ctx, void *out )
+{
+   size_t c;
+   c = ctx->c >> 3;
+
+   ctx->t[0] += ctx->c;
+   if ( ctx->t[0] < ctx->c )
+      ctx->t[1]++;
+
+   while ( ctx->c < 128 )
+   {
+      ctx->b[c++] = m512_zero;
+      ctx->c += 8;
+   }
+
+   blake2b_8way_compress( ctx, 1 );           // final block flag = 1
+
+   casti_m512i( out, 0 ) = ctx->h[0];
+   casti_m512i( out, 1 ) = ctx->h[1];
+   casti_m512i( out, 2 ) = ctx->h[2];
+   casti_m512i( out, 3 ) = ctx->h[3];
+}
+
+#endif
+
+#if defined(__AVX2__)
+
+// G Mixing function.
+
+#define B2B_G(a, b, c, d, x, y) \
+{ \
+   v[a] = _mm256_add_epi64( _mm256_add_epi64( v[a], v[b] ), x ); \
+	v[d] = mm256_ror_64( _mm256_xor_si256( v[d], v[a] ), 32 ); \
+	v[c] = _mm256_add_epi64( v[c], v[d] ); \
+	v[b] = mm256_ror_64( _mm256_xor_si256( v[b], v[c] ), 24 ); \
+	v[a] = _mm256_add_epi64( _mm256_add_epi64( v[a], v[b] ), y ); \
+	v[d] = mm256_ror_64( _mm256_xor_si256( v[d], v[a] ), 16 ); \
+	v[c] = _mm256_add_epi64( v[c], v[d] ); \
+	v[b] = mm256_ror_64( _mm256_xor_si256( v[b], v[c] ), 63 ); \
+}
+
+// Initialization Vector.
+/*
+static const uint64_t blake2b_iv[8] = {
+	0x6A09E667F3BCC908, 0xBB67AE8584CAA73B,
+	0x3C6EF372FE94F82B, 0xA54FF53A5F1D36F1,
+	0x510E527FADE682D1, 0x9B05688C2B3E6C1F,
+	0x1F83D9ABFB41BD6B, 0x5BE0CD19137E2179
+};
+*/
+
+static void blake2b_4way_compress( blake2b_4way_ctx *ctx, int last )
+{
+	__m256i v[16], m[16];
+
+   v[ 0] = ctx->h[0];
+   v[ 1] = ctx->h[1];
+   v[ 2] = ctx->h[2];
+   v[ 3] = ctx->h[3];
+   v[ 4] = ctx->h[4];
+   v[ 5] = ctx->h[5];
+   v[ 6] = ctx->h[6];
+   v[ 7] = ctx->h[7];
+   v[ 8] = m256_const1_64( 0x6A09E667F3BCC908 );
+   v[ 9] = m256_const1_64( 0xBB67AE8584CAA73B );
+   v[10] = m256_const1_64( 0x3C6EF372FE94F82B );
+   v[11] = m256_const1_64( 0xA54FF53A5F1D36F1 );
+   v[12] = m256_const1_64( 0x510E527FADE682D1 );
+   v[13] = m256_const1_64( 0x9B05688C2B3E6C1F );
+   v[14] = m256_const1_64( 0x1F83D9ABFB41BD6B );
+   v[15] = m256_const1_64( 0x5BE0CD19137E2179 );
+
+   v[12] = _mm256_xor_si256( v[12], _mm256_set1_epi64x( ctx->t[0] ) );
+   v[13] = _mm256_xor_si256( v[13], _mm256_set1_epi64x( ctx->t[1] ) );
+
+   if ( last )   
+		v[14] = mm256_not( v[14] );
+
+   m[ 0] = ctx->b[ 0];
+   m[ 1] = ctx->b[ 1];
+   m[ 2] = ctx->b[ 2];
+   m[ 3] = ctx->b[ 3];
+   m[ 4] = ctx->b[ 4];
+   m[ 5] = ctx->b[ 5];
+   m[ 6] = ctx->b[ 6];
+   m[ 7] = ctx->b[ 7];
+   m[ 8] = ctx->b[ 8];
+   m[ 9] = ctx->b[ 9];
+   m[10] = ctx->b[10];
+   m[11] = ctx->b[11];
+   m[12] = ctx->b[12];
+   m[13] = ctx->b[13];
+   m[14] = ctx->b[14];
+   m[15] = ctx->b[15];
+   
+	for ( int i = 0; i < 12; i++ )
+   { 
+		B2B_G( 0, 4,  8, 12, m[ sigma[i][ 0] ], m[ sigma[i][ 1] ] );
+		B2B_G( 1, 5,  9, 13, m[ sigma[i][ 2] ], m[ sigma[i][ 3] ] );
+		B2B_G( 2, 6, 10, 14, m[ sigma[i][ 4] ], m[ sigma[i][ 5] ] );
+		B2B_G( 3, 7, 11, 15, m[ sigma[i][ 6] ], m[ sigma[i][ 7] ] );
+		B2B_G( 0, 5, 10, 15, m[ sigma[i][ 8] ], m[ sigma[i][ 9] ] );
+		B2B_G( 1, 6, 11, 12, m[ sigma[i][10] ], m[ sigma[i][11] ] );
+		B2B_G( 2, 7,  8, 13, m[ sigma[i][12] ], m[ sigma[i][13] ] );
+		B2B_G( 3, 4,  9, 14, m[ sigma[i][14] ], m[ sigma[i][15] ] );
+	}
+
+   ctx->h[0] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[0], v[0] ), v[ 8] );
+   ctx->h[1] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[1], v[1] ), v[ 9] );
+   ctx->h[2] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[2], v[2] ), v[10] );
+   ctx->h[3] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[3], v[3] ), v[11] );
+   ctx->h[4] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[4], v[4] ), v[12] );
+   ctx->h[5] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[5], v[5] ), v[13] );
+   ctx->h[6] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[6], v[6] ), v[14] );
+   ctx->h[7] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[7], v[7] ), v[15] );
+}
+
+int blake2b_4way_init( blake2b_4way_ctx *ctx ) 
+{
+	size_t i;
+
+   ctx->h[0] = m256_const1_64( 0x6A09E667F3BCC908 );
+   ctx->h[1] = m256_const1_64( 0xBB67AE8584CAA73B );
+   ctx->h[2] = m256_const1_64( 0x3C6EF372FE94F82B );
+   ctx->h[3] = m256_const1_64( 0xA54FF53A5F1D36F1 );
+   ctx->h[4] = m256_const1_64( 0x510E527FADE682D1 );
+   ctx->h[5] = m256_const1_64( 0x9B05688C2B3E6C1F );
+   ctx->h[6] = m256_const1_64( 0x1F83D9ABFB41BD6B );
+   ctx->h[7] = m256_const1_64( 0x5BE0CD19137E2179 );
+
+   ctx->h[0] = _mm256_xor_si256( ctx->h[0], m256_const1_64( 0x01010020 ) );
+
+	ctx->t[0] = 0;
+	ctx->t[1] = 0;
+	ctx->c = 0;
+	ctx->outlen = 32;
+
+   for ( i = 0; i < 16; i++ )
+     ctx->b[i] = m256_zero;
+
+	return 0;
+}
+
+void blake2b_4way_update( blake2b_4way_ctx *ctx, const void *input,
+                          size_t inlen ) 
+{
+   __m256i* in =(__m256i*)input;
+
+	size_t i, c;
+   c = ctx->c >> 3; 
+
+	for ( i = 0; i < (inlen >> 3); i++ )
+   {
+		if ( ctx->c == 128 )
+      { 
+			ctx->t[0] += ctx->c;
+			if ( ctx->t[0] < ctx->c )
+				ctx->t[1]++;
+			blake2b_4way_compress( ctx, 0 );
+			ctx->c = 0;
+		}
+      ctx->b[ c++ ] = in[i];
+      ctx->c += 8;
+   }
+}
+
+void blake2b_4way_final( blake2b_4way_ctx *ctx, void *out )
+{
+	size_t c;
+   c = ctx->c >> 3;
+
+	ctx->t[0] += ctx->c;
+	if ( ctx->t[0] < ctx->c )
+		ctx->t[1]++;
+
+	while ( ctx->c < 128 )
+   {
+      ctx->b[c++] = m256_zero;
+      ctx->c += 8;
+   }
+
+   blake2b_4way_compress( ctx, 1 );           // final block flag = 1
+
+   casti_m256i( out, 0 ) = ctx->h[0];
+   casti_m256i( out, 1 ) = ctx->h[1];
+   casti_m256i( out, 2 ) = ctx->h[2];
+   casti_m256i( out, 3 ) = ctx->h[3];
+}
+
+#endif
--- a/algo/blake/blake2b-hash-4way.h
+++ b/algo/blake/blake2b-hash-4way.h
@@ -0,0 +1,53 @@
+#pragma once
+#ifndef __BLAKE2B_HASH_4WAY_H__
+#define __BLAKE2B_HASH_4WAY_H__
+
+#include "simd-utils.h"
+#include <stddef.h>
+#include <stdint.h>
+
+#if defined(_MSC_VER)
+#include <inttypes.h>
+#define inline __inline
+#define ALIGN(x) __declspec(align(x))
+#else
+#define ALIGN(x) __attribute__((aligned(x)))
+#endif
+
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+ALIGN(128) typedef struct {
+   __m512i b[16]; // input buffer
+   __m512i h[8];  // chained state
+   uint64_t t[2];  // total number of bytes
+   size_t c;       // pointer for b[]
+   size_t outlen;  // digest size
+} blake2b_8way_ctx;
+
+int blake2b_8way_init( blake2b_8way_ctx *ctx );
+void blake2b_8way_update( blake2b_8way_ctx *ctx, const void *input,
+                          size_t inlen );
+void blake2b_8way_final( blake2b_8way_ctx *ctx, void *out );
+
+#endif
+
+#if defined(__AVX2__)
+
+// state context
+ALIGN(128) typedef struct {
+	__m256i b[16]; // input buffer
+	__m256i h[8];  // chained state
+	uint64_t t[2];  // total number of bytes
+	size_t c;       // pointer for b[]
+	size_t outlen;  // digest size
+} blake2b_4way_ctx;
+
+int blake2b_4way_init( blake2b_4way_ctx *ctx );
+void blake2b_4way_update( blake2b_4way_ctx *ctx, const void *input,
+                          size_t inlen );
+void blake2b_4way_final( blake2b_4way_ctx *ctx, void *out );
+
+#endif
+
+#endif
--- a/algo/blake/blake2b.c
+++ b/algo/blake/blake2b.c
@@ -3,13 +3,11 @@
 * tpruvot@github 2015-2016
 */

-#include "algo-gate-api.h"
+#include "blake2b-gate.h"
 #include <string.h>
 #include <stdint.h>
 #include "algo/blake/sph_blake2b.h"

-//static __thread sph_blake2b_ctx s_midstate;
-//static __thread sph_blake2b_ctx s_ctx;
 #define MIDLEN 76
 #define A 64

@@ -25,16 +23,6 @@ void blake2b_hash(void *output, const void *input)
 	memcpy(output, hash, 32);
 }

-/*
-static void blake2b_hash_end(uint32_t *output, const uint32_t *input)
-{
-	s_ctx.outlen = MIDLEN;
-	memcpy(&s_ctx, &s_midstate, 32 + 16 + MIDLEN);
-	sph_blake2b_update(&s_ctx, (uint8_t*) &input[MIDLEN/4], 80 - MIDLEN);
-	sph_blake2b_final(&s_ctx, (uint8_t*) output);
-}
-*/
-
 int scanhash_blake2b( struct work *work, uint32_t max_nonce,
                      uint64_t *hashes_done, struct thr_info *mythr )
 {
@@ -45,7 +33,7 @@ int scanhash_blake2b( struct work *work, uint32_t max_nonce,
   int thr_id = mythr->id;  // thr_id arg is deprecated

 	const uint32_t Htarg = ptarget[7];
-	const uint32_t first_nonce = pdata[8];
+	const uint32_t first_nonce = pdata[19];

 	uint32_t n = first_nonce;

@@ -53,179 +41,23 @@ int scanhash_blake2b( struct work *work, uint32_t max_nonce,
 		be32enc(&endiandata[i], pdata[i]);
 	}

-	// midstate (untested yet)
-	//blake2b_init(&s_midstate, 32, NULL, 0);
-	//blake2b_update(&s_midstate, (uint8_t*) endiandata, MIDLEN);
-	//memcpy(&s_ctx, &s_midstate, sizeof(blake2b_ctx));
-
 	do {
-		be32enc(&endiandata[8], n);
+		be32enc(&endiandata[19], n);
 		//blake2b_hash_end(vhashcpu, endiandata);
 		blake2b_hash(vhashcpu, endiandata);

 		if (vhashcpu[7] < Htarg && fulltest(vhashcpu, ptarget)) {
 			work_set_target_ratio(work, vhashcpu);
 			*hashes_done = n - first_nonce + 1;
-			pdata[8] = n;
+			pdata[19] = n;
 			return 1;
 		}
 		n++;

 	} while (n < max_nonce && !work_restart[thr_id].restart);
 	*hashes_done = n - first_nonce + 1;
-	pdata[8] = n;
+	pdata[19] = n;

 	return 0;
 }

-static inline void swab256(void *dest_p, const void *src_p)
-{
-	uint32_t *dest = (uint32_t *)dest_p;
-	const uint32_t *src = (uint32_t *)src_p;
-
-	dest[0] = swab32(src[7]);
-	dest[1] = swab32(src[6]);
-	dest[2] = swab32(src[5]);
-	dest[3] = swab32(src[4]);
-	dest[4] = swab32(src[3]);
-	dest[5] = swab32(src[2]);
-	dest[6] = swab32(src[1]);
-	dest[7] = swab32(src[0]);
-}
-
-/* compute nbits to get the network diff */
-void blake2b_calc_network_diff(struct work *work)
-{
-        // sample for diff 43.281 : 1c05ea29
-        uint32_t nbits = work->data[11]; // unsure if correct
-        uint32_t bits = (nbits & 0xffffff);
-        int16_t shift = (swab32(nbits) & 0xff); // 0x1c = 28
-
-        double d = (double)0x0000ffff / (double)bits;
-        for (int m=shift; m < 29; m++) d *= 256.0;
-        for (int m=29; m < shift; m++) d /= 256.0;
-        if (opt_debug_diff)
-                applog(LOG_DEBUG, "net diff: %f -> shift %u, bits %08x", d, shift, bits);
-        net_diff = d;
-}
-
-void blake2b_be_build_stratum_request( char *req, struct work *work )
-{
-   unsigned char *xnonce2str;
-   uint32_t ntime,       nonce;
-   char     ntimestr[9], noncestr[9];
-   be32enc( &ntime, work->data[ algo_gate.ntime_index ] );
-   be32enc( &nonce, work->data[ algo_gate.nonce_index ] );
-   bin2hex( ntimestr, (char*)(&ntime), sizeof(uint32_t) );
-   bin2hex( noncestr, (char*)(&nonce), sizeof(uint32_t) );
-   uint16_t high_nonce = swab32(work->data[9]) >> 16;
-   xnonce2str = abin2hex((unsigned char*)(&high_nonce), 2);
-   snprintf( req, JSON_BUF_LEN,
-        "{\"method\": \"mining.submit\", \"params\": [\"%s\", \"%s\", \"%s\", \"%s\", \"%s\"], \"id\":4}",
-         rpc_user, work->job_id, xnonce2str, ntimestr, noncestr );
-   free( xnonce2str );
-}
-
-#define min(a,b) (a>b ? (b) :(a))
-
-// merkle root handled here, no need for gen_merkle_root gate target
-void blake2b_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
-{
-    uchar merkle_root[64] = { 0 };
-    uint32_t extraheader[32] = { 0 };
-    int headersize = 0;
-    size_t t;
-    int i;
-
-    // merkle root
-    memcpy( merkle_root, sctx->job.coinbase, 32 );
-    headersize = min( (int)sctx->job.coinbase_size - 32, sizeof(extraheader) );
-    memcpy( extraheader, &sctx->job.coinbase[32], headersize );
-    // Increment extranonce2 
-    for ( t = 0; t < sctx->xnonce2_size && !( ++sctx->job.xnonce2[t] ); t++ );
-    // Assemble block header 
-    memset( g_work->data, 0, sizeof(g_work->data) );
-//    g_work->data[0] = le32dec( sctx->job.version );
-//    for ( i = 0; i < 8; i++ )
-//       g_work->data[1 + i] = le32dec( (uint32_t *) sctx->job.prevhash + i );
-    for ( i = 0; i < 8; i++ )
-       g_work->data[i] = ((uint32_t*)sctx->job.prevhash)[7-i];
-//    for ( i = 0; i < 8; i++ )
-//       g_work->data[9 + i] = be32dec( (uint32_t *) merkle_root + i );
-    g_work->data[8]  = 0; // nonce
-    g_work->data[9]  = swab32( extraheader[0] ) | ( rand() & 0xf0 );
-    g_work->data[10] = be32dec( sctx->job.ntime );
-    g_work->data[11] = be32dec( sctx->job.nbits );
-    for ( i = 0; i < 8; i++ )
-       g_work->data[12+i] = ( (uint32_t*)merkle_root )[i];
-}
-
-#undef min
-
-void blake2b_get_new_work( struct work* work, struct work* g_work, int thr_id,
-                           uint32_t* end_nonce_ptr, bool clean_job )
-{
-   const int wkcmp_sz = 32;  // bytes
-   const int wkcmp_off = 32 + 16; 
-   uint32_t *nonceptr = algo_gate.get_nonceptr( work->data );
-
-   if ( memcmp( &work->data[ wkcmp_off ], &g_work->data[ wkcmp_off ], wkcmp_sz )
-      && ( clean_job || ( *nonceptr >= *end_nonce_ptr ) 
-      || strcmp( work->job_id, g_work->job_id ) ) )
-   {
-      work_free( work );
-      work_copy( work, g_work );
-      *nonceptr = ( 0xffffffffU / opt_n_threads ) * thr_id;
-      if ( opt_randomize )
-         *nonceptr += ( (rand() *4 ) & UINT32_MAX ) / opt_n_threads;
-      *end_nonce_ptr = ( 0xffffffffU / opt_n_threads ) * (thr_id+1) - 0x20;
-   }
-   else
-       ++(*nonceptr);
-
-   // suprnova job_id check without data/target/height change...
-   // we just may have copied new g_wwork to work so why this test here?
-//   if (  have_stratum && strcmp( work->job_id, g_work->job_id ) )
-      // exit thread loop
-//      continue;
-//   else
-//   {
-//      nonceptr[1] += 0x10;
-//      nonceptr[1] |= thr_id;
-//   }
-}
-
-bool blake2b_ready_to_mine( struct work* work, struct stratum_ctx* stratum,
-                           int thr_id )
-{
-   if ( have_stratum && strcmp( stratum->job.job_id, work->job_id ) )
-      // need to regen g_work..
-      return false;
-   // extradata: prevent duplicates
-   work->data[ 8     ] += 0x10;
-   work->data[ 8 + 1 ] |= thr_id;
-   return true;
-}
-
-double blake2b_get_max64() { return 0x1fffffLL; }
-
-bool register_blake2b_algo( algo_gate_t* gate )
-{
-  algo_not_tested();
-  gate->ntime_index   = 10;
-  gate->nbits_index   = 11;
-  gate->nonce_index   =  8;
-  gate->work_cmp_size = 32;
-  gate->scanhash              = (void*)&scanhash_blake2b;
-  gate->hash                  = (void*)&blake2b_hash;
-  gate->calc_network_diff     = (void*)&blake2b_calc_network_diff;
-  gate->build_stratum_request = (void*)&blake2b_be_build_stratum_request;
-  gate->work_decode           = (void*)&std_be_work_decode;
-  gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;
-  gate->build_extraheader     = (void*)&blake2b_build_extraheader;
-  gate->get_new_work          = (void*)&blake2b_get_new_work;
-  gate->get_max64             = (void*)&blake2b_get_max64;
-  gate->ready_to_mine         = (void*)&blake2b_ready_to_mine;
-  have_gbt = false;
-  return true;
-}
--- a/algo/blake/blake2s-4way.c
+++ b/algo/blake/blake2s-4way.c
@@ -3,22 +3,72 @@
 #include <string.h>
 #include <stdint.h>

-#if defined(BLAKE2S_8WAY)
+#if defined(BLAKE2S_16WAY)
+
+static __thread blake2s_16way_state blake2s_16w_ctx;
+
+void blake2s_16way_hash( void *output, const void *input )
+{
+   blake2s_16way_state ctx;
+   memcpy( &ctx, &blake2s_16w_ctx, sizeof ctx );
+   blake2s_16way_update( &ctx, input + (64<<4), 16 );
+   blake2s_16way_final( &ctx, output, BLAKE2S_OUTBYTES );
+}
+
+int scanhash_blake2s_16way( struct work *work, uint32_t max_nonce,
+                            uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t vdata[20*16] __attribute__ ((aligned (128)));
+   uint32_t hash[8*16] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[7<<4]);
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+   __m512i  *noncev = (__m512i*)vdata + 19;   // aligned
+   uint32_t n = first_nonce;
+   int thr_id = mythr->id;  
+
+   mm512_bswap32_intrlv80_16x32( vdata, pdata );
+   blake2s_16way_init( &blake2s_16w_ctx, BLAKE2S_OUTBYTES );
+   blake2s_16way_update( &blake2s_16w_ctx, vdata, 64 );
+
+   do {
+      *noncev = mm512_bswap_32( _mm512_set_epi32(
+	                  n+15, n+14, n+13, n+12, n+11, n+10, n+ 9, n+ 8,
+	                  n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n+ 1, n ) );
+      pdata[19] = n;
+
+      blake2s_16way_hash( hash, vdata );
+
+      for ( int lane = 0; lane < 16; lane++ )
+      if ( unlikely( hash7[lane] <= Htarg ) )
+      {
+         extr_lane_16x32( lane_hash, hash, lane, 256 );
+         if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
+         {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+         }
+      }
+      n += 16;
+   } while ( (n < max_nonce-16) && !work_restart[thr_id].restart );
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#elif defined(BLAKE2S_8WAY)

 static __thread blake2s_8way_state blake2s_8w_ctx;

 void blake2s_8way_hash( void *output, const void *input )
 {
-   uint32_t vhash[8*8] __attribute__ ((aligned (64)));
   blake2s_8way_state ctx;
   memcpy( &ctx, &blake2s_8w_ctx, sizeof ctx );
-
   blake2s_8way_update( &ctx, input + (64<<3), 16 );
-   blake2s_8way_final( &ctx, vhash, BLAKE2S_OUTBYTES );
-
-   mm256_dintrlv_8x32( output,     output+ 32, output+ 64, output+ 96,
-                            output+128, output+160, output+192, output+224,
-                            vhash, 256 );
+   blake2s_8way_final( &ctx, output, BLAKE2S_OUTBYTES );
 }

 int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
@@ -26,44 +76,38 @@ int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
 {
   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
   uint32_t hash[8*8] __attribute__ ((aligned (32)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash7 = &(hash[7<<3]);
   uint32_t *pdata = work->data;
   uint32_t *ptarget = work->target;
-   uint32_t _ALIGN(64) edata[20];
   const uint32_t Htarg = ptarget[7];
   const uint32_t first_nonce = pdata[19];
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
   uint32_t n = first_nonce;
-   uint32_t *noncep = vdata + 152;   // 19*8
-   int thr_id = mythr->id;  // thr_id arg is deprecated
+   int thr_id = mythr->id; 

-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
   blake2s_8way_init( &blake2s_8w_ctx, BLAKE2S_OUTBYTES );
   blake2s_8way_update( &blake2s_8w_ctx, vdata, 64 );

   do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
-      be32enc( noncep +4, n+4 );
-      be32enc( noncep +5, n+5 );
-      be32enc( noncep +6, n+6 );
-      be32enc( noncep +7, n+7 );
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
      pdata[19] = n;

      blake2s_8way_hash( hash, vdata );

-
-      for ( int i = 0; i < 8; i++ )
-      if (  (hash+(i<<3))[7] <= Htarg )
-      if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
+      for ( int lane = 0; lane < 8; lane++ )
+      if ( unlikely( hash7[lane] <= Htarg ) )
      {
-          pdata[19] = n+i;
-          submit_lane_solution( work, hash+(i<<3), mythr, i );
+         extr_lane_8x32( lane_hash, hash, lane, 256 );
+         if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
+         {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+         }
      }
      n += 8;
-
   } while ( (n < max_nonce) && !work_restart[thr_id].restart );

   *hashes_done = n - first_nonce + 1;
@@ -76,15 +120,10 @@ static __thread blake2s_4way_state blake2s_4w_ctx;

 void blake2s_4way_hash( void *output, const void *input )
 {
-   uint32_t vhash[8*4] __attribute__ ((aligned (64)));
   blake2s_4way_state ctx;
   memcpy( &ctx, &blake2s_4w_ctx, sizeof ctx );
-
   blake2s_4way_update( &ctx, input + (64<<2), 16 );
-   blake2s_4way_final( &ctx, vhash, BLAKE2S_OUTBYTES );
-
-   dintrlv_4x32( output, output+32, output+64, output+96,
-		            vhash, 256 );
+   blake2s_4way_final( &ctx, output, BLAKE2S_OUTBYTES );
 }

 int scanhash_blake2s_4way( struct work *work, uint32_t max_nonce,
@@ -92,38 +131,36 @@ int scanhash_blake2s_4way( struct work *work, uint32_t max_nonce,
 {
   uint32_t vdata[20*4] __attribute__ ((aligned (64)));
   uint32_t hash[8*4] __attribute__ ((aligned (32)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash7 = &(hash[7<<2]);
   uint32_t *pdata = work->data;
   uint32_t *ptarget = work->target;
-   uint32_t _ALIGN(64) edata[20];
   const uint32_t Htarg = ptarget[7];
   const uint32_t first_nonce = pdata[19];
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
   uint32_t n = first_nonce;
-   uint32_t *noncep = vdata + 76;   // 19*4
-   int thr_id = mythr->id;  // thr_id arg is deprecated
+   int thr_id = mythr->id; 

-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
   blake2s_4way_init( &blake2s_4w_ctx, BLAKE2S_OUTBYTES );
   blake2s_4way_update( &blake2s_4w_ctx, vdata, 64 );

   do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
      pdata[19] = n;

      blake2s_4way_hash( hash, vdata );

-      for ( int i = 0; i < 4; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg )
-      if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
+      for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
      {
-          pdata[19] = n+i;
-          submit_lane_solution( work, hash+(i<<3), mythr, i );
+         extr_lane_4x32( lane_hash, hash, lane, 256 );
+         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+         {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+              }
      }
      n += 4;
-
   } while ( (n < max_nonce) && !work_restart[thr_id].restart );

   *hashes_done = n - first_nonce + 1;
--- a/algo/blake/blake2s-gate.c
+++ b/algo/blake/blake2s-gate.c
@@ -1,15 +1,12 @@
 #include "blake2s-gate.h"

-
-// changed to get_max64_0x3fffffLL in cpuminer-multi-decred
-int64_t blake2s_get_max64 ()
-{
-   return 0x7ffffLL;
-}
-
 bool register_blake2s_algo( algo_gate_t* gate )
 {
-#if defined(BLAKE2S_8WAY)
+#if defined(BLAKE2S_16WAY)
+  gate->scanhash  = (void*)&scanhash_blake2s_16way;
+  gate->hash      = (void*)&blake2s_16way_hash;
+#elif defined(BLAKE2S_8WAY)
+//#if defined(BLAKE2S_8WAY)
  gate->scanhash  = (void*)&scanhash_blake2s_8way;
  gate->hash      = (void*)&blake2s_8way_hash;
 #elif defined(BLAKE2S_4WAY)
@@ -19,8 +16,7 @@ bool register_blake2s_algo( algo_gate_t* gate )
  gate->scanhash  = (void*)&scanhash_blake2s;
  gate->hash      = (void*)&blake2s_hash;
 #endif
-  gate->get_max64 = (void*)&blake2s_get_max64;
-  gate->optimizations = SSE42_OPT | AVX2_OPT;
+  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
  return true;
 };

--- a/algo/blake/blake2s-gate.h
+++ b/algo/blake/blake2s-gate.h
@@ -4,16 +4,30 @@
 #include <stdint.h>
 #include "algo-gate-api.h"

-#if defined(__SSE4_2__)
+//#if defined(__SSE4_2__)
+#if defined(__SSE2__)
  #define BLAKE2S_4WAY
 #endif
+
 #if defined(__AVX2__)
  #define BLAKE2S_8WAY
 #endif

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define BLAKE2S_16WAY
+#endif
+
 bool register_blake2s_algo( algo_gate_t* gate );

-#if defined(BLAKE2S_8WAY)
+#if defined(BLAKE2S_16WAY)
+
+void blake2s_16way_hash( void *state, const void *input );
+int scanhash_blake2s_16way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+#elif defined (BLAKE2S_8WAY)
+
+//#if defined(BLAKE2S_8WAY)

 void blake2s_8way_hash( void *state, const void *input );
 int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
--- a/algo/blake/blake2s-hash-4way.c
+++ b/algo/blake/blake2s-hash-4way.c
@@ -17,13 +17,16 @@
 #include <string.h>
 #include <stdio.h>

-#if defined(__SSE4_2__)
+//#if defined(__SSE4_2__)
+#if defined(__SSE2__)

+/*
 static const uint32_t blake2s_IV[8] =
 {
 	0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
 	0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
 };
+*/

 static const uint8_t blake2s_sigma[10][16] =
 {
@@ -39,6 +42,7 @@ static const uint8_t blake2s_sigma[10][16] =
 	{ 10,  2,  8,  4,  7,  6,  1,  5, 15, 11,  9, 14,  3, 12, 13 , 0 } ,
 };

+
 // define a constant for initial param.

 int blake2s_4way_init( blake2s_4way_state *S, const uint8_t outlen )
@@ -57,8 +61,18 @@ int blake2s_4way_init( blake2s_4way_state *S, const uint8_t outlen )
   memset( P->personal, 0, sizeof( P->personal ) );

   memset( S, 0, sizeof( blake2s_4way_state ) );
-   for( int i = 0; i < 8; ++i )
-      S->h[i] = _mm_set1_epi32( blake2s_IV[i] );
+
+   S->h[0] = m128_const1_64( 0x6A09E6676A09E667ULL );
+   S->h[1] = m128_const1_64( 0xBB67AE85BB67AE85ULL );
+   S->h[2] = m128_const1_64( 0x3C6EF3723C6EF372ULL );
+   S->h[3] = m128_const1_64( 0xA54FF53AA54FF53AULL );
+   S->h[4] = m128_const1_64( 0x510E527F510E527FULL );
+   S->h[5] = m128_const1_64( 0x9B05688C9B05688CULL );
+   S->h[6] = m128_const1_64( 0x1F83D9AB1F83D9ABULL );
+   S->h[7] = m128_const1_64( 0x5BE0CD195BE0CD19ULL );
+   
+//   for( int i = 0; i < 8; ++i )
+//      S->h[i] = _mm_set1_epi32( blake2s_IV[i] );

   uint32_t *p = ( uint32_t * )( P );

@@ -76,41 +90,45 @@ int blake2s_4way_compress( blake2s_4way_state *S, const __m128i* block )
   memcpy_128( m, block, 16 );
   memcpy_128( v, S->h, 8 );

-   v[ 8] = _mm_set1_epi32( blake2s_IV[0] );
-   v[ 9] = _mm_set1_epi32( blake2s_IV[1] );
-   v[10] = _mm_set1_epi32( blake2s_IV[2] );
-   v[11] = _mm_set1_epi32( blake2s_IV[3] );
+   v[ 8] = m128_const1_64( 0x6A09E6676A09E667ULL );
+   v[ 9] = m128_const1_64( 0xBB67AE85BB67AE85ULL );
+   v[10] = m128_const1_64( 0x3C6EF3723C6EF372ULL );
+   v[11] = m128_const1_64( 0xA54FF53AA54FF53AULL );
   v[12] = _mm_xor_si128( _mm_set1_epi32( S->t[0] ),
-                          _mm_set1_epi32( blake2s_IV[4] ) );
+                          m128_const1_64( 0x510E527F510E527FULL ) );
   v[13] = _mm_xor_si128( _mm_set1_epi32( S->t[1] ),
-                          _mm_set1_epi32( blake2s_IV[5] ) );
+                          m128_const1_64( 0x9B05688C9B05688CULL ) );
   v[14] = _mm_xor_si128( _mm_set1_epi32( S->f[0] ),
-                          _mm_set1_epi32( blake2s_IV[6] ) );
+                          m128_const1_64( 0x1F83D9AB1F83D9ABULL ) );
   v[15] = _mm_xor_si128( _mm_set1_epi32( S->f[1] ),
-                          _mm_set1_epi32( blake2s_IV[7] ) );
+                          m128_const1_64( 0x5BE0CD195BE0CD19ULL ) );

-#define G4W(r,i,a,b,c,d) \
+#define G4W( sigma0, sigma1, a, b, c, d ) \
 do { \
-   a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ blake2s_sigma[r][2*i+0] ] ); \
+   uint8_t s0 = sigma0; \
+   uint8_t s1 = sigma1; \
+   a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ s0 ] ); \
   d = mm128_ror_32( _mm_xor_si128( d, a ), 16 ); \
   c = _mm_add_epi32( c, d ); \
   b = mm128_ror_32( _mm_xor_si128( b, c ), 12 ); \
-   a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ blake2s_sigma[r][2*i+1] ] ); \
+   a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ s1 ] ); \
   d = mm128_ror_32( _mm_xor_si128( d, a ),  8 ); \
   c = _mm_add_epi32( c, d ); \
   b = mm128_ror_32( _mm_xor_si128( b, c ),  7 ); \
 } while(0)

+
 #define ROUND4W(r)  \
 do { \
-   G4W( r, 0, v[ 0], v[ 4], v[ 8], v[12] ); \
-   G4W( r, 1, v[ 1], v[ 5], v[ 9], v[13] ); \
-   G4W( r, 2, v[ 2], v[ 6], v[10], v[14] ); \
-   G4W( r, 3, v[ 3], v[ 7], v[11], v[15] ); \
-   G4W( r, 4, v[ 0], v[ 5], v[10], v[15] ); \
-   G4W( r, 5, v[ 1], v[ 6], v[11], v[12] ); \
-   G4W( r, 6, v[ 2], v[ 7], v[ 8], v[13] ); \
-   G4W( r, 7, v[ 3], v[ 4], v[ 9], v[14] ); \
+   uint8_t *sigma = (uint8_t*)&blake2s_sigma[r]; \
+   G4W( sigma[ 0], sigma[ 1], v[ 0], v[ 4], v[ 8], v[12] ); \
+   G4W( sigma[ 2], sigma[ 3], v[ 1], v[ 5], v[ 9], v[13] ); \
+   G4W( sigma[ 4], sigma[ 5], v[ 2], v[ 6], v[10], v[14] ); \
+   G4W( sigma[ 6], sigma[ 7], v[ 3], v[ 7], v[11], v[15] ); \
+   G4W( sigma[ 8], sigma[ 9], v[ 0], v[ 5], v[10], v[15] ); \
+   G4W( sigma[10], sigma[11], v[ 1], v[ 6], v[11], v[12] ); \
+   G4W( sigma[12], sigma[13], v[ 2], v[ 7], v[ 8], v[13] ); \
+   G4W( sigma[14], sigma[15], v[ 3], v[ 4], v[ 9], v[14] ); \
 } while(0)

   ROUND4W( 0 );
@@ -132,26 +150,47 @@ do { \
   return 0;
 }

+// There is a problem that can't be resolved internally.
+// If the last block is a full 64 bytes it should not be compressed in
+// update but left for final. However, when streaming, it isn't known
+// which block is last. There may be a subsequent call to update to add
+// more data.
+//
+// The reference code handled this by juggling 2 blocks at a time at
+// a significant performance penalty.
+//
+// Instead a new function is introduced called full_blocks which combines
+// update and final and is to be used in non-streaming mode where the data
+// is a multiple of 64 bytes.
+// 
+// Supported:
+//    64 + 16 bytes  (blake2s with midstate optimization)
+//    80 bytes       (blake2s without midstate optimization)
+//    Any multiple of 64 bytes in one shot (x25x)
+//
+// Unsupported:
+//    Stream of full 64 byte blocks one at a time.   
+
+// use only when streaming more data or final block not full.
 int blake2s_4way_update( blake2s_4way_state *S, const void *in,
                         uint64_t inlen )
 {
-  __m128i *input = (__m128i*)in;
-  __m128i *buf = (__m128i*)S->buf;
-  const int bsize = BLAKE2S_BLOCKBYTES;
+   __m128i *input = (__m128i*)in;
+   __m128i *buf = (__m128i*)S->buf;

   while( inlen > 0 )
   {
      size_t left = S->buflen;
-      if( inlen >= bsize - left )
+      if( inlen >= BLAKE2S_BLOCKBYTES - left )
      {
-         memcpy_128( buf + (left>>2), input, (bsize - left) >> 2 );
-         S->buflen += bsize - left;
+         memcpy_128( buf + (left>>2), input, (BLAKE2S_BLOCKBYTES - left) >> 2 );
+         S->buflen += BLAKE2S_BLOCKBYTES - left;
         S->t[0] += BLAKE2S_BLOCKBYTES;
         S->t[1] += ( S->t[0] < BLAKE2S_BLOCKBYTES );
         blake2s_4way_compress( S, buf ); 
         S->buflen = 0;
-         input += ( bsize >> 2 );
-         inlen -= bsize;
+         input += ( BLAKE2S_BLOCKBYTES >> 2 );
+         inlen -= BLAKE2S_BLOCKBYTES;
      }
      else
      {
@@ -183,8 +222,45 @@ int blake2s_4way_final( blake2s_4way_state *S, void *out, uint8_t outlen )
   return 0;
 }

+// Update and final when inlen is a multiple of 64 bytes
+int blake2s_4way_full_blocks( blake2s_4way_state *S, void *out,
+                              const void *input, uint64_t inlen )
+{
+    __m128i *in = (__m128i*)input;
+    __m128i *buf = (__m128i*)S->buf;
+
+    while( inlen > BLAKE2S_BLOCKBYTES )
+    {
+       memcpy_128( buf, in, BLAKE2S_BLOCKBYTES >> 2 );
+       S->buflen = BLAKE2S_BLOCKBYTES;
+       inlen -= BLAKE2S_BLOCKBYTES;
+       S->t[0] += BLAKE2S_BLOCKBYTES;
+       S->t[1] += ( S->t[0] < BLAKE2S_BLOCKBYTES );
+       blake2s_4way_compress( S, buf );
+       S->buflen = 0;
+       in += ( BLAKE2S_BLOCKBYTES >> 2 );
+    }
+
+    // last block
+    memcpy_128( buf, in, BLAKE2S_BLOCKBYTES >> 2 );
+    S->buflen = BLAKE2S_BLOCKBYTES;
+    S->t[0] += S->buflen;
+    S->t[1] += ( S->t[0] < S->buflen );
+    if ( S->last_node )  S->f[1] = ~0U;
+    S->f[0] = ~0U;
+    blake2s_4way_compress( S, buf );
+
+    for ( int i = 0; i < 8; ++i )
+      casti_m128i( out, i ) = S->h[ i ];
+    return 0;
+}
+
 #if defined(__AVX2__)

+// The commented code below is slower on Intel but faster on
+// Zen1 AVX2. It's also faster than Zen1 AVX.
+// Ryzen gen2 is unknown at this time.
+ 
 int blake2s_8way_compress( blake2s_8way_state *S, const __m256i *block )
 {
   __m256i m[16];
@@ -193,6 +269,23 @@ int blake2s_8way_compress( blake2s_8way_state *S, const __m256i *block )
   memcpy_256( m, block, 16 );
   memcpy_256( v, S->h, 8 );

+   v[ 8] = m256_const1_64( 0x6A09E6676A09E667ULL );
+   v[ 9] = m256_const1_64( 0xBB67AE85BB67AE85ULL );
+   v[10] = m256_const1_64( 0x3C6EF3723C6EF372ULL );
+   v[11] = m256_const1_64( 0xA54FF53AA54FF53AULL );
+   v[12] = _mm256_xor_si256( _mm256_set1_epi32( S->t[0] ),
+                          m256_const1_64( 0x510E527F510E527FULL ) );
+
+   v[13] = _mm256_xor_si256( _mm256_set1_epi32( S->t[1] ),
+                          m256_const1_64( 0x9B05688C9B05688CULL ) );
+
+   v[14] = _mm256_xor_si256( _mm256_set1_epi32( S->f[0] ),
+                          m256_const1_64( 0x1F83D9AB1F83D9ABULL ) );
+
+   v[15] = _mm256_xor_si256( _mm256_set1_epi32( S->f[1] ),
+                          m256_const1_64( 0x5BE0CD195BE0CD19ULL ) );
+
+/*
   v[ 8] = _mm256_set1_epi32( blake2s_IV[0] );
   v[ 9] = _mm256_set1_epi32( blake2s_IV[1] );
   v[10] = _mm256_set1_epi32( blake2s_IV[2] );
@@ -206,6 +299,7 @@ int blake2s_8way_compress( blake2s_8way_state *S, const __m256i *block )
   v[15] = _mm256_xor_si256( _mm256_set1_epi32( S->f[1] ),
                             _mm256_set1_epi32( blake2s_IV[7] ) );

+
 #define G8W(r,i,a,b,c,d) \
 do { \
   a = _mm256_add_epi32( _mm256_add_epi32( a, b ), \
@@ -219,7 +313,36 @@ do { \
   c = _mm256_add_epi32( c, d ); \
   b = mm256_ror_32( _mm256_xor_si256( b, c ),  7 ); \
 } while(0)
+*/

+#define G8W( sigma0, sigma1, a, b, c, d) \
+do { \
+   uint8_t s0 = sigma0; \
+   uint8_t s1 = sigma1; \
+   a = _mm256_add_epi32( _mm256_add_epi32( a, b ), m[ s0 ] ); \
+   d = mm256_ror_32( _mm256_xor_si256( d, a ), 16 ); \
+   c = _mm256_add_epi32( c, d ); \
+   b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
+   a = _mm256_add_epi32( _mm256_add_epi32( a, b ), m[ s1 ] ); \
+   d = mm256_ror_32( _mm256_xor_si256( d, a ),  8 ); \
+   c = _mm256_add_epi32( c, d ); \
+   b = mm256_ror_32( _mm256_xor_si256( b, c ),  7 ); \
+} while(0)
+
+#define ROUND8W(r)  \
+do { \
+   uint8_t *sigma = (uint8_t*)&blake2s_sigma[r]; \
+   G8W( sigma[ 0], sigma[ 1], v[ 0], v[ 4], v[ 8], v[12] ); \
+   G8W( sigma[ 2], sigma[ 3], v[ 1], v[ 5], v[ 9], v[13] ); \
+   G8W( sigma[ 4], sigma[ 5], v[ 2], v[ 6], v[10], v[14] ); \
+   G8W( sigma[ 6], sigma[ 7], v[ 3], v[ 7], v[11], v[15] ); \
+   G8W( sigma[ 8], sigma[ 9], v[ 0], v[ 5], v[10], v[15] ); \
+   G8W( sigma[10], sigma[11], v[ 1], v[ 6], v[11], v[12] ); \
+   G8W( sigma[12], sigma[13], v[ 2], v[ 7], v[ 8], v[13] ); \
+   G8W( sigma[14], sigma[15], v[ 3], v[ 4], v[ 9], v[14] ); \
+} while(0)
+
+/*
 #define ROUND8W(r)  \
 do { \
   G8W( r, 0, v[ 0], v[ 4], v[ 8], v[12] ); \
@@ -231,6 +354,7 @@ do { \
   G8W( r, 6, v[ 2], v[ 7], v[ 8], v[13] ); \
   G8W( r, 7, v[ 3], v[ 4], v[ 9], v[14] ); \
 } while(0)
+*/

   ROUND8W( 0 );
   ROUND8W( 1 );
@@ -267,8 +391,18 @@ int blake2s_8way_init( blake2s_8way_state *S, const uint8_t outlen )
   memset( P->personal, 0, sizeof( P->personal ) );

   memset( S, 0, sizeof( blake2s_8way_state ) );
-   for( int i = 0; i < 8; ++i )
-      S->h[i] = _mm256_set1_epi32( blake2s_IV[i] );
+   S->h[0] = m256_const1_64( 0x6A09E6676A09E667ULL );
+   S->h[1] = m256_const1_64( 0xBB67AE85BB67AE85ULL );
+   S->h[2] = m256_const1_64( 0x3C6EF3723C6EF372ULL );
+   S->h[3] = m256_const1_64( 0xA54FF53AA54FF53AULL );
+   S->h[4] = m256_const1_64( 0x510E527F510E527FULL );
+   S->h[5] = m256_const1_64( 0x9B05688C9B05688CULL );
+   S->h[6] = m256_const1_64( 0x1F83D9AB1F83D9ABULL );
+   S->h[7] = m256_const1_64( 0x5BE0CD195BE0CD19ULL );
+
+
+//   for( int i = 0; i < 8; ++i )
+//      S->h[i] = _mm256_set1_epi32( blake2s_IV[i] );

   uint32_t *p = ( uint32_t * )( P );

@@ -329,9 +463,203 @@ int blake2s_8way_final( blake2s_8way_state *S, void *out, uint8_t outlen )
   return 0;
 }

+// Update and final when inlen is a multiple of 64 bytes
+int blake2s_8way_full_blocks( blake2s_8way_state *S, void *out,
+                              const void *input, uint64_t inlen )
+{
+    __m256i *in = (__m256i*)input;
+    __m256i *buf = (__m256i*)S->buf;
+
+    while( inlen > BLAKE2S_BLOCKBYTES )
+    {
+       memcpy_256( buf, in, BLAKE2S_BLOCKBYTES >> 2 );
+       S->buflen = BLAKE2S_BLOCKBYTES;
+       inlen -= BLAKE2S_BLOCKBYTES;
+       S->t[0] += BLAKE2S_BLOCKBYTES;
+       S->t[1] += ( S->t[0] < BLAKE2S_BLOCKBYTES );
+       blake2s_8way_compress( S, buf );
+       S->buflen = 0;
+       in += ( BLAKE2S_BLOCKBYTES >> 2 );
+    }
+
+    // last block
+    memcpy_256( buf, in, BLAKE2S_BLOCKBYTES >> 2 );
+    S->buflen = BLAKE2S_BLOCKBYTES;
+    S->t[0] += S->buflen;
+    S->t[1] += ( S->t[0] < S->buflen );
+    if ( S->last_node )  S->f[1] = ~0U;
+    S->f[0] = ~0U;
+    blake2s_8way_compress( S, buf );
+
+    for ( int i = 0; i < 8; ++i )
+      casti_m256i( out, i ) = S->h[ i ];
+    return 0;
+}

 #endif // __AVX2__

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// Blake2s-256 16 way
+
+int blake2s_16way_compress( blake2s_16way_state *S, const __m512i *block )
+{
+   __m512i m[16];
+   __m512i v[16];
+
+   memcpy_512( m, block, 16 );
+   memcpy_512( v, S->h, 8 );
+
+   v[ 8] = m512_const1_64( 0x6A09E6676A09E667ULL );
+   v[ 9] = m512_const1_64( 0xBB67AE85BB67AE85ULL );
+   v[10] = m512_const1_64( 0x3C6EF3723C6EF372ULL );
+   v[11] = m512_const1_64( 0xA54FF53AA54FF53AULL );
+   v[12] = _mm512_xor_si512( _mm512_set1_epi32( S->t[0] ),
+                          m512_const1_64( 0x510E527F510E527FULL ) );
+
+   v[13] = _mm512_xor_si512( _mm512_set1_epi32( S->t[1] ),
+                          m512_const1_64( 0x9B05688C9B05688CULL ) );
+
+   v[14] = _mm512_xor_si512( _mm512_set1_epi32( S->f[0] ),
+                          m512_const1_64( 0x1F83D9AB1F83D9ABULL ) );
+
+   v[15] = _mm512_xor_si512( _mm512_set1_epi32( S->f[1] ),
+                          m512_const1_64( 0x5BE0CD195BE0CD19ULL ) );
+
+
+#define G16W( sigma0, sigma1, a, b, c, d) \
+do { \
+   uint8_t s0 = sigma0; \
+   uint8_t s1 = sigma1; \
+   a = _mm512_add_epi32( _mm512_add_epi32( a, b ), m[ s0 ] ); \
+   d = mm512_ror_32( _mm512_xor_si512( d, a ), 16 ); \
+   c = _mm512_add_epi32( c, d ); \
+   b = mm512_ror_32( _mm512_xor_si512( b, c ), 12 ); \
+   a = _mm512_add_epi32( _mm512_add_epi32( a, b ), m[ s1 ] ); \
+   d = mm512_ror_32( _mm512_xor_si512( d, a ),  8 ); \
+   c = _mm512_add_epi32( c, d ); \
+   b = mm512_ror_32( _mm512_xor_si512( b, c ),  7 ); \
+} while(0)
+
+#define ROUND16W(r)  \
+do { \
+   uint8_t *sigma = (uint8_t*)&blake2s_sigma[r]; \
+   G16W( sigma[ 0], sigma[ 1], v[ 0], v[ 4], v[ 8], v[12] ); \
+   G16W( sigma[ 2], sigma[ 3], v[ 1], v[ 5], v[ 9], v[13] ); \
+   G16W( sigma[ 4], sigma[ 5], v[ 2], v[ 6], v[10], v[14] ); \
+   G16W( sigma[ 6], sigma[ 7], v[ 3], v[ 7], v[11], v[15] ); \
+   G16W( sigma[ 8], sigma[ 9], v[ 0], v[ 5], v[10], v[15] ); \
+   G16W( sigma[10], sigma[11], v[ 1], v[ 6], v[11], v[12] ); \
+   G16W( sigma[12], sigma[13], v[ 2], v[ 7], v[ 8], v[13] ); \
+   G16W( sigma[14], sigma[15], v[ 3], v[ 4], v[ 9], v[14] ); \
+} while(0)
+
+   ROUND16W( 0 );
+   ROUND16W( 1 );
+   ROUND16W( 2 );
+   ROUND16W( 3 );
+   ROUND16W( 4 );
+   ROUND16W( 5 );
+   ROUND16W( 6 );
+   ROUND16W( 7 );
+   ROUND16W( 8 );
+   ROUND16W( 9 );
+
+   for( size_t i = 0; i < 8; ++i )
+      S->h[i] = _mm512_xor_si512( _mm512_xor_si512( S->h[i], v[i] ), v[i + 8] );
+
+#undef G16W
+#undef ROUND16W
+   return 0;
+}
+
+int blake2s_16way_init( blake2s_16way_state *S, const uint8_t outlen )
+{
+   blake2s_nway_param P[1];
+
+   P->digest_length = outlen;
+   P->key_length    = 0;
+   P->fanout        = 1;
+   P->depth         = 1;
+   P->leaf_length   = 0;
+   *((uint64_t*)(P->node_offset)) = 0;
+   P->node_depth    = 0;
+   P->inner_length  = 0;
+   memset( P->salt,     0, sizeof( P->salt ) );
+   memset( P->personal, 0, sizeof( P->personal ) );
+
+   memset( S, 0, sizeof( blake2s_16way_state ) );
+   S->h[0] = m512_const1_64( 0x6A09E6676A09E667ULL );
+   S->h[1] = m512_const1_64( 0xBB67AE85BB67AE85ULL );
+   S->h[2] = m512_const1_64( 0x3C6EF3723C6EF372ULL );
+   S->h[3] = m512_const1_64( 0xA54FF53AA54FF53AULL );
+   S->h[4] = m512_const1_64( 0x510E527F510E527FULL );
+   S->h[5] = m512_const1_64( 0x9B05688C9B05688CULL );
+   S->h[6] = m512_const1_64( 0x1F83D9AB1F83D9ABULL );
+   S->h[7] = m512_const1_64( 0x5BE0CD195BE0CD19ULL );
+
+   uint32_t *p = ( uint32_t * )( P );
+
+   /* IV XOR ParamBlock */
+   for ( size_t i = 0; i < 8; ++i )
+      S->h[i] = _mm512_xor_si512( S->h[i], _mm512_set1_epi32( p[i] ) );
+   return 0;
+}
+
+int blake2s_16way_update( blake2s_16way_state *S, const void *in,
+                         uint64_t inlen )
+{
+  __m512i *input = (__m512i*)in;
+  __m512i *buf = (__m512i*)S->buf;
+  const int bsize = BLAKE2S_BLOCKBYTES;
+
+   while( inlen > 0 )
+   {
+      size_t left = S->buflen;
+      if( inlen >= bsize - left )
+      {
+         memcpy_512( buf + (left>>2), input, (bsize - left) >> 2 );
+         S->buflen += bsize - left;
+         S->t[0] += BLAKE2S_BLOCKBYTES;
+         S->t[1] += ( S->t[0] < BLAKE2S_BLOCKBYTES );
+         blake2s_16way_compress( S, buf );
+         S->buflen = 0;
+         input += ( bsize >> 2 );
+         inlen -= bsize;
+      }
+      else
+      {
+          memcpy_512( buf + ( left>>2 ), input, inlen>>2 );
+          S->buflen += (size_t) inlen;
+          input += ( inlen>>2 );
+          inlen -= inlen;
+      }
+   }
+   return 0;
+}
+
+int blake2s_16way_final( blake2s_16way_state *S, void *out, uint8_t outlen )
+{
+   __m512i *buf = (__m512i*)S->buf;
+
+   S->t[0] += S->buflen;
+   S->t[1] += ( S->t[0] < S->buflen );
+   if ( S->last_node )
+      S->f[1] = ~0U;
+   S->f[0] = ~0U;
+
+   memset_zero_512( buf + ( S->buflen>>2 ),
+                    ( BLAKE2S_BLOCKBYTES - S->buflen ) >> 2 );
+   blake2s_16way_compress( S, buf );
+
+   for ( int i = 0; i < 8; ++i )
+      casti_m512i( out, i ) = S->h[ i ];
+   return 0;
+}
+
+#endif   // AVX512
+
+
 #if 0
 int blake2s( uint8_t *out, const void *in, const void *key, const uint8_t outlen, const uint64_t inlen, uint8_t keylen )
 {
--- a/algo/blake/blake2s-hash-4way.h
+++ b/algo/blake/blake2s-hash-4way.h
@@ -14,7 +14,7 @@
 #ifndef __BLAKE2S_HASH_4WAY_H__
 #define __BLAKE2S_HASH_4WAY_H__ 1

-#if defined(__SSE4_2__)
+#if defined(__SSE2__)

 #include "simd-utils.h"

@@ -74,6 +74,9 @@ int blake2s_4way_init( blake2s_4way_state *S, const uint8_t outlen );
 int blake2s_4way_update( blake2s_4way_state *S, const void *in,
                         uint64_t inlen );
 int blake2s_4way_final( blake2s_4way_state *S, void *out, uint8_t outlen );
+int blake2s_4way_full_blocks( blake2s_4way_state *S, void *out,
+                              const void *input, uint64_t inlen );
+

 #if defined(__AVX2__)

@@ -91,6 +94,27 @@ int blake2s_8way_init( blake2s_8way_state *S, const uint8_t outlen );
 int blake2s_8way_update( blake2s_8way_state *S, const void *in,
                         uint64_t inlen );
 int blake2s_8way_final( blake2s_8way_state *S, void *out, uint8_t outlen );
+int blake2s_8way_full_blocks( blake2s_8way_state *S, void *out,
+                              const void *input, uint64_t inlen );
+
+#endif
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+ALIGN( 128 ) typedef struct __blake2s_16way_state
+{
+   __m512i h[8];
+   uint8_t  buf[ BLAKE2S_BLOCKBYTES * 16 ];
+   uint32_t t[2];
+   uint32_t f[2];
+   size_t   buflen;
+   uint8_t  last_node;
+} blake2s_16way_state ;
+
+int blake2s_16way_init( blake2s_16way_state *S, const uint8_t outlen );
+int blake2s_16way_update( blake2s_16way_state *S, const void *in,
+                         uint64_t inlen );
+int blake2s_16way_final( blake2s_16way_state *S, void *out, uint8_t outlen );

 #endif

@@ -107,6 +131,6 @@ int blake2s_8way_final( blake2s_8way_state *S, void *out, uint8_t outlen );
 }
 #endif

-#endif  // __SSE4_2__
+#endif  // __SSE2__

 #endif
--- a/algo/blake/blake2s.c
+++ b/algo/blake/blake2s.c
@@ -70,18 +70,3 @@ int scanhash_blake2s( struct work *work,

 	return 0;
 }
-/*
-// changed to get_max64_0x3fffffLL in cpuminer-multi-decred
-int64_t blake2s_get_max64 ()
-{
-   return 0x7ffffLL;
-}
-
-bool register_blake2s_algo( algo_gate_t* gate )
-{
-  gate->scanhash  = (void*)&scanhash_blake2s;
-  gate->hash      = (void*)&blake2s_hash;
-  gate->get_max64 = (void*)&blake2s_get_max64;
-  return true;
-};
-*/
--- a/algo/blake/blake512-hash-4way.c
+++ b/algo/blake/blake512-hash-4way.c
@@ -42,21 +42,13 @@
 extern "C"{
 #endif

-#if SPH_SMALL_FOOTPRINT && !defined SPH_SMALL_FOOTPRINT_BLAKE
-#define SPH_SMALL_FOOTPRINT_BLAKE   1
-#endif
-
-#if SPH_64 && (SPH_SMALL_FOOTPRINT_BLAKE || !SPH_64_TRUE)
-#define SPH_COMPACT_BLAKE_64   1
-#endif
-
 #ifdef _MSC_VER
 #pragma warning (disable: 4146)
 #endif

-
-// Blake-512
-
+// Blake-512 common
+   
+/*
 static const sph_u64 IV512[8] = {
 	SPH_C64(0x6A09E667F3BCC908), SPH_C64(0xBB67AE8584CAA73B),
 	SPH_C64(0x3C6EF372FE94F82B), SPH_C64(0xA54FF53A5F1D36F1),
@@ -64,10 +56,7 @@ static const sph_u64 IV512[8] = {
 	SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
 };

-
-#if SPH_COMPACT_BLAKE_32 || SPH_COMPACT_BLAKE_64
-
-// Blake-256 4 & 8 way, Blake-512 4 way
+static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };

 static const unsigned sigma[16][16] = {
 	{  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 },
@@ -88,7 +77,17 @@ static const unsigned sigma[16][16] = {
 	{  2, 12,  6, 10,  0, 11,  8,  3,  4, 13,  7,  5, 15, 14,  1,  9 }
 };

-#endif
+static const sph_u64 CB[16] = {
+   SPH_C64(0x243F6A8885A308D3), SPH_C64(0x13198A2E03707344),
+   SPH_C64(0xA4093822299F31D0), SPH_C64(0x082EFA98EC4E6C89),
+   SPH_C64(0x452821E638D01377), SPH_C64(0xBE5466CF34E90C6C),
+   SPH_C64(0xC0AC29B7C97C50DD), SPH_C64(0x3F84D5B5B5470917),
+   SPH_C64(0x9216D5D98979FB1B), SPH_C64(0xD1310BA698DFB5AC),
+   SPH_C64(0x2FFD72DBD01ADFB7), SPH_C64(0xB8E1AFED6A267E96),
+   SPH_C64(0xBA7C9045F12C7F99), SPH_C64(0x24A19947B3916CF7),
+   SPH_C64(0x0801F2E2858EFC16), SPH_C64(0x636920D871574E69)
+
+*/

 #define Z00   0
 #define Z01   1
@@ -264,8 +263,6 @@ static const unsigned sigma[16][16] = {
 #define Mx_(n)      Mx__(n)
 #define Mx__(n)     M ## n

-// Blake-512 4 way
-
 #define CBx(r, i)   CBx_(Z ## r ## i)
 #define CBx_(n)     CBx__(n)
 #define CBx__(n)    CB ## n
@@ -287,60 +284,304 @@ static const unsigned sigma[16][16] = {
 #define CBE   SPH_C64(0x0801F2E2858EFC16)
 #define CBF   SPH_C64(0x636920D871574E69)

-#if SPH_COMPACT_BLAKE_64
-// not used
-static const sph_u64 CB[16] = {
-	SPH_C64(0x243F6A8885A308D3), SPH_C64(0x13198A2E03707344),
-	SPH_C64(0xA4093822299F31D0), SPH_C64(0x082EFA98EC4E6C89),
-	SPH_C64(0x452821E638D01377), SPH_C64(0xBE5466CF34E90C6C),
-	SPH_C64(0xC0AC29B7C97C50DD), SPH_C64(0x3F84D5B5B5470917),
-	SPH_C64(0x9216D5D98979FB1B), SPH_C64(0xD1310BA698DFB5AC),
-	SPH_C64(0x2FFD72DBD01ADFB7), SPH_C64(0xB8E1AFED6A267E96),
-	SPH_C64(0xBA7C9045F12C7F99), SPH_C64(0x24A19947B3916CF7),
-	SPH_C64(0x0801F2E2858EFC16), SPH_C64(0x636920D871574E69)
-};
+#define READ_STATE64(state)   do { \
+      H0 = (state)->H[0]; \
+      H1 = (state)->H[1]; \
+      H2 = (state)->H[2]; \
+      H3 = (state)->H[3]; \
+      H4 = (state)->H[4]; \
+      H5 = (state)->H[5]; \
+      H6 = (state)->H[6]; \
+      H7 = (state)->H[7]; \
+      S0 = (state)->S[0]; \
+      S1 = (state)->S[1]; \
+      S2 = (state)->S[2]; \
+      S3 = (state)->S[3]; \
+      T0 = (state)->T0; \
+      T1 = (state)->T1; \
+   } while (0)

-#endif
+#define WRITE_STATE64(state)   do { \
+      (state)->H[0] = H0; \
+      (state)->H[1] = H1; \
+      (state)->H[2] = H2; \
+      (state)->H[3] = H3; \
+      (state)->H[4] = H4; \
+      (state)->H[5] = H5; \
+      (state)->H[6] = H6; \
+      (state)->H[7] = H7; \
+      (state)->S[0] = S0; \
+      (state)->S[1] = S1; \
+      (state)->S[2] = S2; \
+      (state)->S[3] = S3; \
+      (state)->T0 = T0; \
+      (state)->T1 = T1; \
+   } while (0)

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// Blake-512 8 way AVX512
+
+#define GB_8WAY(m0, m1, c0, c1, a, b, c, d)   do { \
+   a = _mm512_add_epi64( _mm512_add_epi64( _mm512_xor_si512( \
+                 _mm512_set1_epi64( c1 ), m0 ), b ), a ); \
+   d = mm512_ror_64( _mm512_xor_si512( d, a ), 32 ); \
+   c = _mm512_add_epi64( c, d ); \
+   b = mm512_ror_64( _mm512_xor_si512( b, c ), 25 ); \
+   a = _mm512_add_epi64( _mm512_add_epi64( _mm512_xor_si512( \
+                 _mm512_set1_epi64( c0 ), m1 ), b ), a ); \
+   d = mm512_ror_64( _mm512_xor_si512( d, a ), 16 ); \
+   c = _mm512_add_epi64( c, d ); \
+   b = mm512_ror_64( _mm512_xor_si512( b, c ), 11 ); \
+} while (0)
+
+#define ROUND_B_8WAY(r)   do { \
+   GB_8WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
+   GB_8WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
+   GB_8WAY(Mx(r, 4), Mx(r, 5), CBx(r, 4), CBx(r, 5), V2, V6, VA, VE); \
+   GB_8WAY(Mx(r, 6), Mx(r, 7), CBx(r, 6), CBx(r, 7), V3, V7, VB, VF); \
+   GB_8WAY(Mx(r, 8), Mx(r, 9), CBx(r, 8), CBx(r, 9), V0, V5, VA, VF); \
+   GB_8WAY(Mx(r, A), Mx(r, B), CBx(r, A), CBx(r, B), V1, V6, VB, VC); \
+   GB_8WAY(Mx(r, C), Mx(r, D), CBx(r, C), CBx(r, D), V2, V7, V8, VD); \
+   GB_8WAY(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
+   } while (0)
+
+#define DECL_STATE64_8WAY \
+   __m512i H0, H1, H2, H3, H4, H5, H6, H7; \
+        __m512i S0, S1, S2, S3; \
+   sph_u64 T0, T1;
+
+#define COMPRESS64_8WAY   do \
+{ \
+  __m512i M0, M1, M2, M3, M4, M5, M6, M7; \
+  __m512i M8, M9, MA, MB, MC, MD, ME, MF; \
+  __m512i V0, V1, V2, V3, V4, V5, V6, V7; \
+  __m512i V8, V9, VA, VB, VC, VD, VE, VF; \
+  __m512i shuf_bswap64; \
+  V0 = H0; \
+  V1 = H1; \
+  V2 = H2; \
+  V3 = H3; \
+  V4 = H4; \
+  V5 = H5; \
+  V6 = H6; \
+  V7 = H7; \
+  V8 = _mm512_xor_si512( S0, m512_const1_64( CB0 ) );  \
+  V9 = _mm512_xor_si512( S1, m512_const1_64( CB1 ) );  \
+  VA = _mm512_xor_si512( S2, m512_const1_64( CB2 ) );  \
+  VB = _mm512_xor_si512( S3, m512_const1_64( CB3 ) );  \
+  VC = _mm512_xor_si512( _mm512_set1_epi64( T0 ), \
+                         m512_const1_64( CB4 ) );  \
+  VD = _mm512_xor_si512( _mm512_set1_epi64( T0 ), \
+                         m512_const1_64( CB5 ) );  \
+  VE = _mm512_xor_si512( _mm512_set1_epi64( T1 ), \
+                         m512_const1_64( CB6 ) );  \
+  VF = _mm512_xor_si512( _mm512_set1_epi64( T1 ), \
+                         m512_const1_64( CB7 ) );  \
+  shuf_bswap64 = m512_const_64( 0x38393a3b3c3d3e3f, 0x3031323334353637, \
+                                0x28292a2b2c2d2e2f, 0x2021222324252627, \
+                                0x18191a1b1c1d1e1f, 0x1011121314151617, \
+                                0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
+  M0 = _mm512_shuffle_epi8( *(buf+ 0), shuf_bswap64 ); \
+  M1 = _mm512_shuffle_epi8( *(buf+ 1), shuf_bswap64 ); \
+  M2 = _mm512_shuffle_epi8( *(buf+ 2), shuf_bswap64 ); \
+  M3 = _mm512_shuffle_epi8( *(buf+ 3), shuf_bswap64 ); \
+  M4 = _mm512_shuffle_epi8( *(buf+ 4), shuf_bswap64 ); \
+  M5 = _mm512_shuffle_epi8( *(buf+ 5), shuf_bswap64 ); \
+  M6 = _mm512_shuffle_epi8( *(buf+ 6), shuf_bswap64 ); \
+  M7 = _mm512_shuffle_epi8( *(buf+ 7), shuf_bswap64 ); \
+  M8 = _mm512_shuffle_epi8( *(buf+ 8), shuf_bswap64 ); \
+  M9 = _mm512_shuffle_epi8( *(buf+ 9), shuf_bswap64 ); \
+  MA = _mm512_shuffle_epi8( *(buf+10), shuf_bswap64 ); \
+  MB = _mm512_shuffle_epi8( *(buf+11), shuf_bswap64 ); \
+  MC = _mm512_shuffle_epi8( *(buf+12), shuf_bswap64 ); \
+  MD = _mm512_shuffle_epi8( *(buf+13), shuf_bswap64 ); \
+  ME = _mm512_shuffle_epi8( *(buf+14), shuf_bswap64 ); \
+  MF = _mm512_shuffle_epi8( *(buf+15), shuf_bswap64 ); \
+  ROUND_B_8WAY(0); \
+  ROUND_B_8WAY(1); \
+  ROUND_B_8WAY(2); \
+  ROUND_B_8WAY(3); \
+  ROUND_B_8WAY(4); \
+  ROUND_B_8WAY(5); \
+  ROUND_B_8WAY(6); \
+  ROUND_B_8WAY(7); \
+  ROUND_B_8WAY(8); \
+  ROUND_B_8WAY(9); \
+  ROUND_B_8WAY(0); \
+  ROUND_B_8WAY(1); \
+  ROUND_B_8WAY(2); \
+  ROUND_B_8WAY(3); \
+  ROUND_B_8WAY(4); \
+  ROUND_B_8WAY(5); \
+  H0 = mm512_xor4( V8, V0, S0, H0 ); \
+  H1 = mm512_xor4( V9, V1, S1, H1 ); \
+  H2 = mm512_xor4( VA, V2, S2, H2 ); \
+  H3 = mm512_xor4( VB, V3, S3, H3 ); \
+  H4 = mm512_xor4( VC, V4, S0, H4 ); \
+  H5 = mm512_xor4( VD, V5, S1, H5 ); \
+  H6 = mm512_xor4( VE, V6, S2, H6 ); \
+  H7 = mm512_xor4( VF, V7, S3, H7 ); \
+} while (0)
+
+void blake512_8way_init( blake_8way_big_context *sc )
+{
+   __m512i zero = m512_zero;
+   casti_m512i( sc->H, 0 ) = m512_const1_64( 0x6A09E667F3BCC908 );
+   casti_m512i( sc->H, 1 ) = m512_const1_64( 0xBB67AE8584CAA73B );
+   casti_m512i( sc->H, 2 ) = m512_const1_64( 0x3C6EF372FE94F82B );
+   casti_m512i( sc->H, 3 ) = m512_const1_64( 0xA54FF53A5F1D36F1 );
+   casti_m512i( sc->H, 4 ) = m512_const1_64( 0x510E527FADE682D1 );
+   casti_m512i( sc->H, 5 ) = m512_const1_64( 0x9B05688C2B3E6C1F );
+   casti_m512i( sc->H, 6 ) = m512_const1_64( 0x1F83D9ABFB41BD6B );
+   casti_m512i( sc->H, 7 ) = m512_const1_64( 0x5BE0CD19137E2179 );
+
+   casti_m512i( sc->S, 0 ) = zero;
+   casti_m512i( sc->S, 1 ) = zero;
+   casti_m512i( sc->S, 2 ) = zero;
+   casti_m512i( sc->S, 3 ) = zero;
+
+   sc->T0 = sc->T1 = 0;
+   sc->ptr = 0;
+}
+
+static void
+blake64_8way( blake_8way_big_context *sc, const void *data, size_t len )
+{
+   __m512i *vdata = (__m512i*)data;
+   __m512i *buf;
+   size_t ptr;
+   DECL_STATE64_8WAY
+
+   const int buf_size = 128;  //  sizeof/8
+
+   buf = sc->buf;
+   ptr = sc->ptr;
+   if ( len < (buf_size - ptr) )
+   {
+   memcpy_512( buf + (ptr>>3), vdata, len>>3 );
+   ptr += len;
+   sc->ptr = ptr;
+   return;
+   }
+
+   READ_STATE64(sc);
+   while ( len > 0 )
+   {
+   size_t clen;
+
+   clen = buf_size - ptr;
+   if ( clen > len )
+      clen = len;
+   memcpy_512( buf + (ptr>>3), vdata, clen>>3 );
+   ptr += clen;
+   vdata = vdata + (clen>>3);
+   len -= clen;
+   if ( ptr == buf_size )
+        {
+      if ( ( T0 = SPH_T64(T0 + 1024) ) < 1024 )
+         T1 = SPH_T64(T1 + 1);
+      COMPRESS64_8WAY;
+      ptr = 0;
+   }
+   }
+   WRITE_STATE64(sc);
+   sc->ptr = ptr;
+}
+
+static void
+blake64_8way_close( blake_8way_big_context *sc, void *dst )
+{
+   __m512i buf[16];
+   size_t ptr;
+   unsigned bit_len;
+//   uint64_t z, zz;
+   sph_u64 th, tl;
+
+   ptr = sc->ptr;
+   bit_len = ((unsigned)ptr << 3);
+//   z = 0x80 >> n;
+//   zz = ((ub & -z) | z) & 0xFF;
+//   buf[ptr>>3] = _mm512_set1_epi64( zz );
+   buf[ptr>>3] = m512_const1_64( 0x80 );
+   tl = sc->T0 + bit_len;
+   th = sc->T1;
+   if (ptr == 0 )
+   {
+   sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
+   sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
+   }
+   else if ( sc->T0 == 0 )
+   {
+   sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL) + bit_len;
+   sc->T1 = SPH_T64(sc->T1 - 1);
+   }
+   else
+   {
+        sc->T0 -= 1024 - bit_len;
+   }
+   if ( ptr <= 104 )
+   {
+       memset_zero_512( buf + (ptr>>3) + 1, (104-ptr) >> 3 );
+       buf[104>>3] = _mm512_or_si512( buf[104>>3],
+                                 m512_const1_64( 0x0100000000000000ULL ) );
+       buf[112>>3] = m512_const1_64( bswap_64( th ) );
+       buf[120>>3] = m512_const1_64( bswap_64( tl ) );
+
+       blake64_8way( sc, buf + (ptr>>3), 128 - ptr );
+   }
+   else
+  {
+       memset_zero_512( buf + (ptr>>3) + 1, (120 - ptr) >> 3 );
+
+       blake64_8way( sc, buf + (ptr>>3), 128 - ptr );
+       sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
+       sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
+       memset_zero_512( buf, 112>>3 );
+       buf[104>>3] = m512_const1_64( 0x0100000000000000ULL );
+       buf[112>>3] = m512_const1_64( bswap_64( th ) );
+       buf[120>>3] = m512_const1_64( bswap_64( tl ) );
+
+       blake64_8way( sc, buf, 128 );
+   }
+   mm512_block_bswap_64( (__m512i*)dst, sc->H );
+}
+
+void
+blake512_8way_update(void *cc, const void *data, size_t len)
+{
+   blake64_8way(cc, data, len);
+}
+
+void
+blake512_8way_close(void *cc, void *dst)
+{
+   blake512_8way_addbits_and_close(cc, 0, 0, dst);
+}
+
+void
+blake512_8way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
+{
+   blake64_8way_close(cc, dst);
+}
+
+#endif  // AVX512

 // Blake-512 4 way

 #define GB_4WAY(m0, m1, c0, c1, a, b, c, d)   do { \
   a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
-                 _mm256_set_epi64x( c1, c1, c1, c1 ), m0 ), b ), a ); \
+                 _mm256_set1_epi64x( c1 ), m0 ), b ), a ); \
   d = mm256_ror_64( _mm256_xor_si256( d, a ), 32 ); \
   c = _mm256_add_epi64( c, d ); \
   b = mm256_ror_64( _mm256_xor_si256( b, c ), 25 ); \
   a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
-                 _mm256_set_epi64x( c0, c0, c0, c0 ), m1 ), b ), a ); \
+                 _mm256_set1_epi64x( c0 ), m1 ), b ), a ); \
   d = mm256_ror_64( _mm256_xor_si256( d, a ), 16 ); \
   c = _mm256_add_epi64( c, d ); \
   b = mm256_ror_64( _mm256_xor_si256( b, c ), 11 ); \
 } while (0)

-#if SPH_COMPACT_BLAKE_64
-// not used
-#define ROUND_B_4WAY(r)   do { \
-	GB_4WAY(M[sigma[r][0x0]], M[sigma[r][0x1]], \
-		CB[sigma[r][0x0]], CB[sigma[r][0x1]], V0, V4, V8, VC); \
-	GB_4WAY(M[sigma[r][0x2]], M[sigma[r][0x3]], \
-		CB[sigma[r][0x2]], CB[sigma[r][0x3]], V1, V5, V9, VD); \
-	GB_4WAY(M[sigma[r][0x4]], M[sigma[r][0x5]], \
-		CB[sigma[r][0x4]], CB[sigma[r][0x5]], V2, V6, VA, VE); \
-	GB_4WAY(M[sigma[r][0x6]], M[sigma[r][0x7]], \
-		CB[sigma[r][0x6]], CB[sigma[r][0x7]], V3, V7, VB, VF); \
-	GB_4WAY(M[sigma[r][0x8]], M[sigma[r][0x9]], \
-		CB[sigma[r][0x8]], CB[sigma[r][0x9]], V0, V5, VA, VF); \
-	GB_4WAY(M[sigma[r][0xA]], M[sigma[r][0xB]], \
-		CB[sigma[r][0xA]], CB[sigma[r][0xB]], V1, V6, VB, VC); \
-	GB_4WAY(M[sigma[r][0xC]], M[sigma[r][0xD]], \
-		CB[sigma[r][0xC]], CB[sigma[r][0xD]], V2, V7, V8, VD); \
-	GB_4WAY(M[sigma[r][0xE]], M[sigma[r][0xF]], \
-		CB[sigma[r][0xE]], CB[sigma[r][0xF]], V3, V4, V9, VE); \
-} while (0)
-
-#else
-//current_impl
 #define ROUND_B_4WAY(r)   do { \
 	GB_4WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
 	GB_4WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
@@ -352,208 +593,100 @@ static const sph_u64 CB[16] = {
 	GB_4WAY(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
 	} while (0)

-#endif
-
-
-// Blake-512 4 way
-
 #define DECL_STATE64_4WAY \
 	__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
        __m256i S0, S1, S2, S3; \
 	sph_u64 T0, T1;

-#define READ_STATE64_4WAY(state)   do { \
-		H0 = (state)->H[0]; \
-		H1 = (state)->H[1]; \
-		H2 = (state)->H[2]; \
-		H3 = (state)->H[3]; \
-		H4 = (state)->H[4]; \
-		H5 = (state)->H[5]; \
-		H6 = (state)->H[6]; \
-		H7 = (state)->H[7]; \
-		S0 = (state)->S[0]; \
-		S1 = (state)->S[1]; \
-		S2 = (state)->S[2]; \
-		S3 = (state)->S[3]; \
-		T0 = (state)->T0; \
-		T1 = (state)->T1; \
-	} while (0)
+#define COMPRESS64_4WAY   do \
+{ \
+  __m256i M0, M1, M2, M3, M4, M5, M6, M7; \
+  __m256i M8, M9, MA, MB, MC, MD, ME, MF; \
+  __m256i V0, V1, V2, V3, V4, V5, V6, V7; \
+  __m256i V8, V9, VA, VB, VC, VD, VE, VF; \
+  __m256i shuf_bswap64; \
+  V0 = H0; \
+  V1 = H1; \
+  V2 = H2; \
+  V3 = H3; \
+  V4 = H4; \
+  V5 = H5; \
+  V6 = H6; \
+  V7 = H7; \
+  V8 = _mm256_xor_si256( S0, m256_const1_64( CB0 ) );  \
+  V9 = _mm256_xor_si256( S1, m256_const1_64( CB1 ) );  \
+  VA = _mm256_xor_si256( S2, m256_const1_64( CB2 ) );  \
+  VB = _mm256_xor_si256( S3, m256_const1_64( CB3 ) );  \
+  VC = _mm256_xor_si256( _mm256_set1_epi64x( T0 ), \
+                         m256_const1_64( CB4 ) );  \
+  VD = _mm256_xor_si256( _mm256_set1_epi64x( T0 ), \
+                         m256_const1_64( CB5 ) );  \
+  VE = _mm256_xor_si256( _mm256_set1_epi64x( T1 ), \
+                         m256_const1_64( CB6 ) );  \
+  VF = _mm256_xor_si256( _mm256_set1_epi64x( T1 ), \
+                         m256_const1_64( CB7 ) );  \
+  shuf_bswap64 = m256_const_64( 0x18191a1b1c1d1e1f, 0x1011121314151617, \
+                                0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
+  M0 = _mm256_shuffle_epi8( *(buf+ 0), shuf_bswap64 ); \
+  M1 = _mm256_shuffle_epi8( *(buf+ 1), shuf_bswap64 ); \
+  M2 = _mm256_shuffle_epi8( *(buf+ 2), shuf_bswap64 ); \
+  M3 = _mm256_shuffle_epi8( *(buf+ 3), shuf_bswap64 ); \
+  M4 = _mm256_shuffle_epi8( *(buf+ 4), shuf_bswap64 ); \
+  M5 = _mm256_shuffle_epi8( *(buf+ 5), shuf_bswap64 ); \
+  M6 = _mm256_shuffle_epi8( *(buf+ 6), shuf_bswap64 ); \
+  M7 = _mm256_shuffle_epi8( *(buf+ 7), shuf_bswap64 ); \
+  M8 = _mm256_shuffle_epi8( *(buf+ 8), shuf_bswap64 ); \
+  M9 = _mm256_shuffle_epi8( *(buf+ 9), shuf_bswap64 ); \
+  MA = _mm256_shuffle_epi8( *(buf+10), shuf_bswap64 ); \
+  MB = _mm256_shuffle_epi8( *(buf+11), shuf_bswap64 ); \
+  MC = _mm256_shuffle_epi8( *(buf+12), shuf_bswap64 ); \
+  MD = _mm256_shuffle_epi8( *(buf+13), shuf_bswap64 ); \
+  ME = _mm256_shuffle_epi8( *(buf+14), shuf_bswap64 ); \
+  MF = _mm256_shuffle_epi8( *(buf+15), shuf_bswap64 ); \
+  ROUND_B_4WAY(0); \
+  ROUND_B_4WAY(1); \
+  ROUND_B_4WAY(2); \
+  ROUND_B_4WAY(3); \
+  ROUND_B_4WAY(4); \
+  ROUND_B_4WAY(5); \
+  ROUND_B_4WAY(6); \
+  ROUND_B_4WAY(7); \
+  ROUND_B_4WAY(8); \
+  ROUND_B_4WAY(9); \
+  ROUND_B_4WAY(0); \
+  ROUND_B_4WAY(1); \
+  ROUND_B_4WAY(2); \
+  ROUND_B_4WAY(3); \
+  ROUND_B_4WAY(4); \
+  ROUND_B_4WAY(5); \
+  H0 = mm256_xor4( V8, V0, S0, H0 ); \
+  H1 = mm256_xor4( V9, V1, S1, H1 ); \
+  H2 = mm256_xor4( VA, V2, S2, H2 ); \
+  H3 = mm256_xor4( VB, V3, S3, H3 ); \
+  H4 = mm256_xor4( VC, V4, S0, H4 ); \
+  H5 = mm256_xor4( VD, V5, S1, H5 ); \
+  H6 = mm256_xor4( VE, V6, S2, H6 ); \
+  H7 = mm256_xor4( VF, V7, S3, H7 ); \
+} while (0)

-#define WRITE_STATE64_4WAY(state)   do { \
-		(state)->H[0] = H0; \
-		(state)->H[1] = H1; \
-		(state)->H[2] = H2; \
-		(state)->H[3] = H3; \
-		(state)->H[4] = H4; \
-		(state)->H[5] = H5; \
-		(state)->H[6] = H6; \
-		(state)->H[7] = H7; \
-		(state)->S[0] = S0; \
-		(state)->S[1] = S1; \
-		(state)->S[2] = S2; \
-		(state)->S[3] = S3; \
-		(state)->T0 = T0; \
-		(state)->T1 = T1; \
-	} while (0)

-#if SPH_COMPACT_BLAKE_64
-
-// not used
-#define COMPRESS64_4WAY   do { \
-	__m256i M[16]; \
-	__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
-	__m256i V8, V9, VA, VB, VC, VD, VE, VF; \
-	unsigned r; \
-	V0 = H0; \
-	V1 = H1; \
-	V2 = H2; \
-	V3 = H3; \
-	V4 = H4; \
-	V5 = H5; \
-	V6 = H6; \
-	V7 = H7; \
-        V8 = _mm256_xor_si256( S0, _mm256_set_epi64x( CB0, CB0, CB0, CB0 ) ); \
-        V9 = _mm256_xor_si256( S1, _mm256_set_epi64x( CB1, CB1, CB1, CB1 ) ); \
-        VA = _mm256_xor_si256( S2, _mm256_set_epi64x( CB2, CB2, CB2, CB2 ) ); \
-        VB = _mm256_xor_si256( S3, _mm256_set_epi64x( CB3, CB3, CB3, CB3 ) ); \
-        VC = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
-                               _mm256_set_epi64x( CB4, CB4, CB4, CB4 ) ); \
-        VD = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
-                               _mm256_set_epi64x( CB5, CB5, CB5, CB5 ) ); \
-        VE = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
-                               _mm256_set_epi64x( CB6, CB6, CB6, CB6 ) ); \
-        VF = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
-                               _mm256_set_epi64x( CB7, CB7, CB7, CB7 ) ); \
-	M[0x0] = mm256_bswap_64( *(buf+0) ); \
-	M[0x1] = mm256_bswap_64( *(buf+1) ); \
-	M[0x2] = mm256_bswap_64( *(buf+2) ); \
-	M[0x3] = mm256_bswap_64( *(buf+3) ); \
-	M[0x4] = mm256_bswap_64( *(buf+4) ); \
-	M[0x5] = mm256_bswap_64( *(buf+5) ); \
-	M[0x6] = mm256_bswap_64( *(buf+6) ); \
-	M[0x7] = mm256_bswap_64( *(buf+7) ); \
-	M[0x8] = mm256_bswap_64( *(buf+8) ); \
-	M[0x9] = mm256_bswap_64( *(buf+9) ); \
-	M[0xA] = mm256_bswap_64( *(buf+10) ); \
-	M[0xB] = mm256_bswap_64( *(buf+11) ); \
-	M[0xC] = mm256_bswap_64( *(buf+12) ); \
-	M[0xD] = mm256_bswap_64( *(buf+13) ); \
-	M[0xE] = mm256_bswap_64( *(buf+14) ); \
-	M[0xF] = mm256_bswap_64( *(buf+15) ); \
-	for (r = 0; r < 16; r ++) \
-		ROUND_B_4WAY(r); \
-        H0 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
-        H1 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
-        H2 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S2, V2 ), VA ), H2 ); \
-        H3 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S3, V3 ), VB ), H3 ); \
-        H4 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S0, V4 ), VC ), H4 ); \
-        H5 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S1, V5 ), VD ), H5 ); \
-        H6 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S2, V6 ), VE ), H6 ); \
-        H7 = _mm256_xor_si256( _mm256_xor_si256( \
-                    _mm256_xor_si256( S3, V7 ), VF ), H7 ); \
-	} while (0)
-
-#else
-
-//current impl
-
-#define COMPRESS64_4WAY   do { \
-     __m256i M0, M1, M2, M3, M4, M5, M6, M7; \
-     __m256i M8, M9, MA, MB, MC, MD, ME, MF; \
-     __m256i V0, V1, V2, V3, V4, V5, V6, V7; \
-     __m256i V8, V9, VA, VB, VC, VD, VE, VF; \
-     V0 = H0; \
-     V1 = H1; \
-     V2 = H2; \
-     V3 = H3; \
-     V4 = H4; \
-     V5 = H5; \
-     V6 = H6; \
-     V7 = H7; \
-     V8 = _mm256_xor_si256( S0, _mm256_set_epi64x( CB0, CB0, CB0, CB0 ) );  \
-     V9 = _mm256_xor_si256( S1, _mm256_set_epi64x( CB1, CB1, CB1, CB1 ) );  \
-     VA = _mm256_xor_si256( S2, _mm256_set_epi64x( CB2, CB2, CB2, CB2 ) );  \
-     VB = _mm256_xor_si256( S3, _mm256_set_epi64x( CB3, CB3, CB3, CB3 ) );  \
-     VC = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
-                            _mm256_set_epi64x( CB4, CB4, CB4, CB4 ) );  \
-     VD = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
-                            _mm256_set_epi64x( CB5, CB5, CB5, CB5 ) );  \
-     VE = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
-                            _mm256_set_epi64x( CB6, CB6, CB6, CB6 ) );  \
-     VF = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
-                            _mm256_set_epi64x( CB7, CB7, CB7, CB7 ) );  \
-     M0 = mm256_bswap_64( *(buf + 0) ); \
-     M1 = mm256_bswap_64( *(buf + 1) ); \
-     M2 = mm256_bswap_64( *(buf + 2) ); \
-     M3 = mm256_bswap_64( *(buf + 3) ); \
-     M4 = mm256_bswap_64( *(buf + 4) ); \
-     M5 = mm256_bswap_64( *(buf + 5) ); \
-     M6 = mm256_bswap_64( *(buf + 6) ); \
-     M7 = mm256_bswap_64( *(buf + 7) ); \
-     M8 = mm256_bswap_64( *(buf + 8) ); \
-     M9 = mm256_bswap_64( *(buf + 9) ); \
-     MA = mm256_bswap_64( *(buf + 10) ); \
-     MB = mm256_bswap_64( *(buf + 11) ); \
-     MC = mm256_bswap_64( *(buf + 12) ); \
-     MD = mm256_bswap_64( *(buf + 13) ); \
-     ME = mm256_bswap_64( *(buf + 14) ); \
-     MF = mm256_bswap_64( *(buf + 15) ); \
-     ROUND_B_4WAY(0); \
-     ROUND_B_4WAY(1); \
-     ROUND_B_4WAY(2); \
-     ROUND_B_4WAY(3); \
-     ROUND_B_4WAY(4); \
-     ROUND_B_4WAY(5); \
-     ROUND_B_4WAY(6); \
-     ROUND_B_4WAY(7); \
-     ROUND_B_4WAY(8); \
-     ROUND_B_4WAY(9); \
-     ROUND_B_4WAY(0); \
-     ROUND_B_4WAY(1); \
-     ROUND_B_4WAY(2); \
-     ROUND_B_4WAY(3); \
-     ROUND_B_4WAY(4); \
-     ROUND_B_4WAY(5); \
-     H0 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
-     H1 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
-     H2 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S2, V2 ), VA ), H2 ); \
-     H3 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S3, V3 ), VB ), H3 ); \
-     H4 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S0, V4 ), VC ), H4 ); \
-     H5 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S1, V5 ), VD ), H5 ); \
-     H6 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S2, V6 ), VE ), H6 ); \
-     H7 = _mm256_xor_si256( _mm256_xor_si256( \
-                            _mm256_xor_si256( S3, V7 ), VF ), H7 ); \
-	} while (0)
-
-#endif
-
-static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };
-
-static void
-blake64_4way_init( blake_4way_big_context *sc, const sph_u64 *iv,
-              const sph_u64 *salt )
+void blake512_4way_init( blake_4way_big_context *sc )
 {
-        int i;
-        for ( i = 0; i < 8; i++ )
-           sc->H[i] = _mm256_set1_epi64x( iv[i] );
-        for ( i = 0; i < 4; i++ )
-           sc->S[i] = _mm256_set1_epi64x( salt[i] );
-        sc->T0 = sc->T1 = 0;
-        sc->ptr = 0;
+   __m256i zero = m256_zero;
+   casti_m256i( sc->H, 0 ) = m256_const1_64( 0x6A09E667F3BCC908 );
+   casti_m256i( sc->H, 1 ) = m256_const1_64( 0xBB67AE8584CAA73B );
+   casti_m256i( sc->H, 2 ) = m256_const1_64( 0x3C6EF372FE94F82B );
+   casti_m256i( sc->H, 3 ) = m256_const1_64( 0xA54FF53A5F1D36F1 );
+   casti_m256i( sc->H, 4 ) = m256_const1_64( 0x510E527FADE682D1 );
+   casti_m256i( sc->H, 5 ) = m256_const1_64( 0x9B05688C2B3E6C1F );
+   casti_m256i( sc->H, 6 ) = m256_const1_64( 0x1F83D9ABFB41BD6B );
+   casti_m256i( sc->H, 7 ) = m256_const1_64( 0x5BE0CD19137E2179 );
+   casti_m256i( sc->S, 0 ) = zero;
+   casti_m256i( sc->S, 1 ) = zero;
+   casti_m256i( sc->S, 2 ) = zero;
+   casti_m256i( sc->S, 3 ) = zero;
+   sc->T0 = sc->T1 = 0;
+   sc->ptr = 0;
 }

 static void
@@ -576,7 +709,7 @@ blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
 	return;
   }

-   READ_STATE64_4WAY(sc);
+   READ_STATE64(sc);
   while ( len > 0 )
   {
 	size_t clen;
@@ -596,29 +729,21 @@ blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
 		ptr = 0;
 	}
   }
-   WRITE_STATE64_4WAY(sc);
+   WRITE_STATE64(sc);
   sc->ptr = ptr;
 }

 static void
-blake64_4way_close( blake_4way_big_context *sc,
-	unsigned ub, unsigned n, void *dst, size_t out_size_w64)
+blake64_4way_close( blake_4way_big_context *sc, void *dst )
 {
-//   union {
-      __m256i buf[16];
-//      sph_u64 dummy;
-//   } u;
-   size_t ptr, k;
+   __m256i buf[16];
+   size_t ptr;
   unsigned bit_len;
-   uint64_t z, zz;
   sph_u64 th, tl;
-   __m256i *out;

   ptr = sc->ptr;
   bit_len = ((unsigned)ptr << 3);
-   z = 0x80 >> n;
-   zz = ((ub & -z) | z) & 0xFF;
-   buf[ptr>>3] = _mm256_set_epi64x( zz, zz, zz, zz );
+   buf[ptr>>3] = m256_const1_64( 0x80 );
   tl = sc->T0 + bit_len;
   th = sc->T1;
   if (ptr == 0 )
@@ -635,49 +760,44 @@ blake64_4way_close( blake_4way_big_context *sc,
   {
        sc->T0 -= 1024 - bit_len;
   }
+
   if ( ptr <= 104 )
   {
       memset_zero_256( buf + (ptr>>3) + 1, (104-ptr) >> 3 );
-       if ( out_size_w64 == 8 )
-          buf[(104>>3)] = _mm256_or_si256( buf[(104>>3)],
-                                 _mm256_set1_epi64x( 0x0100000000000000ULL ) );
-       *(buf+(112>>3)) = mm256_bswap_64(
-                                    _mm256_set_epi64x( th, th, th, th ) );
-       *(buf+(120>>3)) = mm256_bswap_64(
-                                    _mm256_set_epi64x( tl, tl, tl, tl ) );
+       buf[104>>3] = _mm256_or_si256( buf[104>>3],
+                                 m256_const1_64( 0x0100000000000000ULL ) );
+       buf[112>>3] = m256_const1_64( bswap_64( th ) );
+       buf[120>>3] = m256_const1_64( bswap_64( tl ) );

       blake64_4way( sc, buf + (ptr>>3), 128 - ptr );
   }
   else
-  {
+   {
       memset_zero_256( buf + (ptr>>3) + 1, (120 - ptr) >> 3 );

       blake64_4way( sc, buf + (ptr>>3), 128 - ptr );
       sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
       sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
       memset_zero_256( buf, 112>>3 ); 
-       if ( out_size_w64 == 8 )
-           buf[104>>3] = _mm256_set1_epi64x( 0x0100000000000000ULL );
-       *(buf+(112>>3)) = mm256_bswap_64(
-                                    _mm256_set_epi64x( th, th, th, th ) );
-       *(buf+(120>>3)) = mm256_bswap_64(
-                                    _mm256_set_epi64x( tl, tl, tl, tl ) );
+       buf[104>>3] = m256_const1_64( 0x0100000000000000ULL );
+       buf[112>>3] = m256_const1_64( bswap_64( th ) );
+       buf[120>>3] = m256_const1_64( bswap_64( tl ) );

       blake64_4way( sc, buf, 128 );
   }
-   out = (__m256i*)dst;
-   for ( k = 0; k < out_size_w64; k++ )
-       out[k] = mm256_bswap_64( sc->H[k] );
+   mm256_block_bswap_64( (__m256i*)dst, sc->H );
 }

+/*
 void
 blake512_4way_init(void *cc)
 {
 	blake64_4way_init(cc, IV512, salt_zero_big);
 }
+*/

 void
-blake512_4way(void *cc, const void *data, size_t len)
+blake512_4way_update(void *cc, const void *data, size_t len)
 {
 	blake64_4way(cc, data, len);
 }
@@ -685,15 +805,18 @@ blake512_4way(void *cc, const void *data, size_t len)
 void
 blake512_4way_close(void *cc, void *dst)
 {
-	blake512_4way_addbits_and_close(cc, 0, 0, dst);
+   blake64_4way_close( cc, dst );
+
+//   blake512_4way_addbits_and_close(cc, dst);
 }

+/*
 void
 blake512_4way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
 {
 	blake64_4way_close(cc, ub, n, dst, 8);
 }
-
+*/
 #ifdef __cplusplus
 }
 #endif
--- a/algo/blake/blakecoin-4way.c
+++ b/algo/blake/blakecoin-4way.c
@@ -29,23 +29,18 @@ int scanhash_blakecoin_4way( struct work *work, uint32_t max_nonce,
   uint32_t *ptarget = work->target;
   const uint32_t first_nonce = pdata[19];
   uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
   uint32_t n = first_nonce;
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
   int thr_id = mythr->id;  // thr_id arg is deprecated
   if ( opt_benchmark )
      HTarget = 0x7f;

-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
   blake256r8_4way_init( &blakecoin_4w_ctx );
   blake256r8_4way( &blakecoin_4w_ctx, vdata, 64 );

-   uint32_t *noncep = vdata + 76;   // 19*4
   do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
      pdata[19] = n;
      blakecoin_4way_hash( hash, vdata );

@@ -79,9 +74,8 @@ void blakecoin_8way_hash( void *state, const void *input )
     blake256r8_8way( &ctx, input + (64<<3), 16 );
     blake256r8_8way_close( &ctx, vhash );

-     mm256_dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
-                              state+128, state+160, state+192, state+224,
-                              vhash, 256 );
+     dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96, state+128,
+                   state+160, state+192, state+224, vhash, 256 );
 }

 int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
@@ -93,29 +87,19 @@ int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
   uint32_t *ptarget = work->target;
   const uint32_t first_nonce = pdata[19];
   uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
   uint32_t n = first_nonce;
-   uint32_t *noncep = vdata + 152;   // 19*8
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
   int thr_id = mythr->id;  // thr_id arg is deprecated
   if ( opt_benchmark )
      HTarget = 0x7f;

-   // we need big endian data...
-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
   blake256r8_8way_init( &blakecoin_8w_ctx );
   blake256r8_8way( &blakecoin_8w_ctx, vdata, 64 );

   do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
-      be32enc( noncep +4, n+4 );
-      be32enc( noncep +5, n+5 );
-      be32enc( noncep +6, n+6 );
-      be32enc( noncep +7, n+7 );
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
      pdata[19] = n;
      blakecoin_8way_hash( hash, vdata );

--- a/algo/blake/blakecoin-gate.c
+++ b/algo/blake/blakecoin-gate.c
@@ -1,13 +1,6 @@
 #include "blakecoin-gate.h"
 #include <memory.h>

-// changed to get_max64_0x3fffffLL in cpuminer-multi-decred
-int64_t blakecoin_get_max64 ()
-{
-  return 0x7ffffLL;
-//  return 0x3fffffLL;
-}
-
 // vanilla uses default gen merkle root, otherwise identical to blakecoin
 bool register_vanilla_algo( algo_gate_t* gate )
 {
@@ -23,7 +16,6 @@ bool register_vanilla_algo( algo_gate_t* gate )
  gate->hash     = (void*)&blakecoinhash;
 #endif
  gate->optimizations = SSE42_OPT | AVX2_OPT;
-  gate->get_max64 = (void*)&blakecoin_get_max64;
  return true;
 }

--- a/algo/blake/blakecoin.c
+++ b/algo/blake/blakecoin.c
@@ -93,33 +93,3 @@ int scanhash_blakecoin( struct work *work, uint32_t max_nonce,
 	return 0;
 }

-/*
-void blakecoin_gen_merkle_root ( char* merkle_root, struct stratum_ctx* sctx )
-{
- SHA256( sctx->job.coinbase, (int)sctx->job.coinbase_size, merkle_root );
-}
-*/
-/*
-// changed to get_max64_0x3fffffLL in cpuminer-multi-decred
-int64_t blakecoin_get_max64 ()
-{
-  return 0x7ffffLL;
-}
-
-// vanilla uses default gen merkle root, otherwise identical to blakecoin
-bool register_vanilla_algo( algo_gate_t* gate )
-{
-    gate->scanhash = (void*)&scanhash_blakecoin;
-    gate->hash     = (void*)&blakecoinhash;
-    gate->get_max64 = (void*)&blakecoin_get_max64;
-    blakecoin_init( &blake_init_ctx );
-    return true;
-}
-
-bool register_blakecoin_algo( algo_gate_t* gate )
-{
-  register_vanilla_algo( gate );
-  gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
-  return true;
-}
-*/
--- a/algo/blake/decred-gate.c
+++ b/algo/blake/decred-gate.c
@@ -38,7 +38,7 @@ void decred_decode_extradata( struct work* work, uint64_t* net_blocks )
   if (!have_longpoll && work->height > *net_blocks + 1)
   {
      char netinfo[64] = { 0 };
-      if (opt_showdiff && net_diff > 0.)
+      if ( net_diff > 0. )
      {
         if (net_diff != work->targetdiff)
            sprintf(netinfo, ", diff %.3f, target %.1f", net_diff,
@@ -116,7 +116,7 @@ void decred_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
   // block header suffix from coinb2 (stake version)
   memcpy( &g_work->data[44],
           &sctx->job.coinbase[ sctx->job.coinbase_size-4 ], 4 );
-   sctx->bloc_height = g_work->data[32];
+   sctx->block_height = g_work->data[32];
   //applog_hex(work->data, 180);
   //applog_hex(&work->data[36], 36);
 }
@@ -154,7 +154,6 @@ bool register_decred_algo( algo_gate_t* gate )
 #endif
  gate->optimizations = AVX2_OPT;
  gate->get_nonceptr          = (void*)&decred_get_nonceptr;
-  gate->get_max64             = (void*)&get_max64_0x3fffffLL;
  gate->decode_extra_data     = (void*)&decred_decode_extradata;
  gate->build_stratum_request = (void*)&decred_be_build_stratum_request;
  gate->work_decode           = (void*)&std_be_work_decode;
--- a/algo/blake/decred.c
+++ b/algo/blake/decred.c
@@ -143,7 +143,7 @@ void decred_decode_extradata( struct work* work, uint64_t* net_blocks )
   if (!have_longpoll && work->height > *net_blocks + 1)
   {
      char netinfo[64] = { 0 };
-      if (opt_showdiff && net_diff > 0.)
+      if (net_diff > 0.)
      {
         if (net_diff != work->targetdiff)
 	    sprintf(netinfo, ", diff %.3f, target %.1f", net_diff,
@@ -269,7 +269,6 @@ bool register_decred_algo( algo_gate_t* gate )
  gate->scanhash              = (void*)&scanhash_decred;
  gate->hash                  = (void*)&decred_hash;
  gate->get_nonceptr          = (void*)&decred_get_nonceptr;
-  gate->get_max64             = (void*)&get_max64_0x3fffffLL;
  gate->decode_extra_data     = (void*)&decred_decode_extradata;
  gate->build_stratum_request = (void*)&decred_be_build_stratum_request;
  gate->work_decode           = (void*)&std_be_work_decode;
--- a/algo/blake/pentablake-4way.c
+++ b/algo/blake/pentablake-4way.c
@@ -10,13 +10,8 @@
 #include "blake-hash-4way.h"
 #include "sph_blake.h"

-//#define DEBUG_ALGO
-
 extern void pentablakehash_4way( void *output, const void *input )
 {
-//	unsigned char _ALIGN(32) hash[128];
-//	// same as uint32_t hashA[16], hashB[16];
-//	#define hashB hash+64

     uint64_t hash0[8] __attribute__ ((aligned (64)));
     uint64_t hash1[8] __attribute__ ((aligned (64)));
@@ -29,22 +24,7 @@ extern void pentablakehash_4way( void *output, const void *input )
     blake512_4way_init( &ctx );
     blake512_4way( &ctx, input, 80 );
     blake512_4way_close( &ctx, vhash );
-/*
-uint64_t sin0[10], sin1[10], sin2[10], sin3[10];
-mm256_deinterleave_4x64( sin0, sin1, sin2, sin3, input, 640 );
-sph_blake512_context ctx2_blake;
-sph_blake512_init(&ctx2_blake);
-sph_blake512(&ctx2_blake, sin0, 80);
-sph_blake512_close(&ctx2_blake, (void*) hash);

-mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
-uint64_t* hash64 = (uint64_t*)hash;
-for( int i = 0; i < 8; i++ )
-{
-   if ( hash0[i] != hash64[i] )
-      printf("hash mismatch %u\n",i);
-}
-*/
     blake512_4way_init( &ctx );
     blake512_4way( &ctx, vhash, 64 );
     blake512_4way_close( &ctx, vhash );
@@ -61,42 +41,10 @@ for( int i = 0; i < 8; i++ )
     blake512_4way( &ctx, vhash, 64 );
     blake512_4way_close( &ctx, vhash );

-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
     memcpy( output,    hash0, 32 );
     memcpy( output+32, hash1, 32 );
     memcpy( output+64, hash2, 32 );
     memcpy( output+96, hash3, 32 );
-
-/*
-     uint64_t sin0[10] __attribute__ ((aligned (64)));
-     uint64_t sin1[10] __attribute__ ((aligned (64)));
-     uint64_t sin2[10] __attribute__ ((aligned (64)));
-     uint64_t sin3[10] __attribute__ ((aligned (64)));
-
-	sph_blake512_context     ctx_blake;
-
-	sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, input, 80);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-	memcpy(output, hash, 32);
-*/
 }

 int scanhash_pentablake_4way( struct work *work,
@@ -137,7 +85,7 @@ int scanhash_pentablake_4way( struct work *work,
    swab32_array( endiandata, pdata, 20 );

    uint64_t *edata = (uint64_t*)endiandata;
-    mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+    intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );

    for ( int m=0; m < 6; m++ )
    {
--- a/algo/blake/pentablake-gate.c
+++ b/algo/blake/pentablake-gate.c
@@ -10,7 +10,6 @@ bool register_pentablake_algo( algo_gate_t* gate )
    gate->hash      = (void*)&pentablakehash;
 #endif
    gate->optimizations = AVX2_OPT;
-    gate->get_max64 = (void*)&get_max64_0x3ffff;
    return true;
 };

--- a/algo/blake/sph_blake2b.c
+++ b/algo/blake/sph_blake2b.c
@@ -103,7 +103,6 @@ static void blake2b_compress( sph_blake2b_ctx *ctx, int last )
 	v[13] ^= ctx->t[1];                 // high 64 bits
 	if (last)                           // last block flag set ?
 		v[14] = ~v[14];
-
 	for (i = 0; i < 16; i++)            // get little-endian words
 		m[i] = B2B_GET64(&ctx->b[8 * i]);

@@ -184,7 +183,8 @@ void sph_blake2b_final( sph_blake2b_ctx *ctx, void *out )

 	while (ctx->c < 128)                // fill up with zeros
 		ctx->b[ctx->c++] = 0;
-	blake2b_compress(ctx, 1);           // final block flag = 1
+
+   blake2b_compress(ctx, 1);           // final block flag = 1

 	// little endian convert and store
 	for (i = 0; i < ctx->outlen; i++) {
--- a/algo/blake/sse2/blake.c
+++ b/algo/blake/sse2/blake.c
@@ -1,476 +0,0 @@
-/* $Id: blake.c 252 2011-06-07 17:55:14Z tp $ */
-/*
- * BLAKE implementation.
- *
- * ==========================(LICENSE BEGIN)============================
- *
- * Copyright (c) 2007-2010  Projet RNRT SAPHIR
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
- * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * ===========================(LICENSE END)=============================
- *
- * @author   Thomas Pornin <thomas.pornin@cryptolog.com>
- */
-#include <stddef.h>
-#include <string.h>
-#include <limits.h>
-
-#include "../sph_blake.h"
-
-#ifdef __cplusplus
-extern "C"{
-#endif
-
-#ifdef _MSC_VER
-#pragma warning (disable: 4146)
-#endif
-
-static const sph_u64 blkIV512[8] = {
-	SPH_C64(0x6A09E667F3BCC908), SPH_C64(0xBB67AE8584CAA73B),
-	SPH_C64(0x3C6EF372FE94F82B), SPH_C64(0xA54FF53A5F1D36F1),
-	SPH_C64(0x510E527FADE682D1), SPH_C64(0x9B05688C2B3E6C1F),
-	SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
-};
-
-#define Z00   0
-#define Z01   1
-#define Z02   2
-#define Z03   3
-#define Z04   4
-#define Z05   5
-#define Z06   6
-#define Z07   7
-#define Z08   8
-#define Z09   9
-#define Z0A   A
-#define Z0B   B
-#define Z0C   C
-#define Z0D   D
-#define Z0E   E
-#define Z0F   F
-
-#define Z10   E
-#define Z11   A
-#define Z12   4
-#define Z13   8
-#define Z14   9
-#define Z15   F
-#define Z16   D
-#define Z17   6
-#define Z18   1
-#define Z19   C
-#define Z1A   0
-#define Z1B   2
-#define Z1C   B
-#define Z1D   7
-#define Z1E   5
-#define Z1F   3
-
-#define Z20   B
-#define Z21   8
-#define Z22   C
-#define Z23   0
-#define Z24   5
-#define Z25   2
-#define Z26   F
-#define Z27   D
-#define Z28   A
-#define Z29   E
-#define Z2A   3
-#define Z2B   6
-#define Z2C   7
-#define Z2D   1
-#define Z2E   9
-#define Z2F   4
-
-#define Z30   7
-#define Z31   9
-#define Z32   3
-#define Z33   1
-#define Z34   D
-#define Z35   C
-#define Z36   B
-#define Z37   E
-#define Z38   2
-#define Z39   6
-#define Z3A   5
-#define Z3B   A
-#define Z3C   4
-#define Z3D   0
-#define Z3E   F
-#define Z3F   8
-
-#define Z40   9
-#define Z41   0
-#define Z42   5
-#define Z43   7
-#define Z44   2
-#define Z45   4
-#define Z46   A
-#define Z47   F
-#define Z48   E
-#define Z49   1
-#define Z4A   B
-#define Z4B   C
-#define Z4C   6
-#define Z4D   8
-#define Z4E   3
-#define Z4F   D
-
-#define Z50   2
-#define Z51   C
-#define Z52   6
-#define Z53   A
-#define Z54   0
-#define Z55   B
-#define Z56   8
-#define Z57   3
-#define Z58   4
-#define Z59   D
-#define Z5A   7
-#define Z5B   5
-#define Z5C   F
-#define Z5D   E
-#define Z5E   1
-#define Z5F   9
-
-#define Z60   C
-#define Z61   5
-#define Z62   1
-#define Z63   F
-#define Z64   E
-#define Z65   D
-#define Z66   4
-#define Z67   A
-#define Z68   0
-#define Z69   7
-#define Z6A   6
-#define Z6B   3
-#define Z6C   9
-#define Z6D   2
-#define Z6E   8
-#define Z6F   B
-
-#define Z70   D
-#define Z71   B
-#define Z72   7
-#define Z73   E
-#define Z74   C
-#define Z75   1
-#define Z76   3
-#define Z77   9
-#define Z78   5
-#define Z79   0
-#define Z7A   F
-#define Z7B   4
-#define Z7C   8
-#define Z7D   6
-#define Z7E   2
-#define Z7F   A
-
-#define Z80   6
-#define Z81   F
-#define Z82   E
-#define Z83   9
-#define Z84   B
-#define Z85   3
-#define Z86   0
-#define Z87   8
-#define Z88   C
-#define Z89   2
-#define Z8A   D
-#define Z8B   7
-#define Z8C   1
-#define Z8D   4
-#define Z8E   A
-#define Z8F   5
-
-#define Z90   A
-#define Z91   2
-#define Z92   8
-#define Z93   4
-#define Z94   7
-#define Z95   6
-#define Z96   1
-#define Z97   5
-#define Z98   F
-#define Z99   B
-#define Z9A   9
-#define Z9B   E
-#define Z9C   3
-#define Z9D   C
-#define Z9E   D
-#define Z9F   0
-
-#define Mx(r, i)    Mx_(Z ## r ## i)
-#define Mx_(n)      Mx__(n)
-#define Mx__(n)     M ## n
-
-#define CSx(r, i)   CSx_(Z ## r ## i)
-#define CSx_(n)     CSx__(n)
-#define CSx__(n)    CS ## n
-
-#define CS0   SPH_C32(0x243F6A88)
-#define CS1   SPH_C32(0x85A308D3)
-#define CS2   SPH_C32(0x13198A2E)
-#define CS3   SPH_C32(0x03707344)
-#define CS4   SPH_C32(0xA4093822)
-#define CS5   SPH_C32(0x299F31D0)
-#define CS6   SPH_C32(0x082EFA98)
-#define CS7   SPH_C32(0xEC4E6C89)
-#define CS8   SPH_C32(0x452821E6)
-#define CS9   SPH_C32(0x38D01377)
-#define CSA   SPH_C32(0xBE5466CF)
-#define CSB   SPH_C32(0x34E90C6C)
-#define CSC   SPH_C32(0xC0AC29B7)
-#define CSD   SPH_C32(0xC97C50DD)
-#define CSE   SPH_C32(0x3F84D5B5)
-#define CSF   SPH_C32(0xB5470917)
-
-
-
-#define CBx(r, i)   CBx_(Z ## r ## i)
-#define CBx_(n)     CBx__(n)
-#define CBx__(n)    CB ## n
-
-#define CB0   SPH_C64(0x243F6A8885A308D3)
-#define CB1   SPH_C64(0x13198A2E03707344)
-#define CB2   SPH_C64(0xA4093822299F31D0)
-#define CB3   SPH_C64(0x082EFA98EC4E6C89)
-#define CB4   SPH_C64(0x452821E638D01377)
-#define CB5   SPH_C64(0xBE5466CF34E90C6C)
-#define CB6   SPH_C64(0xC0AC29B7C97C50DD)
-#define CB7   SPH_C64(0x3F84D5B5B5470917)
-#define CB8   SPH_C64(0x9216D5D98979FB1B)
-#define CB9   SPH_C64(0xD1310BA698DFB5AC)
-#define CBA   SPH_C64(0x2FFD72DBD01ADFB7)
-#define CBB   SPH_C64(0xB8E1AFED6A267E96)
-#define CBC   SPH_C64(0xBA7C9045F12C7F99)
-#define CBD   SPH_C64(0x24A19947B3916CF7)
-#define CBE   SPH_C64(0x0801F2E2858EFC16)
-#define CBF   SPH_C64(0x636920D871574E69)
-
-
-#define GS(m0, m1, c0, c1, a, b, c, d)   do { \
-		a = SPH_T32(a + b + (m0 ^ c1)); \
-		d = SPH_ROTR32(d ^ a, 16); \
-		c = SPH_T32(c + d); \
-		b = SPH_ROTR32(b ^ c, 12); \
-		a = SPH_T32(a + b + (m1 ^ c0)); \
-		d = SPH_ROTR32(d ^ a, 8); \
-		c = SPH_T32(c + d); \
-		b = SPH_ROTR32(b ^ c, 7); \
-	} while (0)
-
-#define ROUND_S(r)   do { \
-		GS(Mx(r, 0), Mx(r, 1), CSx(r, 0), CSx(r, 1), V0, V4, V8, VC); \
-		GS(Mx(r, 2), Mx(r, 3), CSx(r, 2), CSx(r, 3), V1, V5, V9, VD); \
-		GS(Mx(r, 4), Mx(r, 5), CSx(r, 4), CSx(r, 5), V2, V6, VA, VE); \
-		GS(Mx(r, 6), Mx(r, 7), CSx(r, 6), CSx(r, 7), V3, V7, VB, VF); \
-		GS(Mx(r, 8), Mx(r, 9), CSx(r, 8), CSx(r, 9), V0, V5, VA, VF); \
-		GS(Mx(r, A), Mx(r, B), CSx(r, A), CSx(r, B), V1, V6, VB, VC); \
-		GS(Mx(r, C), Mx(r, D), CSx(r, C), CSx(r, D), V2, V7, V8, VD); \
-		GS(Mx(r, E), Mx(r, F), CSx(r, E), CSx(r, F), V3, V4, V9, VE); \
-	} while (0)
-
-
-
-#define GB(m0, m1, c0, c1, a, b, c, d)   do { \
-		a = SPH_T64(a + b + (m0 ^ c1)); \
-		d = SPH_ROTR64(d ^ a, 32); \
-		c = SPH_T64(c + d); \
-		b = SPH_ROTR64(b ^ c, 25); \
-		a = SPH_T64(a + b + (m1 ^ c0)); \
-		d = SPH_ROTR64(d ^ a, 16); \
-		c = SPH_T64(c + d); \
-		b = SPH_ROTR64(b ^ c, 11); \
-	} while (0)
-
-#define ROUND_B(r)   do { \
-		GB(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
-		GB(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
-		GB(Mx(r, 4), Mx(r, 5), CBx(r, 4), CBx(r, 5), V2, V6, VA, VE); \
-		GB(Mx(r, 6), Mx(r, 7), CBx(r, 6), CBx(r, 7), V3, V7, VB, VF); \
-		GB(Mx(r, 8), Mx(r, 9), CBx(r, 8), CBx(r, 9), V0, V5, VA, VF); \
-		GB(Mx(r, A), Mx(r, B), CBx(r, A), CBx(r, B), V1, V6, VB, VC); \
-		GB(Mx(r, C), Mx(r, D), CBx(r, C), CBx(r, D), V2, V7, V8, VD); \
-		GB(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
-	} while (0)
-
-
-#define COMPRESS64   do { \
-                int b=0; \
-		sph_u64 M0, M1, M2, M3, M4, M5, M6, M7; \
-		sph_u64 M8, M9, MA, MB, MC, MD, ME, MF; \
-		sph_u64 V0, V1, V2, V3, V4, V5, V6, V7; \
-		sph_u64 V8, V9, VA, VB, VC, VD, VE, VF; \
-		V0 = blkH0, \
-		V1 = blkH1, \
-		V2 = blkH2, \
-		V3 = blkH3, \
-		V4 = blkH4, \
-		V5 = blkH5, \
-		V6 = blkH6, \
-		V7 = blkH7; \
-		V8 = blkS0 ^ CB0, \
-		V9 = blkS1 ^ CB1, \
-		VA = blkS2 ^ CB2, \
-		VB = blkS3 ^ CB3, \
-		VC = hashctA ^ CB4, \
-		VD = hashctA ^ CB5, \
-		VE = hashctB ^ CB6, \
-		VF = hashctB ^ CB7; \
-		M0 = sph_dec64be_aligned(buf +   0), \
-		M1 = sph_dec64be_aligned(buf +   8), \
-		M2 = sph_dec64be_aligned(buf +  16), \
-		M3 = sph_dec64be_aligned(buf +  24), \
-		M4 = sph_dec64be_aligned(buf +  32), \
-		M5 = sph_dec64be_aligned(buf +  40), \
-		M6 = sph_dec64be_aligned(buf +  48), \
-		M7 = sph_dec64be_aligned(buf +  56), \
-		M8 = sph_dec64be_aligned(buf +  64), \
-		M9 = sph_dec64be_aligned(buf +  72), \
-		MA = sph_dec64be_aligned(buf +  80), \
-		MB = sph_dec64be_aligned(buf +  88), \
-		MC = sph_dec64be_aligned(buf +  96), \
-		MD = sph_dec64be_aligned(buf + 104), \
-		ME = sph_dec64be_aligned(buf + 112), \
-		MF = sph_dec64be_aligned(buf + 120); \
-                /* loop once and a half */ \
-                /* save some space */ \
-                for (;;) { \
-		    ROUND_B(0); \
-		    ROUND_B(1); \
-		    ROUND_B(2); \
-		    ROUND_B(3); \
-		    ROUND_B(4); \
-		    ROUND_B(5); \
-                    if (b)  break; \
-                    b = 1; \
-		    ROUND_B(6); \
-		    ROUND_B(7); \
-		    ROUND_B(8); \
-		    ROUND_B(9); \
-                }; \
-		blkH0 ^= blkS0 ^ V0 ^ V8, \
-		blkH1 ^= blkS1 ^ V1 ^ V9, \
-		blkH2 ^= blkS2 ^ V2 ^ VA, \
-		blkH3 ^= blkS3 ^ V3 ^ VB, \
-		blkH4 ^= blkS0 ^ V4 ^ VC, \
-		blkH5 ^= blkS1 ^ V5 ^ VD, \
-		blkH6 ^= blkS2 ^ V6 ^ VE, \
-		blkH7 ^= blkS3 ^ V7 ^ VF; \
-	} while (0)
-/*
-*/
-#define DECL_BLK \
-	sph_u64 blkH0; \
-	sph_u64 blkH1; \
-	sph_u64 blkH2; \
-	sph_u64 blkH3; \
-	sph_u64 blkH4; \
-	sph_u64 blkH5; \
-	sph_u64 blkH6; \
-	sph_u64 blkH7; \
-	sph_u64 blkS0; \
-	sph_u64 blkS1; \
-	sph_u64 blkS2; \
-	sph_u64 blkS3; \
-
-/* load initial constants */
-#define BLK_I \
-do { \
-    blkH0 = SPH_C64(0x6A09E667F3BCC908); \
-    blkH1 = SPH_C64(0xBB67AE8584CAA73B); \
-    blkH2 = SPH_C64(0x3C6EF372FE94F82B); \
-    blkH3 = SPH_C64(0xA54FF53A5F1D36F1); \
-    blkH4 = SPH_C64(0x510E527FADE682D1); \
-    blkH5 = SPH_C64(0x9B05688C2B3E6C1F); \
-    blkH6 = SPH_C64(0x1F83D9ABFB41BD6B); \
-    blkH7 = SPH_C64(0x5BE0CD19137E2179); \
-    blkS0 = 0; \
-    blkS1 = 0; \
-    blkS2 = 0; \
-    blkS3 = 0; \
-    hashctB = SPH_T64(0- 1); \
-} while (0)
-
-/* copy in 80 for initial hash */
-#define BLK_W \
-do { \
-    memcpy(hashbuf, input, 80); \
-    hashctA = SPH_C64(0xFFFFFFFFFFFFFC00) + 80*8; \
-    hashptr = 80; \
-} while (0)
-
-/* copy in 64 for looped hash */
-#define BLK_U \
-do { \
-    memcpy(hashbuf, hash , 64); \
-    hashctA = SPH_C64(0xFFFFFFFFFFFFFC00) + 64*8; \
-    hashptr = 64; \
-} while (0)
-
-/* blake compress function */
-/* hash = blake512(loaded) */
-#define BLK_C \
-do { \
-    \
-    union { \
-        unsigned char buf[128]; \
-        sph_u64 dummy; \
-    } u; \
-    size_t ptr; \
-    unsigned bit_len; \
- \
-    ptr = hashptr; \
-    bit_len = ((unsigned)ptr << 3) + 0; \
-    u.buf[ptr] = ((0 & -(0x80)) | (0x80)) & 0xFF; \
-    memset(u.buf + ptr + 1, 0, 111 - ptr); \
-    u.buf[111] |= 1; \
-    sph_enc64be_aligned(u.buf + 112, 0); \
-    sph_enc64be_aligned(u.buf + 120, bit_len); \
-    do { \
-    const void *data = u.buf + ptr; \
-    unsigned char *buf; \
-    buf = hashbuf; \
-    size_t clen; \
-    clen = (sizeof(char)*128) - hashptr; \
-    memcpy(buf + hashptr, data, clen); \
-    hashctA = SPH_T64(hashctA + 1024); \
-    hashctB = SPH_T64(hashctB + 1); \
-    COMPRESS64; \
-    } while (0); \
-    /* end blake64(sc, u.buf + ptr, 128 - ptr); */ \
-    sph_enc64be((unsigned char*)(hash) + (0 << 3), blkH0), \
-    sph_enc64be((unsigned char*)(hash) + (1 << 3), blkH1); \
-    sph_enc64be((unsigned char*)(hash) + (2 << 3), blkH2), \
-    sph_enc64be((unsigned char*)(hash) + (3 << 3), blkH3); \
-    sph_enc64be((unsigned char*)(hash) + (4 << 3), blkH4), \
-    sph_enc64be((unsigned char*)(hash) + (5 << 3), blkH5); \
-    sph_enc64be((unsigned char*)(hash) + (6 << 3), blkH6), \
-    sph_enc64be((unsigned char*)(hash) + (7 << 3), blkH7); \
-} while (0) 
-
-
-#ifdef __cplusplus
-}
-#endif
--- a/algo/blake/sse2/blake/sse41/api.h
+++ b/algo/blake/sse2/blake/sse41/api.h
@@ -1,2 +0,0 @@
-#define CRYPTO_BYTES 64
-
--- a/algo/blake/sse2/blake/sse41/architectures
+++ b/algo/blake/sse2/blake/sse41/architectures
@@ -1,2 +0,0 @@
-amd64
-x86
--- a/algo/blake/sse2/blake/sse41/config.h
+++ b/algo/blake/sse2/blake/sse41/config.h
@@ -1,8 +0,0 @@
-#ifndef __BLAKE512_CONFIG_H__
-#define __BLAKE512_CONFIG_H__
-
-#define AVOID_BRANCHING 1
-//#define HAVE_XOP 1
-
-#endif
-
--- a/algo/blake/sse2/blake/sse41/hash.c
+++ b/algo/blake/sse2/blake/sse41/hash.c
@@ -1,287 +0,0 @@
-
-#include "hash.h"
-/*
-#ifndef NOT_SUPERCOP
-
-#include "crypto_hash.h"
-#include "crypto_uint64.h"
-#include "crypto_uint32.h"
-#include "crypto_uint8.h"
-
-typedef crypto_uint64 u64;
-typedef crypto_uint32 u32;
-typedef crypto_uint8 u8; 
-
-#else
-
-typedef unsigned long long u64; 
-typedef unsigned int u32; 
-typedef unsigned char u8; 
-
-#endif
-*/
-#define U8TO32(p) \
-  (((u32)((p)[0]) << 24) | ((u32)((p)[1]) << 16) | \
-   ((u32)((p)[2]) <<  8) | ((u32)((p)[3])      ))
-#define U8TO64(p) \
-  (((u64)U8TO32(p) << 32) | (u64)U8TO32((p) + 4))
-#define U32TO8(p, v) \
-    (p)[0] = (u8)((v) >> 24); (p)[1] = (u8)((v) >> 16); \
-    (p)[2] = (u8)((v) >>  8); (p)[3] = (u8)((v)      ); 
-#define U64TO8(p, v) \
-    U32TO8((p),     (u32)((v) >> 32));	\
-    U32TO8((p) + 4, (u32)((v)      )); 
-/*
-typedef struct  
-{ 
-	__m128i h[4];
-  u64 s[4], t[2];
-  u32 buflen, nullt;
-  u8 buf[128];
-} state __attribute__ ((aligned (64)));
-*/
-static const u8 padding[129] =
-{ 
-	0x80,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
-    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
-    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
-    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
-};
-
-static inline int blake512_compress( hashState_blake * state, const u8 * datablock ) 
-{
-
-  __m128i row1l,row1h;
-  __m128i row2l,row2h;
-  __m128i row3l,row3h;
-  __m128i row4l,row4h;
-
-  const __m128i r16 = _mm_setr_epi8(2,3,4,5,6,7,0,1,10,11,12,13,14,15,8,9);
-  const __m128i u8to64 = _mm_set_epi8(8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);
-
-  __m128i m0, m1, m2, m3, m4, m5, m6, m7;
-  __m128i t0, t1, t2, t3, t4, t5, t6, t7;
-  __m128i b0, b1, b2, b3;
-
-  m0 = _mm_loadu_si128((__m128i*)(datablock +   0));
-  m1 = _mm_loadu_si128((__m128i*)(datablock +  16));
-  m2 = _mm_loadu_si128((__m128i*)(datablock +  32));
-  m3 = _mm_loadu_si128((__m128i*)(datablock +  48));
-  m4 = _mm_loadu_si128((__m128i*)(datablock +  64));
-  m5 = _mm_loadu_si128((__m128i*)(datablock +  80));
-  m6 = _mm_loadu_si128((__m128i*)(datablock +  96));
-  m7 = _mm_loadu_si128((__m128i*)(datablock + 112));
-
-  m0 = BSWAP64(m0);
-  m1 = BSWAP64(m1);
-  m2 = BSWAP64(m2);
-  m3 = BSWAP64(m3);
-  m4 = BSWAP64(m4);
-  m5 = BSWAP64(m5);
-  m6 = BSWAP64(m6);
-  m7 = BSWAP64(m7);
-
-  row1l = state->h[0];
-  row1h = state->h[1];
-  row2l = state->h[2];
-  row2h = state->h[3];
-  row3l = _mm_set_epi64x(0x13198A2E03707344ULL, 0x243F6A8885A308D3ULL);
-  row3h = _mm_set_epi64x(0x082EFA98EC4E6C89ULL, 0xA4093822299F31D0ULL);
-
-  row4l = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0x452821E638D01377ULL);
-  row4h = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0xC0AC29B7C97C50DDULL);
-
-#ifdef AVOID_BRANCHING
-  do
-  {
-    const __m128i mask = _mm_cmpeq_epi32(_mm_setzero_si128(), _mm_set1_epi32(state->nullt));
-    const __m128i xor1 = _mm_and_si128(_mm_set1_epi64x(state->t[0]), mask);
-    const __m128i xor2 = _mm_and_si128(_mm_set1_epi64x(state->t[1]), mask);
-    row4l = _mm_xor_si128(row4l, xor1);
-    row4h = _mm_xor_si128(row4h, xor2);
-  } while(0);
-#else
-  if(!state->nullt)
-  {
-  	row4l = _mm_xor_si128(row4l, _mm_set1_epi64x(state->t[0]));
-  	row4h = _mm_xor_si128(row4h, _mm_set1_epi64x(state->t[1]));
-  }
-#endif
-
-  ROUND( 0);
-  ROUND( 1);
-  ROUND( 2);
-  ROUND( 3);
-  ROUND( 4);
-  ROUND( 5);
-  ROUND( 6);
-  ROUND( 7);
-  ROUND( 8);
-  ROUND( 9);
-  ROUND(10);
-  ROUND(11);
-  ROUND(12);
-  ROUND(13);
-  ROUND(14);
-  ROUND(15);
-
-  row1l = _mm_xor_si128(row3l,row1l);
-  row1h = _mm_xor_si128(row3h,row1h);
-
-  state->h[0] = _mm_xor_si128(row1l, state->h[0]);
-  state->h[1] = _mm_xor_si128(row1h, state->h[1]);
-
-  row2l = _mm_xor_si128(row4l,row2l);
-  row2h = _mm_xor_si128(row4h,row2h);
-
-  state->h[2] = _mm_xor_si128(row2l, state->h[2]);
-  state->h[3] = _mm_xor_si128(row2h, state->h[3]);
-  
-  return 0;
-}
-
-static inline void blake512_init( hashState_blake * S, u64 databitlen )
-{
-  memset(S, 0, sizeof(hashState_blake));
-  S->h[0] = _mm_set_epi64x(0xBB67AE8584CAA73BULL, 0x6A09E667F3BCC908ULL);
-  S->h[1] = _mm_set_epi64x(0xA54FF53A5F1D36F1ULL, 0x3C6EF372FE94F82BULL);
-  S->h[2] = _mm_set_epi64x(0x9B05688C2B3E6C1FULL, 0x510E527FADE682D1ULL);
-  S->h[3] = _mm_set_epi64x(0x5BE0CD19137E2179ULL, 0x1F83D9ABFB41BD6BULL);
-  S->buflen = databitlen;
-}
-
-
-static void blake512_update( hashState_blake * S, const u8 * data, u64 datalen ) 
-{
-
-
-  int left = (S->buflen >> 3); 
-  int fill = 128 - left;
-
-  if( left && ( ((datalen >> 3) & 0x7F) >= fill ) ) {
-    memcpy( (void *) (S->buf + left), (void *) data, fill );
-    S->t[0] += 1024;
-    blake512_compress( S, S->buf );
-    data += fill;
-    datalen  -= (fill << 3);       
-    left = 0;
-  }
-
-  while( datalen >= 1024 ) {  
-    S->t[0] += 1024;
-    blake512_compress( S, data );
-    data += 128;
-    datalen  -= 1024;
-  }
-
-  if( datalen > 0 ) {
-    memcpy( (void *) (S->buf + left), (void *) data, ( datalen>>3 ) & 0x7F );
-    S->buflen = (left<<3) + datalen;
-  }
-  else S->buflen=0;
-}
-
-static inline void blake512_final( hashState_blake * S, u8 * digest ) 
-{
-
-  u8 msglen[16], zo=0x01,oo=0x81;
-  u64 lo=S->t[0] + S->buflen, hi = S->t[1];
-  if ( lo < S->buflen ) hi++;
-  U64TO8(  msglen + 0, hi );
-  U64TO8(  msglen + 8, lo );
-
-  if ( S->buflen == 888 ) /* one padding byte */
-  { 
-    S->t[0] -= 8; 
-    blake512_update( S, &oo, 8 );
-  }
-  else 
-  {
-    if ( S->buflen < 888 ) /* enough space to fill the block */
-    { 
-      if ( S->buflen == 0 ) S->nullt=1;
-      S->t[0] -= 888 - S->buflen;
-      blake512_update( S, padding, 888 - S->buflen );
-    }
-    else /* NOT enough space, need 2 compressions */ 
-    { 
-      S->t[0] -= 1024 - S->buflen; 
-      blake512_update( S, padding, 1024 - S->buflen );
-      S->t[0] -= 888;
-      blake512_update( S, padding+1, 888 );
-      S->nullt = 1;
-    }
-    blake512_update( S, &zo, 8 );
-    S->t[0] -= 8;
-  }
-  S->t[0] -= 128;
-  blake512_update( S, msglen, 128 );    
-
-  do
-  {
-    const __m128i u8to64 = _mm_set_epi8(8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7);
-    _mm_storeu_si128((__m128i*)(digest +  0), BSWAP64(S->h[0]));
-    _mm_storeu_si128((__m128i*)(digest + 16), BSWAP64(S->h[1]));
-    _mm_storeu_si128((__m128i*)(digest + 32), BSWAP64(S->h[2]));
-    _mm_storeu_si128((__m128i*)(digest + 48), BSWAP64(S->h[3]));
-  } while(0);
-}
-
-/*
-int crypto_hash( unsigned char *out, const unsigned char *in, unsigned long long inlen ) 
-{
-
-  hashState_blake S;
-  blake512_init( &S );
-  blake512_update( &S, in, inlen*8 );
-  blake512_final( &S, out );
-  return 0;
-}
-*/
-/*
-#ifdef NOT_SUPERCOP
-
-int main() 
-{
-
-  int i, v;
-  u8 data[144], digest[64];
-  u8 test1[]= {0x97, 0x96, 0x15, 0x87, 0xF6, 0xD9, 0x70, 0xFA, 0xBA, 0x6D, 0x24, 0x78, 0x04, 0x5D, 0xE6, 0xD1, 
-	       0xFA, 0xBD, 0x09, 0xB6, 0x1A, 0xE5, 0x09, 0x32, 0x05, 0x4D, 0x52, 0xBC, 0x29, 0xD3, 0x1B, 0xE4, 
-	       0xFF, 0x91, 0x02, 0xB9, 0xF6, 0x9E, 0x2B, 0xBD, 0xB8, 0x3B, 0xE1, 0x3D, 0x4B, 0x9C, 0x06, 0x09, 
-	       0x1E, 0x5F, 0xA0, 0xB4, 0x8B, 0xD0, 0x81, 0xB6, 0x34, 0x05, 0x8B, 0xE0, 0xEC, 0x49, 0xBE, 0xB3};
-  u8 test2[]= {0x31, 0x37, 0x17, 0xD6, 0x08, 0xE9, 0xCF, 0x75, 0x8D, 0xCB, 0x1E, 0xB0, 0xF0, 0xC3, 0xCF, 0x9F, 
-	       0xC1, 0x50, 0xB2, 0xD5, 0x00, 0xFB, 0x33, 0xF5, 0x1C, 0x52, 0xAF, 0xC9, 0x9D, 0x35, 0x8A, 0x2F, 
-	       0x13, 0x74, 0xB8, 0xA3, 0x8B, 0xBA, 0x79, 0x74, 0xE7, 0xF6, 0xEF, 0x79, 0xCA, 0xB1, 0x6F, 0x22, 
-	       0xCE, 0x1E, 0x64, 0x9D, 0x6E, 0x01, 0xAD, 0x95, 0x89, 0xC2, 0x13, 0x04, 0x5D, 0x54, 0x5D, 0xDE};
-
-  for(i=0; i<144; ++i) data[i]=0;  
-
-  crypto_hash( digest, data, 1 );    
-  v=0;
-  for(i=0; i<64; ++i) {
-    printf("%02X", digest[i]);
-    if ( digest[i] != test1[i]) v=1;
-  }
-  if (v) printf("\nerror\n");
-  else  printf("\nok\n");
-
-  for(i=0; i<144; ++i) data[i]=0;  
-
-  crypto_hash( digest, data, 144 );    
-  v=0;
-  for(i=0; i<64; ++i) {
-    printf("%02X", digest[i]);
-    if ( digest[i] != test2[i]) v=1;
-  }
-  if (v) printf("\nerror\n");
-  else printf("\nok\n");
-
-  return 0;
-}
-
-#endif
-
-*/
-
-
--- a/algo/blake/sse2/blake/sse41/hash.h
+++ b/algo/blake/sse2/blake/sse41/hash.h
@@ -1,74 +0,0 @@
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <x86intrin.h>
-
-#include "config.h"
-#include "rounds.h"
-/*
-#ifndef NOT_SUPERCOP
-
-#include "crypto_hash.h"
-#include "crypto_uint64.h"
-#include "crypto_uint32.h"
-#include "crypto_uint8.h"
-
-typedef crypto_uint64 u64;
-typedef crypto_uint32 u32;
-typedef crypto_uint8 u8; 
-
-#else
-*/
-typedef unsigned long long u64; 
-typedef unsigned int u32; 
-typedef unsigned char u8; 
-
-typedef struct  
-{ 
-	__m128i h[4];
-  u64 s[4], t[2];
-  u32 buflen, nullt;
-  u8 buf[128];
-} hashState_blake __attribute__ ((aligned (64)));
-/*
-#endif
-
-#define U8TO32(p) \
-  (((u32)((p)[0]) << 24) | ((u32)((p)[1]) << 16) | \
-   ((u32)((p)[2]) <<  8) | ((u32)((p)[3])      ))
-#define U8TO64(p) \
-  (((u64)U8TO32(p) << 32) | (u64)U8TO32((p) + 4))
-#define U32TO8(p, v) \
-    (p)[0] = (u8)((v) >> 24); (p)[1] = (u8)((v) >> 16); \
-    (p)[2] = (u8)((v) >>  8); (p)[3] = (u8)((v)      ); 
-#define U64TO8(p, v) \
-    U32TO8((p),     (u32)((v) >> 32));	\
-    U32TO8((p) + 4, (u32)((v)      )); 
-*/
-
-/*
-static const u8 padding[129] =
-{ 
-	0x80,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
-    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
-    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
-    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
-};
-
-*/
-static inline void blake512_init( hashState_blake * S, u64 datalen );
-
-
-static void blake512_update( hashState_blake * S, const u8 * data, u64 datalen ) ;
-
-static inline void blake512_final( hashState_blake * S, u8 * digest ) ;
-
-
-int crypto_hash( unsigned char *out, const unsigned char *in, unsigned long long inlen ) ;
-
-
-
-
-
-
--- a/algo/blake/sse2/blake/sse41/implementors
+++ b/algo/blake/sse2/blake/sse41/implementors
@@ -1,2 +0,0 @@
-Jean-Philippe Aumasson
-Samuel Neves
--- a/algo/blake/sse2/blake/sse41/rounds.h
+++ b/algo/blake/sse2/blake/sse41/rounds.h
@@ -1,871 +0,0 @@
-
-#ifndef __BLAKE512_ROUNDS_H__
-#define __BLAKE512_ROUNDS_H__
-
-#ifndef HAVE_XOP
-	#define BSWAP64(x) _mm_shuffle_epi8((x), u8to64)
-
-	#define _mm_roti_epi64(x, c) \
-	(-(c) == 32) ? _mm_shuffle_epi32((x), _MM_SHUFFLE(2,3,0,1))  \
-	: (-(c) == 16) ? _mm_shuffle_epi8((x), r16) \
-		: _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_slli_epi64((x), 64-(-c))) 
-#else
-	#define BSWAP64(x) _mm_perm_epi8((x),(x),u8to64)
-#endif
-
-
-#define LOAD_MSG_0_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m0, m1); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x13198A2E03707344ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m2, m3); \
-t3 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0xBE5466CF34E90C6CULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_0_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m0, m1); \
-t1 = _mm_set_epi64x(0xA4093822299F31D0ULL, 0x243F6A8885A308D3ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m2, m3); \
-t3 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_0_3(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m4, m5); \
-t1 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0xD1310BA698DFB5ACULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m6, m7); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0x24A19947B3916CF7ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_0_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m4, m5); \
-t1 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0x9216D5D98979FB1BULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m6, m7); \
-t3 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xBA7C9045F12C7F99ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_1_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m7, m2); \
-t1 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x2FFD72DBD01ADFB7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m4, m6); \
-t3 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x636920D871574E69ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_1_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m5, m4); \
-t1 = _mm_set_epi64x(0x452821E638D01377ULL, 0x801F2E2858EFC16ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m3, m7, 8); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0xD1310BA698DFB5ACULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_1_3(b0, b1) \
-do \
-{ \
-t0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
-t1 = _mm_set_epi64x(0xA4093822299F31D0ULL, 0xBA7C9045F12C7F99ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m5, m2); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x3F84D5B5B5470917ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_1_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m6, m1); \
-t1 = _mm_set_epi64x(0x243F6A8885A308D3ULL, 0x13198A2E03707344ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m3, m1); \
-t3 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xB8E1AFED6A267E96ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_2_1(b0, b1) \
-do \
-{ \
-t0 = _mm_alignr_epi8(m6, m5, 8); \
-t1 = _mm_set_epi64x(0x243F6A8885A308D3ULL, 0x9216D5D98979FB1BULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m2, m7); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0xA4093822299F31D0ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_2_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m4, m0); \
-t1 = _mm_set_epi64x(0xBA7C9045F12C7F99ULL, 0xB8E1AFED6A267E96ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m1, m6, 0xF0); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0xBE5466CF34E90C6CULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_2_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m5, m1, 0xF0); \
-t1 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x801F2E2858EFC16ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m3, m4); \
-t3 = _mm_set_epi64x(0x452821E638D01377ULL, 0x13198A2E03707344ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_2_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m7, m3); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x2FFD72DBD01ADFB7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m2, m0, 8); \
-t3 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x3F84D5B5B5470917ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_3_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m3, m1); \
-t1 = _mm_set_epi64x(0x13198A2E03707344ULL, 0xD1310BA698DFB5ACULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m6, m5); \
-t3 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xBA7C9045F12C7F99ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_3_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m4, m0); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x3F84D5B5B5470917ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m6, m7); \
-t3 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0x24A19947B3916CF7ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_3_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m1, m2, 0xF0); \
-t1 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xC0AC29B7C97C50DDULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m2, m7, 0xF0); \
-t3 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x243F6A8885A308D3ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_3_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m3, m5); \
-t1 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xA4093822299F31D0ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m0, m4); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_4_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m4, m2); \
-t1 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0x243F6A8885A308D3ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m1, m5); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_4_2(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m0, m3, 0xF0); \
-t1 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xD1310BA698DFB5ACULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m2, m7, 0xF0); \
-t3 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xA4093822299F31D0ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_4_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m7, m5, 0xF0); \
-t1 = _mm_set_epi64x(0xBA7C9045F12C7F99ULL, 0x13198A2E03707344ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m3, m1, 0xF0); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0x9216D5D98979FB1BULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_4_4(b0, b1) \
-do \
-{ \
-t0 = _mm_alignr_epi8(m6, m0, 8); \
-t1 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0x801F2E2858EFC16ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m4, m6, 0xF0); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0xC0AC29B7C97C50DDULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_5_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m1, m3); \
-t1 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xBA7C9045F12C7F99ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m0, m4); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0xB8E1AFED6A267E96ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_5_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m6, m5); \
-t1 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0xA4093822299F31D0ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m5, m1); \
-t3 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x243F6A8885A308D3ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_5_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m2, m3, 0xF0); \
-t1 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0x24A19947B3916CF7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m7, m0); \
-t3 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x801F2E2858EFC16ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_5_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m6, m2); \
-t1 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0x452821E638D01377ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m7, m4, 0xF0); \
-t3 = _mm_set_epi64x(0x13198A2E03707344ULL, 0x636920D871574E69ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_6_1(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m6, m0, 0xF0); \
-t1 = _mm_set_epi64x(0x636920D871574E69ULL, 0xBE5466CF34E90C6CULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m7, m2); \
-t3 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0x24A19947B3916CF7ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_6_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m2, m7); \
-t1 = _mm_set_epi64x(0x13198A2E03707344ULL, 0xBA7C9045F12C7F99ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m5, m6, 8); \
-t3 = _mm_set_epi64x(0x452821E638D01377ULL, 0x801F2E2858EFC16ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_6_3(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m0, m3); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x3F84D5B5B5470917ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_shuffle_epi32(m4, _MM_SHUFFLE(1,0,3,2)); \
-t3 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0xA4093822299F31D0ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_6_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m3, m1); \
-t1 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x243F6A8885A308D3ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m1, m5, 0xF0); \
-t3 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0xD1310BA698DFB5ACULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_7_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m6, m3); \
-t1 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xB8E1AFED6A267E96ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m6, m1, 0xF0); \
-t3 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x13198A2E03707344ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_7_2(b0, b1) \
-do \
-{ \
-t0 = _mm_alignr_epi8(m7, m5, 8); \
-t1 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0x24A19947B3916CF7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m0, m4); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0xBA7C9045F12C7F99ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_7_3(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m2, m7); \
-t1 = _mm_set_epi64x(0x452821E638D01377ULL, 0x243F6A8885A308D3ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m4, m1); \
-t3 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xC0AC29B7C97C50DDULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_7_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m0, m2); \
-t1 = _mm_set_epi64x(0x636920D871574E69ULL, 0xBE5466CF34E90C6CULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m3, m5); \
-t3 = _mm_set_epi64x(0xA4093822299F31D0ULL, 0x9216D5D98979FB1BULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_8_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m3, m7); \
-t1 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x636920D871574E69ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m0, m5, 8); \
-t3 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x82EFA98EC4E6C89ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_8_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m7, m4); \
-t1 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xC0AC29B7C97C50DDULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m4, m1, 8); \
-t3 = _mm_set_epi64x(0x243F6A8885A308D3ULL, 0xB8E1AFED6A267E96ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_8_3(b0, b1) \
-do \
-{ \
-t0 = m6; \
-t1 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0xA4093822299F31D0ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m5, m0, 8); \
-t3 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_8_4(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m1, m3, 0xF0); \
-t1 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0xBA7C9045F12C7F99ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = m2; \
-t3 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0x13198A2E03707344ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_9_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m5, m4); \
-t1 = _mm_set_epi64x(0x452821E638D01377ULL, 0xA4093822299F31D0ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m3, m0); \
-t3 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xC0AC29B7C97C50DDULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_9_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m1, m2); \
-t1 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x2FFD72DBD01ADFB7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m3, m2, 0xF0); \
-t3 = _mm_set_epi64x(0x13198A2E03707344ULL, 0x3F84D5B5B5470917ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_9_3(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m7, m4); \
-t1 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xB8E1AFED6A267E96ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m1, m6); \
-t3 = _mm_set_epi64x(0x243F6A8885A308D3ULL, 0xBA7C9045F12C7F99ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_9_4(b0, b1) \
-do \
-{ \
-t0 = _mm_alignr_epi8(m7, m5, 8); \
-t1 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x636920D871574E69ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m6, m0); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0x82EFA98EC4E6C89ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_10_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m0, m1); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x13198A2E03707344ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m2, m3); \
-t3 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0xBE5466CF34E90C6CULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_10_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m0, m1); \
-t1 = _mm_set_epi64x(0xA4093822299F31D0ULL, 0x243F6A8885A308D3ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m2, m3); \
-t3 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_10_3(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m4, m5); \
-t1 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0xD1310BA698DFB5ACULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m6, m7); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0x24A19947B3916CF7ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_10_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m4, m5); \
-t1 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0x9216D5D98979FB1BULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m6, m7); \
-t3 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xBA7C9045F12C7F99ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_11_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m7, m2); \
-t1 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x2FFD72DBD01ADFB7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m4, m6); \
-t3 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x636920D871574E69ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_11_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m5, m4); \
-t1 = _mm_set_epi64x(0x452821E638D01377ULL, 0x801F2E2858EFC16ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m3, m7, 8); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0xD1310BA698DFB5ACULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_11_3(b0, b1) \
-do \
-{ \
-t0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
-t1 = _mm_set_epi64x(0xA4093822299F31D0ULL, 0xBA7C9045F12C7F99ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m5, m2); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x3F84D5B5B5470917ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_11_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m6, m1); \
-t1 = _mm_set_epi64x(0x243F6A8885A308D3ULL, 0x13198A2E03707344ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m3, m1); \
-t3 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xB8E1AFED6A267E96ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_12_1(b0, b1) \
-do \
-{ \
-t0 = _mm_alignr_epi8(m6, m5, 8); \
-t1 = _mm_set_epi64x(0x243F6A8885A308D3ULL, 0x9216D5D98979FB1BULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m2, m7); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0xA4093822299F31D0ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_12_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m4, m0); \
-t1 = _mm_set_epi64x(0xBA7C9045F12C7F99ULL, 0xB8E1AFED6A267E96ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m1, m6, 0xF0); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0xBE5466CF34E90C6CULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_12_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m5, m1, 0xF0); \
-t1 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0x801F2E2858EFC16ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m3, m4); \
-t3 = _mm_set_epi64x(0x452821E638D01377ULL, 0x13198A2E03707344ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_12_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m7, m3); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x2FFD72DBD01ADFB7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_alignr_epi8(m2, m0, 8); \
-t3 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x3F84D5B5B5470917ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_13_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m3, m1); \
-t1 = _mm_set_epi64x(0x13198A2E03707344ULL, 0xD1310BA698DFB5ACULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m6, m5); \
-t3 = _mm_set_epi64x(0x801F2E2858EFC16ULL, 0xBA7C9045F12C7F99ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_13_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m4, m0); \
-t1 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0x3F84D5B5B5470917ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m6, m7); \
-t3 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0x24A19947B3916CF7ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_13_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m1, m2, 0xF0); \
-t1 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xC0AC29B7C97C50DDULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m2, m7, 0xF0); \
-t3 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x243F6A8885A308D3ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_13_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m3, m5); \
-t1 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xA4093822299F31D0ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m0, m4); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_14_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m4, m2); \
-t1 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0x243F6A8885A308D3ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m1, m5); \
-t3 = _mm_set_epi64x(0x636920D871574E69ULL, 0x452821E638D01377ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_14_2(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m0, m3, 0xF0); \
-t1 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0xD1310BA698DFB5ACULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m2, m7, 0xF0); \
-t3 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xA4093822299F31D0ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_14_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m7, m5, 0xF0); \
-t1 = _mm_set_epi64x(0xBA7C9045F12C7F99ULL, 0x13198A2E03707344ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m3, m1, 0xF0); \
-t3 = _mm_set_epi64x(0x24A19947B3916CF7ULL, 0x9216D5D98979FB1BULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_14_4(b0, b1) \
-do \
-{ \
-t0 = _mm_alignr_epi8(m6, m0, 8); \
-t1 = _mm_set_epi64x(0xB8E1AFED6A267E96ULL, 0x801F2E2858EFC16ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m4, m6, 0xF0); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0xC0AC29B7C97C50DDULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_15_1(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m1, m3); \
-t1 = _mm_set_epi64x(0x2FFD72DBD01ADFB7ULL, 0xBA7C9045F12C7F99ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpacklo_epi64(m0, m4); \
-t3 = _mm_set_epi64x(0x82EFA98EC4E6C89ULL, 0xB8E1AFED6A267E96ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_15_2(b0, b1) \
-do \
-{ \
-t0 = _mm_unpacklo_epi64(m6, m5); \
-t1 = _mm_set_epi64x(0xC0AC29B7C97C50DDULL, 0xA4093822299F31D0ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m5, m1); \
-t3 = _mm_set_epi64x(0x9216D5D98979FB1BULL, 0x243F6A8885A308D3ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_15_3(b0, b1) \
-do \
-{ \
-t0 = _mm_blend_epi16(m2, m3, 0xF0); \
-t1 = _mm_set_epi64x(0xBE5466CF34E90C6CULL, 0x24A19947B3916CF7ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_unpackhi_epi64(m7, m0); \
-t3 = _mm_set_epi64x(0xD1310BA698DFB5ACULL, 0x801F2E2858EFC16ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-#define LOAD_MSG_15_4(b0, b1) \
-do \
-{ \
-t0 = _mm_unpackhi_epi64(m6, m2); \
-t1 = _mm_set_epi64x(0x3F84D5B5B5470917ULL, 0x452821E638D01377ULL); \
-b0 = _mm_xor_si128(t0, t1); \
-t2 = _mm_blend_epi16(m7, m4, 0xF0); \
-t3 = _mm_set_epi64x(0x13198A2E03707344ULL, 0x636920D871574E69ULL); \
-b1 = _mm_xor_si128(t2, t3); \
-} while(0) 
-
-
-
-
-
-
-#define G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
-  row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
-  row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
-  \
-  row4l = _mm_xor_si128(row4l, row1l); \
-  row4h = _mm_xor_si128(row4h, row1h); \
-  \
-  row4l = _mm_roti_epi64(row4l, -32); \
-  row4h = _mm_roti_epi64(row4h, -32); \
-  \
-  row3l = _mm_add_epi64(row3l, row4l); \
-  row3h = _mm_add_epi64(row3h, row4h); \
-  \
-  row2l = _mm_xor_si128(row2l, row3l); \
-  row2h = _mm_xor_si128(row2h, row3h); \
-  \
-  row2l = _mm_roti_epi64(row2l, -25); \
-  row2h = _mm_roti_epi64(row2h, -25); \
-
-#define G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
-  row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
-  row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
-  \
-  row4l = _mm_xor_si128(row4l, row1l); \
-  row4h = _mm_xor_si128(row4h, row1h); \
-  \
-  row4l = _mm_roti_epi64(row4l, -16); \
-  row4h = _mm_roti_epi64(row4h, -16); \
-  \
-  row3l = _mm_add_epi64(row3l, row4l); \
-  row3h = _mm_add_epi64(row3h, row4h); \
-  \
-  row2l = _mm_xor_si128(row2l, row3l); \
-  row2h = _mm_xor_si128(row2h, row3h); \
-  \
-  row2l = _mm_roti_epi64(row2l, -11); \
-  row2h = _mm_roti_epi64(row2h, -11); \
-
-
-#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
-	t0 = _mm_alignr_epi8(row2h, row2l, 8); \
-	t1 = _mm_alignr_epi8(row2l, row2h, 8); \
-	row2l = t0; \
-	row2h = t1; \
-	\
-	t0 = row3l; \
-	row3l = row3h; \
-	row3h = t0;    \
-	\
-	t0 = _mm_alignr_epi8(row4h, row4l, 8); \
-	t1 = _mm_alignr_epi8(row4l, row4h, 8); \
-	row4l = t1; \
-	row4h = t0; 
-
-#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
-	t0 = _mm_alignr_epi8(row2l, row2h, 8); \
-	t1 = _mm_alignr_epi8(row2h, row2l, 8); \
-	row2l = t0; \
-	row2h = t1; \
-	\
-	t0 = row3l; \
-	row3l = row3h; \
-	row3h = t0; \
-	\
-	t0 = _mm_alignr_epi8(row4l, row4h, 8); \
-	t1 = _mm_alignr_epi8(row4h, row4l, 8); \
-	row4l = t1; \
-	row4h = t0; 
-
-#define ROUND(r) \
-  LOAD_MSG_ ##r ##_1(b0, b1); \
-  G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
-  LOAD_MSG_ ##r ##_2(b0, b1); \
-  G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
-  DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
-  LOAD_MSG_ ##r ##_3(b0, b1); \
-  G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
-  LOAD_MSG_ ##r ##_4(b0, b1); \
-  G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
-  UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
-
-#endif
-
--- a/algo/bmw/bmw-hash-4way.h
+++ b/algo/bmw/bmw-hash-4way.h
@@ -62,9 +62,10 @@ typedef struct {

 typedef bmw_4way_small_context bmw256_4way_context;

-void bmw256_4way_init(void *cc);
+void bmw256_4way_init( bmw256_4way_context *ctx );

-void bmw256_4way(void *cc, const void *data, size_t len);
+void bmw256_4way_update(void *cc, const void *data, size_t len);
+#define bmw256_4way bmw256_4way_update

 void bmw256_4way_close(void *cc, void *dst);

@@ -78,7 +79,7 @@ void bmw256_4way_addbits_and_close(
 // BMW-256 8 way 32

 typedef struct {
-   __m256i buf[64];
+   __m256i buf[16];
   __m256i H[16];
   size_t ptr;
   uint32_t bit_count;  // assume bit_count fits in 32 bits
@@ -87,11 +88,33 @@ typedef struct {
 typedef bmw_8way_small_context bmw256_8way_context;

 void bmw256_8way_init( bmw256_8way_context *ctx );
-void bmw256_8way( bmw256_8way_context *ctx, const void *data, size_t len );
+void bmw256_8way_update( bmw256_8way_context *ctx, const void *data,
+                         size_t len );
+#define bmw256_8way bmw256_8way_update
 void bmw256_8way_close( bmw256_8way_context *ctx, void *dst );

 #endif

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// BMW-256 16 way 32
+
+typedef struct {
+   __m512i buf[16];
+   __m512i H[16];
+   size_t ptr;
+   uint32_t bit_count;  // assume bit_count fits in 32 bits
+} bmw_16way_small_context __attribute__ ((aligned (128)));
+
+typedef bmw_16way_small_context bmw256_16way_context;
+
+void bmw256_16way_init( bmw256_16way_context *ctx );
+void bmw256_16way_update( bmw256_16way_context *ctx, const void *data,
+                          size_t len );
+void bmw256_16way_close( bmw256_16way_context *ctx, void *dst );
+
+#endif
+

 #if defined(__SSE2__)

@@ -107,7 +130,8 @@ typedef struct {
 typedef bmw_2way_big_context bmw512_2way_context;

 void bmw512_2way_init( bmw512_2way_context *ctx );
-void bmw512_2way( bmw512_2way_context *ctx, const void *data, size_t len );
+void bmw512_2way_update( bmw512_2way_context *ctx, const void *data,
+                         size_t len );
 void bmw512_2way_close( bmw512_2way_context *ctx, void *dst );

 #endif // __SSE2__
@@ -121,14 +145,15 @@ typedef struct {
   __m256i H[16];
   size_t ptr;
   sph_u64 bit_count;
-} bmw_4way_big_context;
+} bmw_4way_big_context __attribute__((aligned(128)));

 typedef bmw_4way_big_context bmw512_4way_context;


 void bmw512_4way_init(void *cc);

-void bmw512_4way(void *cc, const void *data, size_t len);
+void bmw512_4way_update(void *cc, const void *data, size_t len);
+#define bmw512_4way bmw512_4way_update

 void bmw512_4way_close(void *cc, void *dst);

@@ -137,6 +162,22 @@ void bmw512_4way_addbits_and_close(

 #endif  // __AVX2__

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+typedef struct {
+   __m512i buf[16];
+   __m512i H[16];
+   size_t ptr;
+   uint64_t bit_count;
+} bmw512_8way_context __attribute__((aligned(128)));
+
+void bmw512_8way_init( bmw512_8way_context *ctx );
+void bmw512_8way_update( bmw512_8way_context *ctx, const void *data,
+                         size_t len );
+void bmw512_8way_close( bmw512_8way_context *ctx, void *dst );
+
+#endif // AVX512
+
 #ifdef __cplusplus
 }
 #endif
--- a/algo/bmw/bmw256-hash-4way.c
+++ b/algo/bmw/bmw256-hash-4way.c
--- a/algo/bmw/bmw512-4way.c
+++ b/algo/bmw/bmw512-4way.c
@@ -0,0 +1,114 @@
+#include "bmw512-gate.h"
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+//#include "sph_keccak.h"
+#include "bmw-hash-4way.h"
+
+#if defined(BMW512_8WAY)
+
+void bmw512hash_8way(void *state, const void *input)
+{
+    bmw512_8way_context ctx;
+    bmw512_8way_init( &ctx );
+    bmw512_8way_update( &ctx, input, 80 );
+    bmw512_8way_close( &ctx, state );
+}
+
+int scanhash_bmw512_8way( struct work *work, uint32_t max_nonce,
+                          uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t vdata[24*8] __attribute__ ((aligned (128)));
+   uint32_t hash[16*8] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[49]);   // 3*16+1
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce - 8;
+   __m512i  *noncev = (__m512i*)vdata + 9;   // aligned
+   const uint32_t Htarg = ptarget[7];
+   int thr_id = mythr->id;
+
+   mm512_bswap32_intrlv80_8x64( vdata, pdata );
+   do {
+       *noncev = mm512_intrlv_blend_32( mm512_bswap_32(
+                _mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0 ,
+                                  n+3, 0, n+2, 0, n+1, 0, n  , 0 ) ), *noncev );
+
+      bmw512hash_8way( hash, vdata );
+
+      for ( int lane = 0; lane < 8; lane++ )
+      if ( unlikely( hash7[ lane<<1 ] < Htarg ) )
+      {
+          extr_lane_8x64( lane_hash, hash, lane, 256 );
+          if ( fulltest( lane_hash, ptarget ) )
+          {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+          }
+      }
+      n += 8;
+
+   } while ( likely( ( n < last_nonce ) && !work_restart[thr_id].restart) );
+
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+   
+#elif defined(BMW512_4WAY)
+
+//#ifdef BMW512_4WAY
+
+void bmw512hash_4way(void *state, const void *input)
+{
+    bmw512_4way_context ctx;
+    bmw512_4way_init( &ctx );
+    bmw512_4way_update( &ctx, input, 80 );
+    bmw512_4way_close( &ctx, state );
+}
+
+int scanhash_bmw512_4way( struct work *work, uint32_t max_nonce,
+                          uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t vdata[24*4] __attribute__ ((aligned (128)));
+   uint32_t hash[16*4] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[25]);   // 3*8+1
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce -  4;
+   __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+   const uint32_t Htarg = ptarget[7];
+    int thr_id = mythr->id;  // thr_id arg is deprecated
+
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
+   do {
+       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
+	
+      bmw512hash_4way( hash, vdata );
+
+      for ( int lane = 0; lane < 4; lane++ )
+      if ( unlikely( hash7[ lane<<1 ] < Htarg ) )
+//      if ( ( ( hash7[ lane<<1 ] & 0xFFFFFF00 ) == 0 ) )
+      {
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
+          if ( fulltest( lane_hash, ptarget ) )
+          {
+              pdata[19] = n + lane;
+              submit_lane_solution( work, lane_hash, mythr, lane );
+          }
+      }
+      n += 4;
+
+   } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
+
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+
+#endif
--- a/algo/bmw/bmw512-gate.c
+++ b/algo/bmw/bmw512-gate.c
@@ -0,0 +1,20 @@
+#include "bmw512-gate.h"
+
+bool register_bmw512_algo( algo_gate_t* gate )
+{
+  gate->optimizations = AVX2_OPT | AVX512_OPT;
+  opt_target_factor = 256.0;
+#if defined (BMW512_8WAY)
+  gate->scanhash  = (void*)&scanhash_bmw512_8way;
+  gate->hash      = (void*)&bmw512hash_8way;
+#elif defined (BMW512_4WAY)
+  gate->scanhash  = (void*)&scanhash_bmw512_4way;
+  gate->hash      = (void*)&bmw512hash_4way;
+#else
+  gate->scanhash        = (void*)&scanhash_bmw512;
+  gate->hash            = (void*)&bmw512hash;
+#endif
+  return true;
+};
+
+
--- a/algo/bmw/bmw512-gate.h
+++ b/algo/bmw/bmw512-gate.h
@@ -0,0 +1,33 @@
+#ifndef BMW512_GATE_H__
+#define BMW512_GATE_H__ 1
+
+#include "algo-gate-api.h"
+#include <stdint.h>
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define BMW512_8WAY 1
+#elif defined(__AVX2__)
+  #define BMW512_4WAY 1
+#endif
+
+#if defined(BMW512_8WAY)
+
+void bmw512hash_8way( void *state, const void *input );
+int scanhash_bmw512_8way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+#elif defined(BMW512_4WAY)
+
+void bmw512hash_4way( void *state, const void *input );
+int scanhash_bmw512_4way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+#else
+
+void bmw512hash( void *state, const void *input );
+int scanhash_bmw512( struct work *work, uint32_t max_nonce,
+                    uint64_t *hashes_done, struct thr_info *mythr );
+
+#endif
+
+#endif
--- a/algo/bmw/bmw512-hash-4way.c
+++ b/algo/bmw/bmw512-hash-4way.c
--- a/algo/bmw/bmw512.c
+++ b/algo/bmw/bmw512.c
@@ -0,0 +1,53 @@
+#include "algo-gate-api.h"
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+
+#include "sph_bmw.h"
+
+void bmw512hash(void *state, const void *input)
+{
+    sph_bmw512_context ctx;
+    uint32_t hash[32];	
+   
+    sph_bmw512_init( &ctx );
+    sph_bmw512( &ctx,input, 80 );
+    sph_bmw512_close( &ctx, hash );
+
+    memcpy( state, hash, 32 );
+}
+
+int scanhash_bmw512( struct work *work, uint32_t max_nonce,
+                     uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+	uint32_t n = pdata[19] - 1;
+	const uint32_t first_nonce = pdata[19];
+	//const uint32_t Htarg = ptarget[7];
+   int thr_id = mythr->id;  // thr_id arg is deprecated
+
+	uint32_t _ALIGN(32) hash64[8];
+	uint32_t endiandata[32];
+
+   for (int i=0; i < 19; i++) 
+           be32enc(&endiandata[i], pdata[i]);
+
+	do {
+	
+		pdata[19] = ++n;
+		be32enc(&endiandata[19], n); 
+		bmw512hash(hash64, endiandata);
+        if (((hash64[7]&0xFFFFFF00)==0) && 
+				fulltest(hash64, ptarget)) {
+            *hashes_done = n - first_nonce + 1;
+			return true;
+		}
+	} while (n < max_nonce && !work_restart[thr_id].restart);
+	
+	*hashes_done = n - first_nonce + 1;
+	pdata[19] = n;
+	return 0;
+}
+
--- a/algo/bmw/sse2/bmw.c
+++ b/algo/bmw/sse2/bmw.c
@@ -1,519 +0,0 @@
-/* $Id: bmw.c 227 2010-06-16 17:28:38Z tp $ */
-/*
- * BMW implementation.
- *
- * ==========================(LICENSE BEGIN)============================
- *
- * Copyright (c) 2007-2010  Projet RNRT SAPHIR
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
- * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * ===========================(LICENSE END)=============================
- *
- * @author   Thomas Pornin <thomas.pornin@cryptolog.com>
- */
-
-#include <stddef.h>
-#include <string.h>
-#include <limits.h>
-
-#ifdef __cplusplus
-extern "C"{
-#endif
-
-#include "../sph_bmw.h"
-
-#ifdef _MSC_VER
-#pragma warning (disable: 4146)
-#endif
-
-static const sph_u64 bmwIV512[] = {
-	SPH_C64(0x8081828384858687), SPH_C64(0x88898A8B8C8D8E8F),
-	SPH_C64(0x9091929394959697), SPH_C64(0x98999A9B9C9D9E9F),
-	SPH_C64(0xA0A1A2A3A4A5A6A7), SPH_C64(0xA8A9AAABACADAEAF),
-	SPH_C64(0xB0B1B2B3B4B5B6B7), SPH_C64(0xB8B9BABBBCBDBEBF),
-	SPH_C64(0xC0C1C2C3C4C5C6C7), SPH_C64(0xC8C9CACBCCCDCECF),
-	SPH_C64(0xD0D1D2D3D4D5D6D7), SPH_C64(0xD8D9DADBDCDDDEDF),
-	SPH_C64(0xE0E1E2E3E4E5E6E7), SPH_C64(0xE8E9EAEBECEDEEEF),
-	SPH_C64(0xF0F1F2F3F4F5F6F7), SPH_C64(0xF8F9FAFBFCFDFEFF)
-};
-
-#define XCAT(x, y)    XCAT_(x, y)
-#define XCAT_(x, y)   x ## y
-
-#define LPAR   (
-
-#define I16_16    0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15
-#define I16_17    1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16
-#define I16_18    2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17
-#define I16_19    3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18
-#define I16_20    4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19
-#define I16_21    5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20
-#define I16_22    6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21
-#define I16_23    7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22
-#define I16_24    8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23
-#define I16_25    9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24
-#define I16_26   10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25
-#define I16_27   11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26
-#define I16_28   12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27
-#define I16_29   13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28
-#define I16_30   14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29
-#define I16_31   15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
-
-#define M16_16    0,  1,  3,  4,  7, 10, 11
-#define M16_17    1,  2,  4,  5,  8, 11, 12
-#define M16_18    2,  3,  5,  6,  9, 12, 13
-#define M16_19    3,  4,  6,  7, 10, 13, 14
-#define M16_20    4,  5,  7,  8, 11, 14, 15
-#define M16_21    5,  6,  8,  9, 12, 15, 16
-#define M16_22    6,  7,  9, 10, 13,  0,  1
-#define M16_23    7,  8, 10, 11, 14,  1,  2
-#define M16_24    8,  9, 11, 12, 15,  2,  3
-#define M16_25    9, 10, 12, 13,  0,  3,  4
-#define M16_26   10, 11, 13, 14,  1,  4,  5
-#define M16_27   11, 12, 14, 15,  2,  5,  6
-#define M16_28   12, 13, 15, 16,  3,  6,  7
-#define M16_29   13, 14,  0,  1,  4,  7,  8
-#define M16_30   14, 15,  1,  2,  5,  8,  9
-#define M16_31   15, 16,  2,  3,  6,  9, 10
-
-#define ss0(x)    (((x) >> 1) ^ SPH_T32((x) << 3) \
-                  ^ SPH_ROTL32(x,  4) ^ SPH_ROTL32(x, 19))
-#define ss1(x)    (((x) >> 1) ^ SPH_T32((x) << 2) \
-                  ^ SPH_ROTL32(x,  8) ^ SPH_ROTL32(x, 23))
-#define ss2(x)    (((x) >> 2) ^ SPH_T32((x) << 1) \
-                  ^ SPH_ROTL32(x, 12) ^ SPH_ROTL32(x, 25))
-#define ss3(x)    (((x) >> 2) ^ SPH_T32((x) << 2) \
-                  ^ SPH_ROTL32(x, 15) ^ SPH_ROTL32(x, 29))
-#define ss4(x)    (((x) >> 1) ^ (x))
-#define ss5(x)    (((x) >> 2) ^ (x))
-#define rs1(x)    SPH_ROTL32(x,  3)
-#define rs2(x)    SPH_ROTL32(x,  7)
-#define rs3(x)    SPH_ROTL32(x, 13)
-#define rs4(x)    SPH_ROTL32(x, 16)
-#define rs5(x)    SPH_ROTL32(x, 19)
-#define rs6(x)    SPH_ROTL32(x, 23)
-#define rs7(x)    SPH_ROTL32(x, 27)
-
-#define Ks(j)   SPH_T32((sph_u32)(j) * SPH_C32(0x05555555))
-
-#define add_elt_s(mf, hf, j0m, j1m, j3m, j4m, j7m, j10m, j11m, j16) \
-	(SPH_T32(SPH_ROTL32(mf(j0m), j1m) + SPH_ROTL32(mf(j3m), j4m) \
-		- SPH_ROTL32(mf(j10m), j11m) + Ks(j16)) ^ hf(j7m))
-
-#define expand1s_inner(qf, mf, hf, i16, \
-		i0, i1, i2, i3, i4, i5, i6, i7, i8, \
-		i9, i10, i11, i12, i13, i14, i15, \
-		i0m, i1m, i3m, i4m, i7m, i10m, i11m) \
-	SPH_T32(ss1(qf(i0)) + ss2(qf(i1)) + ss3(qf(i2)) + ss0(qf(i3)) \
-		+ ss1(qf(i4)) + ss2(qf(i5)) + ss3(qf(i6)) + ss0(qf(i7)) \
-		+ ss1(qf(i8)) + ss2(qf(i9)) + ss3(qf(i10)) + ss0(qf(i11)) \
-		+ ss1(qf(i12)) + ss2(qf(i13)) + ss3(qf(i14)) + ss0(qf(i15)) \
-		+ add_elt_s(mf, hf, i0m, i1m, i3m, i4m, i7m, i10m, i11m, i16))
-
-#define expand1s(qf, mf, hf, i16) \
-	expand1s_(qf, mf, hf, i16, I16_ ## i16, M16_ ## i16)
-#define expand1s_(qf, mf, hf, i16, ix, iy) \
-	expand1s_inner LPAR qf, mf, hf, i16, ix, iy)
-
-#define expand2s_inner(qf, mf, hf, i16, \
-		i0, i1, i2, i3, i4, i5, i6, i7, i8, \
-		i9, i10, i11, i12, i13, i14, i15, \
-		i0m, i1m, i3m, i4m, i7m, i10m, i11m) \
-	SPH_T32(qf(i0) + rs1(qf(i1)) + qf(i2) + rs2(qf(i3)) \
-		+ qf(i4) + rs3(qf(i5)) + qf(i6) + rs4(qf(i7)) \
-		+ qf(i8) + rs5(qf(i9)) + qf(i10) + rs6(qf(i11)) \
-		+ qf(i12) + rs7(qf(i13)) + ss4(qf(i14)) + ss5(qf(i15)) \
-		+ add_elt_s(mf, hf, i0m, i1m, i3m, i4m, i7m, i10m, i11m, i16))
-
-#define expand2s(qf, mf, hf, i16) \
-	expand2s_(qf, mf, hf, i16, I16_ ## i16, M16_ ## i16)
-#define expand2s_(qf, mf, hf, i16, ix, iy) \
-	expand2s_inner LPAR qf, mf, hf, i16, ix, iy)
-
-#if SPH_64
-
-#define sb0(x)    (((x) >> 1) ^ SPH_T64((x) << 3) \
-                  ^ SPH_ROTL64(x,  4) ^ SPH_ROTL64(x, 37))
-#define sb1(x)    (((x) >> 1) ^ SPH_T64((x) << 2) \
-                  ^ SPH_ROTL64(x, 13) ^ SPH_ROTL64(x, 43))
-#define sb2(x)    (((x) >> 2) ^ SPH_T64((x) << 1) \
-                  ^ SPH_ROTL64(x, 19) ^ SPH_ROTL64(x, 53))
-#define sb3(x)    (((x) >> 2) ^ SPH_T64((x) << 2) \
-                  ^ SPH_ROTL64(x, 28) ^ SPH_ROTL64(x, 59))
-#define sb4(x)    (((x) >> 1) ^ (x))
-#define sb5(x)    (((x) >> 2) ^ (x))
-#define rb1(x)    SPH_ROTL64(x,  5)
-#define rb2(x)    SPH_ROTL64(x, 11)
-#define rb3(x)    SPH_ROTL64(x, 27)
-#define rb4(x)    SPH_ROTL64(x, 32)
-#define rb5(x)    SPH_ROTL64(x, 37)
-#define rb6(x)    SPH_ROTL64(x, 43)
-#define rb7(x)    SPH_ROTL64(x, 53)
-
-#define Kb(j)   SPH_T64((sph_u64)(j) * SPH_C64(0x0555555555555555))
-
-#if 0
-
-static const sph_u64 Kb_tab[] = {
-	Kb(16), Kb(17), Kb(18), Kb(19), Kb(20), Kb(21), Kb(22), Kb(23),
-	Kb(24), Kb(25), Kb(26), Kb(27), Kb(28), Kb(29), Kb(30), Kb(31)
-};
-
-#define rol_off(mf, j, off) \
-	SPH_ROTL64(mf(((j) + (off)) & 15), (((j) + (off)) & 15) + 1)
-
-#define add_elt_b(mf, hf, j) \
-	(SPH_T64(rol_off(mf, j, 0) + rol_off(mf, j, 3) \
-		- rol_off(mf, j, 10) + Kb_tab[j]) ^ hf(((j) + 7) & 15))
-
-#define expand1b(qf, mf, hf, i) \
-	SPH_T64(sb1(qf((i) - 16)) + sb2(qf((i) - 15)) \
-		+ sb3(qf((i) - 14)) + sb0(qf((i) - 13)) \
-		+ sb1(qf((i) - 12)) + sb2(qf((i) - 11)) \
-		+ sb3(qf((i) - 10)) + sb0(qf((i) - 9)) \
-		+ sb1(qf((i) - 8)) + sb2(qf((i) - 7)) \
-		+ sb3(qf((i) - 6)) + sb0(qf((i) - 5)) \
-		+ sb1(qf((i) - 4)) + sb2(qf((i) - 3)) \
-		+ sb3(qf((i) - 2)) + sb0(qf((i) - 1)) \
-		+ add_elt_b(mf, hf, (i) - 16))
-
-#define expand2b(qf, mf, hf, i) \
-	SPH_T64(qf((i) - 16) + rb1(qf((i) - 15)) \
-		+ qf((i) - 14) + rb2(qf((i) - 13)) \
-		+ qf((i) - 12) + rb3(qf((i) - 11)) \
-		+ qf((i) - 10) + rb4(qf((i) - 9)) \
-		+ qf((i) - 8) + rb5(qf((i) - 7)) \
-		+ qf((i) - 6) + rb6(qf((i) - 5)) \
-		+ qf((i) - 4) + rb7(qf((i) - 3)) \
-		+ sb4(qf((i) - 2)) + sb5(qf((i) - 1)) \
-		+ add_elt_b(mf, hf, (i) - 16))
-
-#else
-
-#define add_elt_b(mf, hf, j0m, j1m, j3m, j4m, j7m, j10m, j11m, j16) \
-	(SPH_T64(SPH_ROTL64(mf(j0m), j1m) + SPH_ROTL64(mf(j3m), j4m) \
-		- SPH_ROTL64(mf(j10m), j11m) + Kb(j16)) ^ hf(j7m))
-
-#define expand1b_inner(qf, mf, hf, i16, \
-		i0, i1, i2, i3, i4, i5, i6, i7, i8, \
-		i9, i10, i11, i12, i13, i14, i15, \
-		i0m, i1m, i3m, i4m, i7m, i10m, i11m) \
-	SPH_T64(sb1(qf(i0)) + sb2(qf(i1)) + sb3(qf(i2)) + sb0(qf(i3)) \
-		+ sb1(qf(i4)) + sb2(qf(i5)) + sb3(qf(i6)) + sb0(qf(i7)) \
-		+ sb1(qf(i8)) + sb2(qf(i9)) + sb3(qf(i10)) + sb0(qf(i11)) \
-		+ sb1(qf(i12)) + sb2(qf(i13)) + sb3(qf(i14)) + sb0(qf(i15)) \
-		+ add_elt_b(mf, hf, i0m, i1m, i3m, i4m, i7m, i10m, i11m, i16))
-
-#define expand1b(qf, mf, hf, i16) \
-	expand1b_(qf, mf, hf, i16, I16_ ## i16, M16_ ## i16)
-#define expand1b_(qf, mf, hf, i16, ix, iy) \
-	expand1b_inner LPAR qf, mf, hf, i16, ix, iy)
-
-#define expand2b_inner(qf, mf, hf, i16, \
-		i0, i1, i2, i3, i4, i5, i6, i7, i8, \
-		i9, i10, i11, i12, i13, i14, i15, \
-		i0m, i1m, i3m, i4m, i7m, i10m, i11m) \
-	SPH_T64(qf(i0) + rb1(qf(i1)) + qf(i2) + rb2(qf(i3)) \
-		+ qf(i4) + rb3(qf(i5)) + qf(i6) + rb4(qf(i7)) \
-		+ qf(i8) + rb5(qf(i9)) + qf(i10) + rb6(qf(i11)) \
-		+ qf(i12) + rb7(qf(i13)) + sb4(qf(i14)) + sb5(qf(i15)) \
-		+ add_elt_b(mf, hf, i0m, i1m, i3m, i4m, i7m, i10m, i11m, i16))
-
-#define expand2b(qf, mf, hf, i16) \
-	expand2b_(qf, mf, hf, i16, I16_ ## i16, M16_ ## i16)
-#define expand2b_(qf, mf, hf, i16, ix, iy) \
-	expand2b_inner LPAR qf, mf, hf, i16, ix, iy)
-
-#endif
-
-#endif
-
-#define MAKE_W(tt, i0, op01, i1, op12, i2, op23, i3, op34, i4) \
-	tt((M(i0) ^ H(i0)) op01 (M(i1) ^ H(i1)) op12 (M(i2) ^ H(i2)) \
-	op23 (M(i3) ^ H(i3)) op34 (M(i4) ^ H(i4)))
-
-#define Ws0    MAKE_W(SPH_T32,  5, -,  7, +, 10, +, 13, +, 14)
-#define Ws1    MAKE_W(SPH_T32,  6, -,  8, +, 11, +, 14, -, 15)
-#define Ws2    MAKE_W(SPH_T32,  0, +,  7, +,  9, -, 12, +, 15)
-#define Ws3    MAKE_W(SPH_T32,  0, -,  1, +,  8, -, 10, +, 13)
-#define Ws4    MAKE_W(SPH_T32,  1, +,  2, +,  9, -, 11, -, 14)
-#define Ws5    MAKE_W(SPH_T32,  3, -,  2, +, 10, -, 12, +, 15)
-#define Ws6    MAKE_W(SPH_T32,  4, -,  0, -,  3, -, 11, +, 13)
-#define Ws7    MAKE_W(SPH_T32,  1, -,  4, -,  5, -, 12, -, 14)
-#define Ws8    MAKE_W(SPH_T32,  2, -,  5, -,  6, +, 13, -, 15)
-#define Ws9    MAKE_W(SPH_T32,  0, -,  3, +,  6, -,  7, +, 14)
-#define Ws10   MAKE_W(SPH_T32,  8, -,  1, -,  4, -,  7, +, 15)
-#define Ws11   MAKE_W(SPH_T32,  8, -,  0, -,  2, -,  5, +,  9)
-#define Ws12   MAKE_W(SPH_T32,  1, +,  3, -,  6, -,  9, +, 10)
-#define Ws13   MAKE_W(SPH_T32,  2, +,  4, +,  7, +, 10, +, 11)
-#define Ws14   MAKE_W(SPH_T32,  3, -,  5, +,  8, -, 11, -, 12)
-#define Ws15   MAKE_W(SPH_T32, 12, -,  4, -,  6, -,  9, +, 13)
-
-#define MAKE_Qas   do { \
-		qt[ 0] = SPH_T32(ss0(Ws0 ) + H( 1)); \
-		qt[ 1] = SPH_T32(ss1(Ws1 ) + H( 2)); \
-		qt[ 2] = SPH_T32(ss2(Ws2 ) + H( 3)); \
-		qt[ 3] = SPH_T32(ss3(Ws3 ) + H( 4)); \
-		qt[ 4] = SPH_T32(ss4(Ws4 ) + H( 5)); \
-		qt[ 5] = SPH_T32(ss0(Ws5 ) + H( 6)); \
-		qt[ 6] = SPH_T32(ss1(Ws6 ) + H( 7)); \
-		qt[ 7] = SPH_T32(ss2(Ws7 ) + H( 8)); \
-		qt[ 8] = SPH_T32(ss3(Ws8 ) + H( 9)); \
-		qt[ 9] = SPH_T32(ss4(Ws9 ) + H(10)); \
-		qt[10] = SPH_T32(ss0(Ws10) + H(11)); \
-		qt[11] = SPH_T32(ss1(Ws11) + H(12)); \
-		qt[12] = SPH_T32(ss2(Ws12) + H(13)); \
-		qt[13] = SPH_T32(ss3(Ws13) + H(14)); \
-		qt[14] = SPH_T32(ss4(Ws14) + H(15)); \
-		qt[15] = SPH_T32(ss0(Ws15) + H( 0)); \
-	} while (0)
-
-#define MAKE_Qbs   do { \
-		qt[16] = expand1s(Qs, M, H, 16); \
-		qt[17] = expand1s(Qs, M, H, 17); \
-		qt[18] = expand2s(Qs, M, H, 18); \
-		qt[19] = expand2s(Qs, M, H, 19); \
-		qt[20] = expand2s(Qs, M, H, 20); \
-		qt[21] = expand2s(Qs, M, H, 21); \
-		qt[22] = expand2s(Qs, M, H, 22); \
-		qt[23] = expand2s(Qs, M, H, 23); \
-		qt[24] = expand2s(Qs, M, H, 24); \
-		qt[25] = expand2s(Qs, M, H, 25); \
-		qt[26] = expand2s(Qs, M, H, 26); \
-		qt[27] = expand2s(Qs, M, H, 27); \
-		qt[28] = expand2s(Qs, M, H, 28); \
-		qt[29] = expand2s(Qs, M, H, 29); \
-		qt[30] = expand2s(Qs, M, H, 30); \
-		qt[31] = expand2s(Qs, M, H, 31); \
-	} while (0)
-
-#define MAKE_Qs   do { \
-		MAKE_Qas; \
-		MAKE_Qbs; \
-	} while (0)
-
-#define Qs(j)   (qt[j])
-
-#define Wb0    MAKE_W(SPH_T64,  5, -,  7, +, 10, +, 13, +, 14)
-#define Wb1    MAKE_W(SPH_T64,  6, -,  8, +, 11, +, 14, -, 15)
-#define Wb2    MAKE_W(SPH_T64,  0, +,  7, +,  9, -, 12, +, 15)
-#define Wb3    MAKE_W(SPH_T64,  0, -,  1, +,  8, -, 10, +, 13)
-#define Wb4    MAKE_W(SPH_T64,  1, +,  2, +,  9, -, 11, -, 14)
-#define Wb5    MAKE_W(SPH_T64,  3, -,  2, +, 10, -, 12, +, 15)
-#define Wb6    MAKE_W(SPH_T64,  4, -,  0, -,  3, -, 11, +, 13)
-#define Wb7    MAKE_W(SPH_T64,  1, -,  4, -,  5, -, 12, -, 14)
-#define Wb8    MAKE_W(SPH_T64,  2, -,  5, -,  6, +, 13, -, 15)
-#define Wb9    MAKE_W(SPH_T64,  0, -,  3, +,  6, -,  7, +, 14)
-#define Wb10   MAKE_W(SPH_T64,  8, -,  1, -,  4, -,  7, +, 15)
-#define Wb11   MAKE_W(SPH_T64,  8, -,  0, -,  2, -,  5, +,  9)
-#define Wb12   MAKE_W(SPH_T64,  1, +,  3, -,  6, -,  9, +, 10)
-#define Wb13   MAKE_W(SPH_T64,  2, +,  4, +,  7, +, 10, +, 11)
-#define Wb14   MAKE_W(SPH_T64,  3, -,  5, +,  8, -, 11, -, 12)
-#define Wb15   MAKE_W(SPH_T64, 12, -,  4, -,  6, -,  9, +, 13)
-
-#define MAKE_Qab   do { \
-		qt[ 0] = SPH_T64(sb0(Wb0 ) + H( 1)); \
-		qt[ 1] = SPH_T64(sb1(Wb1 ) + H( 2)); \
-		qt[ 2] = SPH_T64(sb2(Wb2 ) + H( 3)); \
-		qt[ 3] = SPH_T64(sb3(Wb3 ) + H( 4)); \
-		qt[ 4] = SPH_T64(sb4(Wb4 ) + H( 5)); \
-		qt[ 5] = SPH_T64(sb0(Wb5 ) + H( 6)); \
-		qt[ 6] = SPH_T64(sb1(Wb6 ) + H( 7)); \
-		qt[ 7] = SPH_T64(sb2(Wb7 ) + H( 8)); \
-		qt[ 8] = SPH_T64(sb3(Wb8 ) + H( 9)); \
-		qt[ 9] = SPH_T64(sb4(Wb9 ) + H(10)); \
-		qt[10] = SPH_T64(sb0(Wb10) + H(11)); \
-		qt[11] = SPH_T64(sb1(Wb11) + H(12)); \
-		qt[12] = SPH_T64(sb2(Wb12) + H(13)); \
-		qt[13] = SPH_T64(sb3(Wb13) + H(14)); \
-		qt[14] = SPH_T64(sb4(Wb14) + H(15)); \
-		qt[15] = SPH_T64(sb0(Wb15) + H( 0)); \
-	} while (0)
-
-#define MAKE_Qbb   do { \
-		qt[16] = expand1b(Qb, M, H, 16); \
-		qt[17] = expand1b(Qb, M, H, 17); \
-		qt[18] = expand2b(Qb, M, H, 18); \
-		qt[19] = expand2b(Qb, M, H, 19); \
-		qt[20] = expand2b(Qb, M, H, 20); \
-		qt[21] = expand2b(Qb, M, H, 21); \
-		qt[22] = expand2b(Qb, M, H, 22); \
-		qt[23] = expand2b(Qb, M, H, 23); \
-		qt[24] = expand2b(Qb, M, H, 24); \
-		qt[25] = expand2b(Qb, M, H, 25); \
-		qt[26] = expand2b(Qb, M, H, 26); \
-		qt[27] = expand2b(Qb, M, H, 27); \
-		qt[28] = expand2b(Qb, M, H, 28); \
-		qt[29] = expand2b(Qb, M, H, 29); \
-		qt[30] = expand2b(Qb, M, H, 30); \
-		qt[31] = expand2b(Qb, M, H, 31); \
-	} while (0)
-
-#define MAKE_Qb   do { \
-		MAKE_Qab; \
-		MAKE_Qbb; \
-	} while (0)
-
-#define Qb(j)   (qt[j])
-
-#define FOLD(type, mkQ, tt, rol, mf, qf, dhf)   do { \
-		type qt[32], xl, xh; \
-		mkQ; \
-		xl = qf(16) ^ qf(17) ^ qf(18) ^ qf(19) \
-			^ qf(20) ^ qf(21) ^ qf(22) ^ qf(23); \
-		xh = xl ^ qf(24) ^ qf(25) ^ qf(26) ^ qf(27) \
-			^ qf(28) ^ qf(29) ^ qf(30) ^ qf(31); \
-		dhf( 0) = tt(((xh <<  5) ^ (qf(16) >>  5) ^ mf( 0)) \
-			+ (xl ^ qf(24) ^ qf( 0))); \
-		dhf( 1) = tt(((xh >>  7) ^ (qf(17) <<  8) ^ mf( 1)) \
-			+ (xl ^ qf(25) ^ qf( 1))); \
-		dhf( 2) = tt(((xh >>  5) ^ (qf(18) <<  5) ^ mf( 2)) \
-			+ (xl ^ qf(26) ^ qf( 2))); \
-		dhf( 3) = tt(((xh >>  1) ^ (qf(19) <<  5) ^ mf( 3)) \
-			+ (xl ^ qf(27) ^ qf( 3))); \
-		dhf( 4) = tt(((xh >>  3) ^ (qf(20) <<  0) ^ mf( 4)) \
-			+ (xl ^ qf(28) ^ qf( 4))); \
-		dhf( 5) = tt(((xh <<  6) ^ (qf(21) >>  6) ^ mf( 5)) \
-			+ (xl ^ qf(29) ^ qf( 5))); \
-		dhf( 6) = tt(((xh >>  4) ^ (qf(22) <<  6) ^ mf( 6)) \
-			+ (xl ^ qf(30) ^ qf( 6))); \
-		dhf( 7) = tt(((xh >> 11) ^ (qf(23) <<  2) ^ mf( 7)) \
-			+ (xl ^ qf(31) ^ qf( 7))); \
-		dhf( 8) = tt(rol(dhf(4),  9) + (xh ^ qf(24) ^ mf( 8)) \
-			+ ((xl << 8) ^ qf(23) ^ qf( 8))); \
-		dhf( 9) = tt(rol(dhf(5), 10) + (xh ^ qf(25) ^ mf( 9)) \
-			+ ((xl >> 6) ^ qf(16) ^ qf( 9))); \
-		dhf(10) = tt(rol(dhf(6), 11) + (xh ^ qf(26) ^ mf(10)) \
-			+ ((xl << 6) ^ qf(17) ^ qf(10))); \
-		dhf(11) = tt(rol(dhf(7), 12) + (xh ^ qf(27) ^ mf(11)) \
-			+ ((xl << 4) ^ qf(18) ^ qf(11))); \
-		dhf(12) = tt(rol(dhf(0), 13) + (xh ^ qf(28) ^ mf(12)) \
-			+ ((xl >> 3) ^ qf(19) ^ qf(12))); \
-		dhf(13) = tt(rol(dhf(1), 14) + (xh ^ qf(29) ^ mf(13)) \
-			+ ((xl >> 4) ^ qf(20) ^ qf(13))); \
-		dhf(14) = tt(rol(dhf(2), 15) + (xh ^ qf(30) ^ mf(14)) \
-			+ ((xl >> 7) ^ qf(21) ^ qf(14))); \
-		dhf(15) = tt(rol(dhf(3), 16) + (xh ^ qf(31) ^ mf(15)) \
-			+ ((xl >> 2) ^ qf(22) ^ qf(15))); \
-	} while (0)
-
-#define FOLDs   FOLD(sph_u32, MAKE_Qs, SPH_T32, SPH_ROTL32, M, Qs, dH)
-
-#define FOLDb   FOLD(sph_u64, MAKE_Qb, SPH_T64, SPH_ROTL64, M, Qb, dH)
-
-#define DECL_BMW \
-    sph_u64 bmwH[16]; \
-
-/* load initial constants */
-#define BMW_I \
-do { \
-    memcpy(bmwH, bmwIV512, sizeof bmwH); \
-    hashptr = 0; \
-    hashctA = 0; \
-} while (0) 
-
-/* load hash for loop */
-#define BMW_U \
-do { \
-    const void *data = hash; \
-    size_t len = 64; \
-    unsigned char *buf; \
-    \
-    hashctA += (sph_u64)len << 3; \
-    buf = hashbuf; \
-    memcpy(buf, data, 64); \
-    hashptr = 64; \
-} while (0)  
-
-
-/* bmw512 hash loaded */
-/* hash = blake512(loaded) */
-#define BMW_C \
-do { \
-    void *dst = hash; \
-    size_t out_size_w64 = 8; \
-    unsigned char *data; \
-    sph_u64 *dh; \
-    unsigned char *out; \
-    size_t ptr, u, v; \
-    unsigned z; \
-    sph_u64 h1[16], h2[16], *h; \
-    data = hashbuf; \
-    ptr = hashptr; \
-    z = 0x80 >> 0; \
-    data[ptr ++] = ((0 & -z) | z) & 0xFF; \
-    memset(data + ptr, 0, (sizeof(char)*128) - 8 - ptr); \
-    sph_enc64le_aligned(data + (sizeof(char)*128) - 8, \
-    SPH_T64(hashctA + 0)); \
-    /* for break loop */ \
-    /* one copy of inline FOLD */ \
-    /* FOLD uses, */ \
-    /* uint64 *h, data */ \
-    /* uint64 dh, state */ \
-        h = bmwH; \
-        dh = h2; \
-    for (;;) { \
-        FOLDb; \
-        /* dh gets changed for 2nd run */ \
-        if (dh == h1) break; \
-        for (u = 0; u < 16; u ++) \
-        sph_enc64le_aligned(data + 8 * u, h2[u]); \
-        dh = h1; \
-        h = (sph_u64*)final_b; \
-    } \
-    /* end wrapped for break loop */ \
-    out = dst; \
-    for (u = 0, v = 16 - out_size_w64; u < out_size_w64; u ++, v ++) \
-    sph_enc64le(out + 8 * u, h1[v]); \
-} while (0) 
-
-/*
-static void
-compress_big(const unsigned char *data, const sph_u64 h[16], sph_u64 dh[16])
-{
-
-#define M(x)    sph_dec64le_aligned(data + 8 * (x))
-#define H(x)    (h[x])
-#define dH(x)   (dh[x])
-
-	FOLDb;
-
-#undef M
-#undef H
-#undef dH
-}
-*/
-
-static const sph_u64 final_b[16] = {
-	SPH_C64(0xaaaaaaaaaaaaaaa0), SPH_C64(0xaaaaaaaaaaaaaaa1),
-	SPH_C64(0xaaaaaaaaaaaaaaa2), SPH_C64(0xaaaaaaaaaaaaaaa3),
-	SPH_C64(0xaaaaaaaaaaaaaaa4), SPH_C64(0xaaaaaaaaaaaaaaa5),
-	SPH_C64(0xaaaaaaaaaaaaaaa6), SPH_C64(0xaaaaaaaaaaaaaaa7),
-	SPH_C64(0xaaaaaaaaaaaaaaa8), SPH_C64(0xaaaaaaaaaaaaaaa9),
-	SPH_C64(0xaaaaaaaaaaaaaaaa), SPH_C64(0xaaaaaaaaaaaaaaab),
-	SPH_C64(0xaaaaaaaaaaaaaaac), SPH_C64(0xaaaaaaaaaaaaaaad),
-	SPH_C64(0xaaaaaaaaaaaaaaae), SPH_C64(0xaaaaaaaaaaaaaaaf)
-};
-
-
-#ifdef __cplusplus
-}
-#endif
--- a/algo/bmw/sse2/sph_bmw.h
+++ b/algo/bmw/sse2/sph_bmw.h
@@ -1,61 +0,0 @@
-/* $Id: sph_bmw.h 216 2010-06-08 09:46:57Z tp $ */
-/**
- * BMW interface. BMW (aka "Blue Midnight Wish") is a family of
- * functions which differ by their output size; this implementation
- * defines BMW for output sizes 224, 256, 384 and 512 bits.
- *
- * ==========================(LICENSE BEGIN)============================
- *
- * Copyright (c) 2007-2010  Projet RNRT SAPHIR
- * 
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice shall be
- * included in all copies or substantial portions of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
- * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * ===========================(LICENSE END)=============================
- *
- * @file     sph_bmw.h
- * @author   Thomas Pornin <thomas.pornin@cryptolog.com>
- */
-
-#ifndef SPH_BMW_H__
-#define SPH_BMW_H__
-
-#ifdef __cplusplus
-extern "C"{
-#endif
-
-#include <stddef.h>
-#include "sph_types.h"
-
-#define SPH_SIZE_bmw512   512
-
-typedef struct {
-#ifndef DOXYGEN_IGNORE
-	sph_u64 bmwH[16];
-#endif
-} sph_bmw_big_context;
-
-typedef sph_bmw_big_context sph_bmw512_context;
-
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif
--- a/algo/cryptonight/cryptolight.c
+++ b/algo/cryptonight/cryptolight.c
@@ -242,6 +242,8 @@ void cryptolight_hash(void* output, const void* input, int len) {
 	free(ctx);
 }

+#if defined(__AES__)
+
 static void cryptolight_hash_ctx_aes_ni(void* output, const void* input,
                       int len, struct cryptonight_ctx* ctx)
 {
@@ -312,6 +314,8 @@ static void cryptolight_hash_ctx_aes_ni(void* output, const void* input,
 	oaes_free((OAES_CTX **) &ctx->aes_ctx);
 }

+#endif
+
 int scanhash_cryptolight( struct work *work,
 		uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr)
 {
@@ -359,7 +363,6 @@ bool register_cryptolight_algo( algo_gate_t* gate )
  gate->scanhash  = (void*)&scanhash_cryptolight;
  gate->hash      = (void*)&cryptolight_hash;
  gate->hash_suw  = (void*)&cryptolight_hash; 
-  gate->get_max64 = (void*)&get_max64_0x40LL;
  return true;
 };

--- a/algo/cryptonight/cryptonight-common.c
+++ b/algo/cryptonight/cryptonight-common.c
@@ -111,7 +111,6 @@ bool register_cryptonight_algo( algo_gate_t* gate )
  gate->scanhash         = (void*)&scanhash_cryptonight;
  gate->hash             = (void*)&cryptonight_hash;
  gate->hash_suw         = (void*)&cryptonight_hash_suw;  
-  gate->get_max64        = (void*)&get_max64_0x40LL;
  return true;
 };

@@ -123,7 +122,6 @@ bool register_cryptonightv7_algo( algo_gate_t* gate )
  gate->scanhash      = (void*)&scanhash_cryptonight;
  gate->hash          = (void*)&cryptonight_hash;
  gate->hash_suw      = (void*)&cryptonight_hash_suw;
-  gate->get_max64     = (void*)&get_max64_0x40LL;
  return true;
 };

--- a/algo/cubehash/cube-hash-2way.c
+++ b/algo/cubehash/cube-hash-2way.c
@@ -7,6 +7,7 @@

 // 2x128

+
 // The result of hashing 10 rounds of initial data which consists of params
 // zero padded.
 static const uint64_t IV256[] =
@@ -25,12 +26,193 @@ static const uint64_t IV512[] =
 0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
 };

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// 4 way 128 is handy to avoid reinterleaving in many algos.
+// If reinterleaving is necessary it may be more efficient to use
+// 2 way 256. The same transform code should work for both.
+
+static void transform_4way( cube_4way_context *sp )
+{
+    int r;
+    const int rounds = sp->rounds;
+
+    __m512i x0, x1, x2, x3, x4, x5, x6, x7, y0, y1;
+
+    x0 = _mm512_load_si512( (__m512i*)sp->h     );
+    x1 = _mm512_load_si512( (__m512i*)sp->h + 1 );
+    x2 = _mm512_load_si512( (__m512i*)sp->h + 2 );
+    x3 = _mm512_load_si512( (__m512i*)sp->h + 3 );
+    x4 = _mm512_load_si512( (__m512i*)sp->h + 4 );
+    x5 = _mm512_load_si512( (__m512i*)sp->h + 5 );
+    x6 = _mm512_load_si512( (__m512i*)sp->h + 6 );
+    x7 = _mm512_load_si512( (__m512i*)sp->h + 7 );
+
+    for ( r = 0; r < rounds; ++r )
+    {
+        x4 = _mm512_add_epi32( x0, x4 );
+        x5 = _mm512_add_epi32( x1, x5 );
+        x6 = _mm512_add_epi32( x2, x6 );
+        x7 = _mm512_add_epi32( x3, x7 );
+        y0 = x0;
+        y1 = x1;
+        x0 = mm512_rol_32( x2, 7 );
+        x1 = mm512_rol_32( x3, 7 );
+        x2 = mm512_rol_32( y0, 7 );
+        x3 = mm512_rol_32( y1, 7 );
+        x0 = _mm512_xor_si512( x0, x4 );
+        x1 = _mm512_xor_si512( x1, x5 );
+        x2 = _mm512_xor_si512( x2, x6 );
+        x3 = _mm512_xor_si512( x3, x7 );
+        x4 = mm512_swap128_64( x4 );
+        x5 = mm512_swap128_64( x5 );
+        x6 = mm512_swap128_64( x6 );
+        x7 = mm512_swap128_64( x7 );
+        x4 = _mm512_add_epi32( x0, x4 );
+        x5 = _mm512_add_epi32( x1, x5 );
+        x6 = _mm512_add_epi32( x2, x6 );
+        x7 = _mm512_add_epi32( x3, x7 );
+        y0 = x0;
+        y1 = x2;
+        x0 = mm512_rol_32( x1, 11 );
+        x1 = mm512_rol_32( y0, 11 );
+        x2 = mm512_rol_32( x3, 11 );
+        x3 = mm512_rol_32( y1, 11 );
+        x0 = _mm512_xor_si512( x0, x4 );
+        x1 = _mm512_xor_si512( x1, x5 );
+        x2 = _mm512_xor_si512( x2, x6 );
+        x3 = _mm512_xor_si512( x3, x7 );
+        x4 = mm512_swap64_32( x4 );
+        x5 = mm512_swap64_32( x5 );
+        x6 = mm512_swap64_32( x6 );
+        x7 = mm512_swap64_32( x7 );
+    }
+
+    _mm512_store_si512( (__m512i*)sp->h,     x0 );
+    _mm512_store_si512( (__m512i*)sp->h + 1, x1 );
+    _mm512_store_si512( (__m512i*)sp->h + 2, x2 );
+    _mm512_store_si512( (__m512i*)sp->h + 3, x3 );
+    _mm512_store_si512( (__m512i*)sp->h + 4, x4 );
+    _mm512_store_si512( (__m512i*)sp->h + 5, x5 );
+    _mm512_store_si512( (__m512i*)sp->h + 6, x6 );
+    _mm512_store_si512( (__m512i*)sp->h + 7, x7 );
+}
+
+int cube_4way_init( cube_4way_context *sp, int hashbitlen, int rounds,
+                    int blockbytes )
+{
+    __m512i *h = (__m512i*)sp->h;
+    __m128i *iv = (__m128i*)( hashbitlen == 512 ? (__m128i*)IV512
+                                                : (__m128i*)IV256 );
+    sp->hashlen   = hashbitlen/128;
+    sp->blocksize = blockbytes/16;
+    sp->rounds    = rounds;
+    sp->pos       = 0;
+
+    h[ 0] = m512_const1_128( iv[0] );
+    h[ 1] = m512_const1_128( iv[1] );
+    h[ 2] = m512_const1_128( iv[2] );
+    h[ 3] = m512_const1_128( iv[3] );
+    h[ 4] = m512_const1_128( iv[4] );
+    h[ 5] = m512_const1_128( iv[5] );
+    h[ 6] = m512_const1_128( iv[6] );
+    h[ 7] = m512_const1_128( iv[7] );
+    h[ 0] = m512_const1_128( iv[0] );
+    h[ 1] = m512_const1_128( iv[1] );
+    h[ 2] = m512_const1_128( iv[2] );
+    h[ 3] = m512_const1_128( iv[3] );
+    h[ 4] = m512_const1_128( iv[4] );
+    h[ 5] = m512_const1_128( iv[5] );
+    h[ 6] = m512_const1_128( iv[6] );
+    h[ 7] = m512_const1_128( iv[7] );
+
+    return 0;
+}
+
+int cube_4way_update( cube_4way_context *sp, const void *data, size_t size )
+{
+    const int len = size >> 4;
+    const __m512i *in = (__m512i*)data;
+    int i;
+
+    for ( i = 0; i < len; i++ )
+    {
+        sp->h[ sp->pos ] = _mm512_xor_si512( sp->h[ sp->pos ], in[i] );
+        sp->pos++;
+        if ( sp->pos == sp->blocksize )
+        {
+           transform_4way( sp );
+           sp->pos = 0;
+        }
+    }
+    return 0;
+}
+
+int cube_4way_close( cube_4way_context *sp, void *output )
+{
+    __m512i *hash = (__m512i*)output;
+    int i;
+
+    // pos is zero for 64 byte data, 1 for 80 byte data.
+    sp->h[ sp->pos ] = _mm512_xor_si512( sp->h[ sp->pos ],
+                                 m512_const2_64( 0, 0x0000000000000080 ) );
+    transform_4way( sp );
+
+    sp->h[7] = _mm512_xor_si512( sp->h[7],
+                                 m512_const2_64( 0x0000000100000000, 0 ) );
+
+    for ( i = 0; i < 10; ++i ) 
+       transform_4way( sp );
+
+    memcpy( hash, sp->h, sp->hashlen<<6 );
+    return 0;
+}
+
+int cube_4way_update_close( cube_4way_context *sp, void *output,
+                               const void *data, size_t size )
+{
+    const int len = size >> 4;
+    const __m512i *in = (__m512i*)data;
+    __m512i *hash = (__m512i*)output;
+    int i;
+
+    for ( i = 0; i < len; i++ )
+    {
+        sp->h[ sp->pos ] = _mm512_xor_si512( sp->h[ sp->pos ], in[i] );
+        sp->pos++;
+        if ( sp->pos == sp->blocksize )
+        {
+           transform_4way( sp );
+           sp->pos = 0;
+        }
+    }
+
+    // pos is zero for 64 byte data, 1 for 80 byte data.
+    sp->h[ sp->pos ] = _mm512_xor_si512( sp->h[ sp->pos ],
+                                    m512_const2_64( 0, 0x0000000000000080 ) );
+    transform_4way( sp );
+
+    sp->h[7] = _mm512_xor_si512( sp->h[7],
+                                    m512_const2_64( 0x0000000100000000, 0 ) );
+
+    for ( i = 0; i < 10; ++i )
+       transform_4way( sp );
+
+    memcpy( hash, sp->h, sp->hashlen<<6);
+    return 0;
+}
+
+
+#endif // AVX512
+
+// 2 way 128 
+
 static void transform_2way( cube_2way_context *sp )
 {
    int r;
    const int rounds = sp->rounds;

-    __m256i x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3;
+    __m256i x0, x1, x2, x3, x4, x5, x6, x7, y0, y1;

    x0 = _mm256_load_si256( (__m256i*)sp->h     );
    x1 = _mm256_load_si256( (__m256i*)sp->h + 1 );
@@ -47,50 +229,38 @@ static void transform_2way( cube_2way_context *sp )
        x5 = _mm256_add_epi32( x1, x5 );
        x6 = _mm256_add_epi32( x2, x6 );
        x7 = _mm256_add_epi32( x3, x7 );
-        y0 = x2;
-        y1 = x3;
-        y2 = x0;
-        y3 = x1;
-        x0 = _mm256_xor_si256( _mm256_slli_epi32( y0,  7 ),
-                               _mm256_srli_epi32( y0, 25 ) );
-        x1 = _mm256_xor_si256( _mm256_slli_epi32( y1,  7 ),
-                               _mm256_srli_epi32( y1, 25 ) );
-        x2 = _mm256_xor_si256( _mm256_slli_epi32( y2,  7 ),
-                               _mm256_srli_epi32( y2, 25 ) );
-        x3 = _mm256_xor_si256( _mm256_slli_epi32( y3,  7 ),
-                               _mm256_srli_epi32( y3, 25 ) );
+        y0 = x0;
+        y1 = x1;
+        x0 = mm256_rol_32( x2, 7 );
+        x1 = mm256_rol_32( x3, 7 );
+        x2 = mm256_rol_32( y0, 7 );
+        x3 = mm256_rol_32( y1, 7 );
        x0 = _mm256_xor_si256( x0, x4 );
        x1 = _mm256_xor_si256( x1, x5 );
        x2 = _mm256_xor_si256( x2, x6 );
        x3 = _mm256_xor_si256( x3, x7 );
-        x4 = mm256_swap64_128( x4 );
-        x5 = mm256_swap64_128( x5 );
-        x6 = mm256_swap64_128( x6 );
-        x7 = mm256_swap64_128( x7 );
+        x4 = mm256_swap128_64( x4 );
+        x5 = mm256_swap128_64( x5 );
+        x6 = mm256_swap128_64( x6 );
+        x7 = mm256_swap128_64( x7 );
        x4 = _mm256_add_epi32( x0, x4 );
        x5 = _mm256_add_epi32( x1, x5 );
        x6 = _mm256_add_epi32( x2, x6 );
        x7 = _mm256_add_epi32( x3, x7 );
-        y0 = x1;
-        y1 = x0;
-        y2 = x3;
-        y3 = x2;
-        x0 = _mm256_xor_si256( _mm256_slli_epi32( y0, 11 ),
-                               _mm256_srli_epi32( y0, 21 ) );
-        x1 = _mm256_xor_si256( _mm256_slli_epi32( y1, 11 ),
-                               _mm256_srli_epi32( y1, 21 ) );
-        x2 = _mm256_xor_si256( _mm256_slli_epi32( y2, 11 ),
-                               _mm256_srli_epi32( y2, 21 ) );
-        x3 = _mm256_xor_si256( _mm256_slli_epi32( y3, 11 ),
-                               _mm256_srli_epi32( y3, 21 ) );
+        y0 = x0;
+        y1 = x2;
+        x0 = mm256_rol_32( x1, 11 );
+        x1 = mm256_rol_32( y0, 11 );
+        x2 = mm256_rol_32( x3, 11 );
+        x3 = mm256_rol_32( y1, 11 );
        x0 = _mm256_xor_si256( x0, x4 );
        x1 = _mm256_xor_si256( x1, x5 );
        x2 = _mm256_xor_si256( x2, x6 );
        x3 = _mm256_xor_si256( x3, x7 );
-        x4 = mm256_swap32_64( x4 );
-        x5 = mm256_swap32_64( x5 );
-        x6 = mm256_swap32_64( x6 );
-        x7 = mm256_swap32_64( x7 );
+        x4 = mm256_swap64_32( x4 );
+        x5 = mm256_swap64_32( x5 );
+        x6 = mm256_swap64_32( x6 );
+        x7 = mm256_swap64_32( x7 );
    }

    _mm256_store_si256( (__m256i*)sp->h,     x0 );
@@ -101,29 +271,36 @@ static void transform_2way( cube_2way_context *sp )
    _mm256_store_si256( (__m256i*)sp->h + 5, x5 );
    _mm256_store_si256( (__m256i*)sp->h + 6, x6 );
    _mm256_store_si256( (__m256i*)sp->h + 7, x7 );
-
 }

 int cube_2way_init( cube_2way_context *sp, int hashbitlen, int rounds,
                    int blockbytes )
 {
-    const uint64_t* iv = hashbitlen == 512 ? IV512 : IV256;
+    __m256i *h = (__m256i*)sp->h;
+    __m128i *iv = (__m128i*)( hashbitlen == 512 ? (__m128i*)IV512
+                                                : (__m128i*)IV256 );
    sp->hashlen   = hashbitlen/128;
    sp->blocksize = blockbytes/16;
    sp->rounds    = rounds;
    sp->pos       = 0;

-    __m256i* h = (__m256i*)sp->h;
-
-    h[0] = _mm256_set_epi64x( iv[ 1], iv[ 0], iv[ 1], iv[ 0] );
-    h[1] = _mm256_set_epi64x( iv[ 3], iv[ 2], iv[ 3], iv[ 2] );
-    h[2] = _mm256_set_epi64x( iv[ 5], iv[ 4], iv[ 5], iv[ 4] );
-    h[3] = _mm256_set_epi64x( iv[ 7], iv[ 6], iv[ 7], iv[ 6] );
-    h[4] = _mm256_set_epi64x( iv[ 9], iv[ 8], iv[ 9], iv[ 8] );
-    h[5] = _mm256_set_epi64x( iv[11], iv[10], iv[11], iv[10] );
-    h[6] = _mm256_set_epi64x( iv[13], iv[12], iv[13], iv[12] );
-    h[7] = _mm256_set_epi64x( iv[15], iv[14], iv[15], iv[14] );
-
+    h[ 0] = m256_const1_128( iv[0] );
+    h[ 1] = m256_const1_128( iv[1] );
+    h[ 2] = m256_const1_128( iv[2] );
+    h[ 3] = m256_const1_128( iv[3] );
+    h[ 4] = m256_const1_128( iv[4] );
+    h[ 5] = m256_const1_128( iv[5] );
+    h[ 6] = m256_const1_128( iv[6] );
+    h[ 7] = m256_const1_128( iv[7] );
+    h[ 0] = m256_const1_128( iv[0] );
+    h[ 1] = m256_const1_128( iv[1] );
+    h[ 2] = m256_const1_128( iv[2] );
+    h[ 3] = m256_const1_128( iv[3] );
+    h[ 4] = m256_const1_128( iv[4] );
+    h[ 5] = m256_const1_128( iv[5] );
+    h[ 6] = m256_const1_128( iv[6] );
+    h[ 7] = m256_const1_128( iv[7] );
+    
    return 0;
 }

@@ -134,9 +311,6 @@ int cube_2way_update( cube_2way_context *sp, const void *data, size_t size )
    const __m256i *in = (__m256i*)data;
    int i;

-    // It is assumed data is aligned to 256 bits and is a multiple of 128 bits.
-    // Current usage sata is either 64 or 80 bytes.
-
    for ( i = 0; i < len; i++ )
    {
        sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ], in[i] );
@@ -157,15 +331,15 @@ int cube_2way_close( cube_2way_context *sp, void *output )

    // pos is zero for 64 byte data, 1 for 80 byte data.
    sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ],
-                                _mm256_set_epi32( 0,0,0,0x80,  0,0,0,0x80 ) );
+                                   m256_const2_64( 0, 0x0000000000000080 ) );
    transform_2way( sp );

    sp->h[7] = _mm256_xor_si256( sp->h[7],
-		                 _mm256_set_epi32( 1,0,0,0,  1,0,0,0 ) );
+                                   m256_const2_64( 0x0000000100000000, 0 ) );

    for ( i = 0; i < 10; ++i )           transform_2way( sp );

-    for ( i = 0; i < sp->hashlen; i++ )  hash[i] = sp->h[i];
+    memcpy( hash, sp->h, sp->hashlen<<5 );
    return 0;
 }

@@ -190,15 +364,15 @@ int cube_2way_update_close( cube_2way_context *sp, void *output,

    // pos is zero for 64 byte data, 1 for 80 byte data.
    sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ],
-                    _mm256_set_epi32( 0,0,0,0x80,  0,0,0,0x80 ) );
+                                    m256_const2_64( 0, 0x0000000000000080 ) );
    transform_2way( sp );

-    sp->h[7] = _mm256_xor_si256( sp->h[7], _mm256_set_epi32( 1,0,0,0,
-                                                             1,0,0,0 ) );
+    sp->h[7] = _mm256_xor_si256( sp->h[7],
+                                    m256_const2_64( 0x0000000100000000, 0 ) );

-    for ( i = 0; i < 10; ++i )            transform_2way( sp );
+    for ( i = 0; i < 10; ++i )    transform_2way( sp );

-    for ( i = 0; i < sp->hashlen; i++ )   hash[i] = sp->h[i];
+    memcpy( hash, sp->h, sp->hashlen<<5 );
    return 0;
 }

--- a/algo/cubehash/cube-hash-2way.h
+++ b/algo/cubehash/cube-hash-2way.h
@@ -1,11 +1,38 @@
 #ifndef CUBE_HASH_2WAY_H__
-#define CUBE_HASH_2WAY_H__
-
-#if defined(__AVX2__)
+#define CUBE_HASH_2WAY_H__ 1

 #include <stdint.h>
 #include "simd-utils.h"

+#if defined(__AVX2__)
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+struct _cube_4way_context
+{
+    __m512i h[8];
+    int hashlen;
+    int rounds;
+    int blocksize;
+    int pos; 
+} __attribute__ ((aligned (128)));
+
+typedef struct _cube_4way_context cube_4way_context;
+
+int cube_4way_init( cube_4way_context* sp, int hashbitlen, int rounds,
+                       int blockbytes );
+// reinitialize context with same parameters, much faster.
+int cube_4way_reinit( cube_4way_context *sp );
+
+int cube_4way_update( cube_4way_context *sp, const void *data, size_t size );
+
+int cube_4way_close( cube_4way_context *sp, void *output );
+
+int cube_4way_update_close( cube_4way_context *sp, void *output,
+                            const void *data, size_t size );
+
+#endif
+
 // 2x128, 2 way parallel SSE2

 struct _cube_2way_context
@@ -15,7 +42,7 @@ struct _cube_2way_context
    int rounds;
    int blocksize;         // __m128i
    int pos;               // number of __m128i read into x from current block
-} __attribute__ ((aligned (64)));
+} __attribute__ ((aligned (128)));

 typedef struct _cube_2way_context cube_2way_context;

--- a/algo/cubehash/cubehash_sse2.c
+++ b/algo/cubehash/cubehash_sse2.c
@@ -16,30 +16,32 @@
 #include "simd-utils.h"
 #include <stdio.h>

-// The result of hashing 10 rounds of initial data which is params and 
-// mostly zeros.
-static const uint64_t IV256[] =
-{
-0xCCD6F29FEA2BD4B4, 0x35481EAE63117E71, 0xE5D94E6322512D5B, 0xF4CC12BE7E624131,
-0x42AF2070C2D0B696, 0x3361DA8CD0720C35, 0x8EF8AD8328CCECA4, 0x40E5FBAB4680AC00,
-0x6107FBD5D89041C3, 0xF0B266796C859D41, 0x5FA2560309392549, 0x93CB628565C892FD,
-0x9E4B4E602AF2B5AE, 0x85254725774ABFDD, 0x4AB6AAD615815AEB, 0xD6032C0A9CDAF8AF
-};
-
-static const uint64_t IV512[] =
-{
-0x50F494D42AEA2A61, 0x4167D83E2D538B8B, 0xC701CF8C3FEE2313, 0x50AC5695CC39968E,
-0xA647A8B34D42C787, 0x825B453797CF0BEF, 0xF22090C4EEF864D2, 0xA23911AED0E5CD33,
-0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
-0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
-};
-
 static void transform( cubehashParam *sp )
 {
    int r;
    const int rounds = sp->rounds;

-#ifdef __AVX2__
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+    register __m512i x0, x1;
+
+    x0 = _mm512_load_si512( (__m512i*)sp->x     );
+    x1 = _mm512_load_si512( (__m512i*)sp->x + 1 );
+
+    for ( r = 0; r < rounds; ++r )
+    { 
+        x1 = _mm512_add_epi32( x0, x1 );
+        x0 = _mm512_xor_si512( mm512_rol_32( mm512_swap_256( x0 ), 7 ), x1 );
+        x1 = _mm512_add_epi32( x0, mm512_swap128_64( x1 ) );
+        x0 = _mm512_xor_si512( mm512_rol_32(
+                                         mm512_swap256_128( x0 ), 11 ), x1 );
+        x1 = mm512_swap64_32( x1 );
+    }
+
+    _mm512_store_si512( (__m512i*)sp->x,     x0 );
+    _mm512_store_si512( (__m512i*)sp->x + 1, x1 );
+
+#elif defined(__AVX2__)

    register __m256i x0, x1, x2, x3, y0, y1;

@@ -53,26 +55,22 @@ static void transform( cubehashParam *sp )
        x2 = _mm256_add_epi32( x0, x2 );
        x3 = _mm256_add_epi32( x1, x3 );
        y0 = x0;
-        x0 = _mm256_xor_si256( _mm256_slli_epi32( x1, 7 ),
-                               _mm256_srli_epi32( x1, 25 ) );
-        x1 = _mm256_xor_si256( _mm256_slli_epi32( y0, 7 ),
-                               _mm256_srli_epi32( y0, 25 ) );
+        x0 = mm256_rol_32( x1, 7 );
+        x1 = mm256_rol_32( y0, 7 );
        x0 = _mm256_xor_si256( x0, x2 );
        x1 = _mm256_xor_si256( x1, x3 );
-        x2 = _mm256_shuffle_epi32( x2, 0x4e );
-        x3 = _mm256_shuffle_epi32( x3, 0x4e );
+        x2 = mm256_swap128_64( x2 );
+        x3 = mm256_swap128_64( x3 );
        x2 = _mm256_add_epi32( x0, x2 );
        x3 = _mm256_add_epi32( x1, x3 );
-        y0 = _mm256_permute4x64_epi64( x0, 0x4e );
-        y1 = _mm256_permute4x64_epi64( x1, 0x4e );
-        x0 = _mm256_xor_si256( _mm256_slli_epi32( y0, 11 ),
-                               _mm256_srli_epi32( y0, 21 ) );
-        x1 = _mm256_xor_si256( _mm256_slli_epi32( y1, 11 ), 
-                               _mm256_srli_epi32( y1, 21 ) );
+        y0 = mm256_swap_128( x0 );
+        y1 = mm256_swap_128( x1 );
+        x0 = mm256_rol_32( y0, 11 );
+        x1 = mm256_rol_32( y1, 11 );
        x0 = _mm256_xor_si256( x0, x2 );
        x1 = _mm256_xor_si256( x1, x3 );
-        x2 = _mm256_shuffle_epi32( x2, 0xb1 );
-        x3 = _mm256_shuffle_epi32( x3, 0xb1 );
+        x2 = mm256_swap64_32( x2 );
+        x3 = mm256_swap64_32( x3 );
    }

    _mm256_store_si256( (__m256i*)sp->x,     x0 );
@@ -147,37 +145,58 @@ static void transform( cubehashParam *sp )
 #endif
 }  // transform

+/*
+// The result of hashing 10 rounds of initial data which is params and
+// mostly zeros.
+static const uint64_t IV256[] =
+{
+0xCCD6F29FEA2BD4B4, 0x35481EAE63117E71, 0xE5D94E6322512D5B, 0xF4CC12BE7E624131,
+0x42AF2070C2D0B696, 0x3361DA8CD0720C35, 0x8EF8AD8328CCECA4, 0x40E5FBAB4680AC00,
+0x6107FBD5D89041C3, 0xF0B266796C859D41, 0x5FA2560309392549, 0x93CB628565C892FD,
+0x9E4B4E602AF2B5AE, 0x85254725774ABFDD, 0x4AB6AAD615815AEB, 0xD6032C0A9CDAF8AF
+};
+
+static const uint64_t IV512[] =
+{
+0x50F494D42AEA2A61, 0x4167D83E2D538B8B, 0xC701CF8C3FEE2313, 0x50AC5695CC39968E,
+0xA647A8B34D42C787, 0x825B453797CF0BEF, 0xF22090C4EEF864D2, 0xA23911AED0E5CD33,
+0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
+0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
+};
+*/
+
 int cubehashInit(cubehashParam *sp, int hashbitlen, int rounds, int blockbytes)
 {
-    const uint64_t* iv = hashbitlen == 512 ? IV512 : IV256;
+    __m128i *x = (__m128i*)sp->x;
    sp->hashlen   = hashbitlen/128;
    sp->blocksize = blockbytes/16;
    sp->rounds    = rounds;
    sp->pos       = 0;
-    
-#if defined(__AVX2__)

-    __m256i* x = (__m256i*)sp->x;
+    if ( hashbitlen == 512 )
+    {

-    x[0] = _mm256_set_epi64x( iv[ 3], iv[ 2], iv[ 1], iv[ 0] );
-    x[1] = _mm256_set_epi64x( iv[ 7], iv[ 6], iv[ 5], iv[ 4] );
-    x[2] = _mm256_set_epi64x( iv[11], iv[10], iv[ 9], iv[ 8] );
-    x[3] = _mm256_set_epi64x( iv[15], iv[14], iv[13], iv[12] );
+       x[0] = m128_const_64( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
+       x[1] = m128_const_64( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
+       x[2] = m128_const_64( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
+       x[3] = m128_const_64( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
+       x[4] = m128_const_64( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
+       x[5] = m128_const_64( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
+       x[6] = m128_const_64( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
+       x[7] = m128_const_64( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
+    }
+    else
+    {
+       x[0] = m128_const_64( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
+       x[1] = m128_const_64( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
+       x[2] = m128_const_64( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
+       x[3] = m128_const_64( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
+       x[4] = m128_const_64( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
+       x[5] = m128_const_64( 0x93CB628565C892FD, 0x5FA2560309392549 );
+       x[6] = m128_const_64( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
+       x[7] = m128_const_64( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
+    }   

-#else
-
-    __m128i* x = (__m128i*)sp->x;
-
-     x[0] = _mm_set_epi64x( iv[ 1], iv[ 0] );
-     x[1] = _mm_set_epi64x( iv[ 3], iv[ 2] );
-     x[2] = _mm_set_epi64x( iv[ 5], iv[ 4] );
-     x[3] = _mm_set_epi64x( iv[ 7], iv[ 6] );
-     x[4] = _mm_set_epi64x( iv[ 9], iv[ 8] );
-     x[5] = _mm_set_epi64x( iv[11], iv[10] );
-     x[6] = _mm_set_epi64x( iv[13], iv[12] );
-     x[7] = _mm_set_epi64x( iv[15], iv[14] );
-
-#endif
    return SUCCESS;
 }

--- a/algo/echo/aes_ni/hash.c
+++ b/algo/echo/aes_ni/hash.c
@@ -179,53 +179,53 @@ void Compress(hashState_echo *ctx, const unsigned char *pmsg, unsigned int uBloc

   for(b = 0; b < uBlockCount; b++)
   {
-	ctx->k = _mm_add_epi64(ctx->k, ctx->const1536);
+   	ctx->k = _mm_add_epi64(ctx->k, ctx->const1536);

-	// load message
-	for(j = ctx->uHashSize / 256; j < 4; j++)
-	{
-	   for(i = 0; i < 4; i++)
+   	// load message
+	   for(j = ctx->uHashSize / 256; j < 4; j++)
 	   {
-		_state[i][j] = _mm_loadu_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
+	      for(i = 0; i < 4; i++)
+	      {
+		     _state[i][j] = _mm_load_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
+	      }
 	   }
-	}

-	// save state
-	SAVESTATE(_statebackup, _state);
+	   // save state
+	   SAVESTATE(_statebackup, _state);

-	k1 = ctx->k;
+	   k1 = ctx->k;

-	for(r = 0; r < ctx->uRounds / 2; r++)
-	{
-		ECHO_ROUND_UNROLL2;
-	}
+	   for(r = 0; r < ctx->uRounds / 2; r++)
+   	{
+	   	ECHO_ROUND_UNROLL2;
+	   }
 		
-	if(ctx->uHashSize == 256)
-	{
-	   for(i = 0; i < 4; i++)
+	   if(ctx->uHashSize == 256)
 	   {
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][1]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][3]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][1]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][3]);
+	      for(i = 0; i < 4; i++)
+	      {
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _state[i][1]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _state[i][3]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][1]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][3]);
+	      }
 	   }
-	}
-	else
-	{
-	   for(i = 0; i < 4; i++)
-	   {
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
-		_state[i][1] = _mm_xor_si128(_state[i][1], _state[i][3]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
-		_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][1]);
-		_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][3]);
-           }
-	}
-	pmsg += ctx->uBlockLength;
+	   else
+    	{
+	      for(i = 0; i < 4; i++)
+	      {
+      		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
+		      _state[i][1] = _mm_xor_si128(_state[i][1], _state[i][3]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
+		      _state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
+		      _state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][1]);
+		      _state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][3]);
+         }
+   	}
+	   pmsg += ctx->uBlockLength;
   }
 	SAVESTATE(ctx->state, _state);

@@ -390,13 +390,13 @@ HashReturn final_echo(hashState_echo *state, BitSequence *hashval)
 	}

 	// Store the hash value
-	_mm_storeu_si128((__m128i*)hashval + 0, state->state[0][0]);
-	_mm_storeu_si128((__m128i*)hashval + 1, state->state[1][0]);
+	_mm_store_si128((__m128i*)hashval + 0, state->state[0][0]);
+	_mm_store_si128((__m128i*)hashval + 1, state->state[1][0]);

 	if(state->uHashSize == 512)
 	{
-		_mm_storeu_si128((__m128i*)hashval + 2, state->state[2][0]);
-		_mm_storeu_si128((__m128i*)hashval + 3, state->state[3][0]);
+		_mm_store_si128((__m128i*)hashval + 2, state->state[2][0]);
+		_mm_store_si128((__m128i*)hashval + 3, state->state[3][0]);
 	}

 	return SUCCESS;
@@ -513,13 +513,13 @@ HashReturn update_final_echo( hashState_echo *state, BitSequence *hashval,
   }

   // Store the hash value
-   _mm_storeu_si128( (__m128i*)hashval + 0, state->state[0][0] );
-   _mm_storeu_si128( (__m128i*)hashval + 1, state->state[1][0] );
+   _mm_store_si128( (__m128i*)hashval + 0, state->state[0][0] );
+   _mm_store_si128( (__m128i*)hashval + 1, state->state[1][0] );

   if( state->uHashSize == 512 )
   {
-        _mm_storeu_si128( (__m128i*)hashval + 2, state->state[2][0] );
-        _mm_storeu_si128( (__m128i*)hashval + 3, state->state[3][0] );
+        _mm_store_si128( (__m128i*)hashval + 2, state->state[2][0] );
+        _mm_store_si128( (__m128i*)hashval + 3, state->state[3][0] );

   }
   return SUCCESS;
--- a/algo/echo/aes_ni/hash.c.test
+++ b/algo/echo/aes_ni/hash.c.test
@@ -0,0 +1,620 @@
+/*
+ * file        : echo_vperm.c
+ * version     : 1.0.208
+ * date        : 14.12.2010
+ * 
+ * - vperm and aes_ni implementations of hash function ECHO
+ * - implements NIST hash api
+ * - assumes that message lenght is multiple of 8-bits
+ * - _ECHO_VPERM_ must be defined if compiling with ../main.c
+ * -  define NO_AES_NI for aes_ni version
+ *
+ * Cagdas Calik
+ * ccalik@metu.edu.tr
+ * Institute of Applied Mathematics, Middle East Technical University, Turkey.
+ *
+ */
+#if defined(__AES__)
+
+#include <memory.h>
+#include "miner.h"
+#include "hash_api.h"
+//#include "vperm.h"
+#include <immintrin.h>
+/*
+#ifndef NO_AES_NI
+#include <wmmintrin.h>
+#else
+#include <tmmintrin.h>
+#endif
+*/
+
+MYALIGN const unsigned int _k_s0F[] = {0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F};
+MYALIGN const unsigned int _k_ipt[] = {0x5A2A7000, 0xC2B2E898, 0x52227808, 0xCABAE090, 0x317C4D00, 0x4C01307D, 0xB0FDCC81, 0xCD80B1FC};
+MYALIGN const unsigned int _k_opt[] = {0xD6B66000, 0xFF9F4929, 0xDEBE6808, 0xF7974121, 0x50BCEC00, 0x01EDBD51, 0xB05C0CE0, 0xE10D5DB1};
+MYALIGN const unsigned int _k_inv[] = {0x0D080180, 0x0E05060F, 0x0A0B0C02, 0x04070309, 0x0F0B0780, 0x01040A06, 0x02050809, 0x030D0E0C};
+MYALIGN const unsigned int _k_sb1[] = {0xCB503E00, 0xB19BE18F, 0x142AF544, 0xA5DF7A6E, 0xFAE22300, 0x3618D415, 0x0D2ED9EF, 0x3BF7CCC1};
+MYALIGN const unsigned int _k_sb2[] = {0x0B712400, 0xE27A93C6, 0xBC982FCD, 0x5EB7E955, 0x0AE12900, 0x69EB8840, 0xAB82234A, 0xC2A163C8};
+MYALIGN const unsigned int _k_sb3[] = {0xC0211A00, 0x53E17249, 0xA8B2DA89, 0xFB68933B, 0xF0030A00, 0x5FF35C55, 0xA6ACFAA5, 0xF956AF09};
+MYALIGN const unsigned int _k_sb4[] = {0x3FD64100, 0xE1E937A0, 0x49087E9F, 0xA876DE97, 0xC393EA00, 0x3D50AED7, 0x876D2914, 0xBA44FE79};
+MYALIGN const unsigned int _k_sb5[] = {0xF4867F00, 0x5072D62F, 0x5D228BDB, 0x0DA9A4F9, 0x3971C900, 0x0B487AC2, 0x8A43F0FB, 0x81B332B8};
+MYALIGN const unsigned int _k_sb7[] = {0xFFF75B00, 0xB20845E9, 0xE1BAA416, 0x531E4DAC, 0x3390E000, 0x62A3F282, 0x21C1D3B1, 0x43125170};
+MYALIGN const unsigned int _k_sbo[] = {0x6FBDC700, 0xD0D26D17, 0xC502A878, 0x15AABF7A, 0x5FBB6A00, 0xCFE474A5, 0x412B35FA, 0x8E1E90D1};
+MYALIGN const unsigned int _k_h63[] = {0x63636363, 0x63636363, 0x63636363, 0x63636363};
+MYALIGN const unsigned int _k_hc6[] = {0xc6c6c6c6, 0xc6c6c6c6, 0xc6c6c6c6, 0xc6c6c6c6};
+MYALIGN const unsigned int _k_h5b[] = {0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b};
+MYALIGN const unsigned int _k_h4e[] = {0x4e4e4e4e, 0x4e4e4e4e, 0x4e4e4e4e, 0x4e4e4e4e};
+MYALIGN const unsigned int _k_h0e[] = {0x0e0e0e0e, 0x0e0e0e0e, 0x0e0e0e0e, 0x0e0e0e0e};
+MYALIGN const unsigned int _k_h15[] = {0x15151515, 0x15151515, 0x15151515, 0x15151515};
+MYALIGN const unsigned int _k_aesmix1[] = {0x0f0a0500, 0x030e0904, 0x07020d08, 0x0b06010c};
+MYALIGN const unsigned int _k_aesmix2[] = {0x000f0a05, 0x04030e09, 0x0807020d, 0x0c0b0601};
+MYALIGN const unsigned int _k_aesmix3[] = {0x05000f0a, 0x0904030e, 0x0d080702, 0x010c0b06};
+MYALIGN const unsigned int _k_aesmix4[] = {0x0a05000f, 0x0e090403, 0x020d0807, 0x06010c0b};
+
+
+MYALIGN const unsigned int 	const1[]		= {0x00000001, 0x00000000, 0x00000000, 0x00000000};
+MYALIGN const unsigned int	mul2mask[]		= {0x00001b00, 0x00000000, 0x00000000, 0x00000000};
+MYALIGN const unsigned int	lsbmask[]		= {0x01010101, 0x01010101, 0x01010101, 0x01010101};
+MYALIGN const unsigned int	invshiftrows[]	= {0x070a0d00, 0x0b0e0104, 0x0f020508, 0x0306090c};
+MYALIGN const unsigned int	zero[]			= {0x00000000, 0x00000000, 0x00000000, 0x00000000};
+MYALIGN const unsigned int	mul2ipt[]		= {0x728efc00, 0x6894e61a, 0x3fc3b14d, 0x25d9ab57, 0xfd5ba600, 0x2a8c71d7, 0x1eb845e3, 0xc96f9234};
+
+
+#define ECHO_SUBBYTES(state, i, j) \
+	state[i][j] = _mm_aesenc_si128(state[i][j], k1);\
+	state[i][j] = _mm_aesenc_si128(state[i][j], M128(zero));\
+	k1 = _mm_add_epi32(k1, M128(const1))
+
+#define ECHO_MIXBYTES(state1, state2, j, t1, t2, s2) \
+	s2 = _mm_add_epi8(state1[0][j], state1[0][j]);\
+	t1 = _mm_srli_epi16(state1[0][j], 7);\
+	t1 = _mm_and_si128(t1, M128(lsbmask));\
+	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
+	s2 = _mm_xor_si128(s2, t2);\
+	state2[0][j] = s2;\
+	state2[1][j] = state1[0][j];\
+	state2[2][j] = state1[0][j];\
+	state2[3][j] = _mm_xor_si128(s2, state1[0][j]);\
+	s2 = _mm_add_epi8(state1[1][(j + 1) & 3], state1[1][(j + 1) & 3]);\
+	t1 = _mm_srli_epi16(state1[1][(j + 1) & 3], 7);\
+	t1 = _mm_and_si128(t1, M128(lsbmask));\
+	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
+	s2 = _mm_xor_si128(s2, t2);\
+	state2[0][j] = _mm_xor_si128(state2[0][j], _mm_xor_si128(s2, state1[1][(j + 1) & 3]));\
+	state2[1][j] = _mm_xor_si128(state2[1][j], s2);\
+	state2[2][j] = _mm_xor_si128(state2[2][j], state1[1][(j + 1) & 3]);\
+	state2[3][j] = _mm_xor_si128(state2[3][j], state1[1][(j + 1) & 3]);\
+	s2 = _mm_add_epi8(state1[2][(j + 2) & 3], state1[2][(j + 2) & 3]);\
+	t1 = _mm_srli_epi16(state1[2][(j + 2) & 3], 7);\
+	t1 = _mm_and_si128(t1, M128(lsbmask));\
+	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
+	s2 = _mm_xor_si128(s2, t2);\
+	state2[0][j] = _mm_xor_si128(state2[0][j], state1[2][(j + 2) & 3]);\
+	state2[1][j] = _mm_xor_si128(state2[1][j], _mm_xor_si128(s2, state1[2][(j + 2) & 3]));\
+	state2[2][j] = _mm_xor_si128(state2[2][j], s2);\
+	state2[3][j] = _mm_xor_si128(state2[3][j], state1[2][(j + 2) & 3]);\
+	s2 = _mm_add_epi8(state1[3][(j + 3) & 3], state1[3][(j + 3) & 3]);\
+	t1 = _mm_srli_epi16(state1[3][(j + 3) & 3], 7);\
+	t1 = _mm_and_si128(t1, M128(lsbmask));\
+	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
+	s2 = _mm_xor_si128(s2, t2);\
+	state2[0][j] = _mm_xor_si128(state2[0][j], state1[3][(j + 3) & 3]);\
+	state2[1][j] = _mm_xor_si128(state2[1][j], state1[3][(j + 3) & 3]);\
+	state2[2][j] = _mm_xor_si128(state2[2][j], _mm_xor_si128(s2, state1[3][(j + 3) & 3]));\
+	state2[3][j] = _mm_xor_si128(state2[3][j], s2)
+
+
+#define ECHO_ROUND_UNROLL2 \
+	ECHO_SUBBYTES(_state, 0, 0);\
+	ECHO_SUBBYTES(_state, 1, 0);\
+	ECHO_SUBBYTES(_state, 2, 0);\
+	ECHO_SUBBYTES(_state, 3, 0);\
+	ECHO_SUBBYTES(_state, 0, 1);\
+	ECHO_SUBBYTES(_state, 1, 1);\
+	ECHO_SUBBYTES(_state, 2, 1);\
+	ECHO_SUBBYTES(_state, 3, 1);\
+	ECHO_SUBBYTES(_state, 0, 2);\
+	ECHO_SUBBYTES(_state, 1, 2);\
+	ECHO_SUBBYTES(_state, 2, 2);\
+	ECHO_SUBBYTES(_state, 3, 2);\
+	ECHO_SUBBYTES(_state, 0, 3);\
+	ECHO_SUBBYTES(_state, 1, 3);\
+	ECHO_SUBBYTES(_state, 2, 3);\
+	ECHO_SUBBYTES(_state, 3, 3);\
+	ECHO_MIXBYTES(_state, _state2, 0, t1, t2, s2);\
+	ECHO_MIXBYTES(_state, _state2, 1, t1, t2, s2);\
+	ECHO_MIXBYTES(_state, _state2, 2, t1, t2, s2);\
+	ECHO_MIXBYTES(_state, _state2, 3, t1, t2, s2);\
+	ECHO_SUBBYTES(_state2, 0, 0);\
+	ECHO_SUBBYTES(_state2, 1, 0);\
+	ECHO_SUBBYTES(_state2, 2, 0);\
+	ECHO_SUBBYTES(_state2, 3, 0);\
+	ECHO_SUBBYTES(_state2, 0, 1);\
+	ECHO_SUBBYTES(_state2, 1, 1);\
+	ECHO_SUBBYTES(_state2, 2, 1);\
+	ECHO_SUBBYTES(_state2, 3, 1);\
+	ECHO_SUBBYTES(_state2, 0, 2);\
+	ECHO_SUBBYTES(_state2, 1, 2);\
+	ECHO_SUBBYTES(_state2, 2, 2);\
+	ECHO_SUBBYTES(_state2, 3, 2);\
+	ECHO_SUBBYTES(_state2, 0, 3);\
+	ECHO_SUBBYTES(_state2, 1, 3);\
+	ECHO_SUBBYTES(_state2, 2, 3);\
+	ECHO_SUBBYTES(_state2, 3, 3);\
+	ECHO_MIXBYTES(_state2, _state, 0, t1, t2, s2);\
+	ECHO_MIXBYTES(_state2, _state, 1, t1, t2, s2);\
+	ECHO_MIXBYTES(_state2, _state, 2, t1, t2, s2);\
+	ECHO_MIXBYTES(_state2, _state, 3, t1, t2, s2)
+
+
+
+#define SAVESTATE(dst, src)\
+	dst[0][0] = src[0][0];\
+	dst[0][1] = src[0][1];\
+	dst[0][2] = src[0][2];\
+	dst[0][3] = src[0][3];\
+	dst[1][0] = src[1][0];\
+	dst[1][1] = src[1][1];\
+	dst[1][2] = src[1][2];\
+	dst[1][3] = src[1][3];\
+	dst[2][0] = src[2][0];\
+	dst[2][1] = src[2][1];\
+	dst[2][2] = src[2][2];\
+	dst[2][3] = src[2][3];\
+	dst[3][0] = src[3][0];\
+	dst[3][1] = src[3][1];\
+	dst[3][2] = src[3][2];\
+	dst[3][3] = src[3][3]
+
+
+void Compress(hashState_echo *ctx, const unsigned char *pmsg, unsigned int uBlockCount)
+{
+   unsigned int r, b, i, j;
+   __m128i t1, t2, s2, k1;
+   __m128i _state[4][4], _state2[4][4], _statebackup[4][4]; 
+
+   for(i = 0; i < 4; i++)
+	for(j = 0; j < ctx->uHashSize / 256; j++)
+		_state[i][j] = ctx->state[i][j];
+
+   for(b = 0; b < uBlockCount; b++)
+   {
+	ctx->k = _mm_add_epi64(ctx->k, ctx->const1536);
+
+	// load message
+	for(j = ctx->uHashSize / 256; j < 4; j++)
+	{
+	   for(i = 0; i < 4; i++)
+	   {
+		_state[i][j] = _mm_loadu_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
+	   }
+	}
+
+uint64_t *b = (uint64_t*)_state;
+//printf("Ss3: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
+   
+	// save state
+	SAVESTATE(_statebackup, _state);
+
+	k1 = ctx->k;
+
+	for(r = 0; r < ctx->uRounds / 2; r++)
+	{
+		ECHO_ROUND_UNROLL2;
+	}
+
+//printf("Ss4: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
+   
+   
+	if(ctx->uHashSize == 256)
+	{
+	   for(i = 0; i < 4; i++)
+	   {
+		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][1]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][3]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][1]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][3]);
+	   }
+	}
+	else
+	{
+	   for(i = 0; i < 4; i++)
+	   {
+		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
+		_state[i][1] = _mm_xor_si128(_state[i][1], _state[i][3]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
+		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
+		_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][1]);
+		_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][3]);
+           }
+	}
+	pmsg += ctx->uBlockLength;
+   }
+	SAVESTATE(ctx->state, _state);
+
+}
+
+
+
+HashReturn init_echo(hashState_echo *ctx, int nHashSize)
+{
+	int i, j;
+
+        ctx->k = _mm_setzero_si128(); 
+	ctx->processed_bits = 0;
+	ctx->uBufferBytes = 0;
+
+	switch(nHashSize)
+	{
+		case 256:
+			ctx->uHashSize = 256;
+			ctx->uBlockLength = 192;
+			ctx->uRounds = 8;
+			ctx->hashsize = _mm_set_epi32(0, 0, 0, 0x00000100);
+			ctx->const1536 = _mm_set_epi32(0x00000000, 0x00000000, 0x00000000, 0x00000600);
+			break;
+
+		case 512:
+			ctx->uHashSize = 512;
+			ctx->uBlockLength = 128;
+			ctx->uRounds = 10;
+			ctx->hashsize = _mm_set_epi32(0, 0, 0, 0x00000200);
+			ctx->const1536 = _mm_set_epi32(0x00000000, 0x00000000, 0x00000000, 0x00000400);
+			break;
+
+		default:
+			return BAD_HASHBITLEN;
+	}
+
+
+	for(i = 0; i < 4; i++)
+		for(j = 0; j < nHashSize / 256; j++)
+			ctx->state[i][j] = ctx->hashsize;
+
+	for(i = 0; i < 4; i++)
+		for(j = nHashSize / 256; j < 4; j++)
+			ctx->state[i][j] = _mm_set_epi32(0, 0, 0, 0);
+
+	return SUCCESS;
+}
+
+HashReturn update_echo(hashState_echo *state, const BitSequence *data, DataLength databitlen)
+{
+	unsigned int uByteLength, uBlockCount, uRemainingBytes;
+
+	uByteLength = (unsigned int)(databitlen / 8);
+
+	if((state->uBufferBytes + uByteLength) >= state->uBlockLength)
+	{
+		if(state->uBufferBytes != 0)
+		{
+			// Fill the buffer
+			memcpy(state->buffer + state->uBufferBytes, (void*)data, state->uBlockLength - state->uBufferBytes);
+
+			// Process buffer
+			Compress(state, state->buffer, 1);
+			state->processed_bits += state->uBlockLength * 8;
+
+			data += state->uBlockLength - state->uBufferBytes;
+			uByteLength -= state->uBlockLength - state->uBufferBytes;
+		}
+
+		// buffer now does not contain any unprocessed bytes
+
+		uBlockCount = uByteLength / state->uBlockLength;
+		uRemainingBytes = uByteLength % state->uBlockLength;
+
+		if(uBlockCount > 0)
+		{
+			Compress(state, data, uBlockCount);
+
+			state->processed_bits += uBlockCount * state->uBlockLength * 8;
+			data += uBlockCount * state->uBlockLength;
+		}
+
+		if(uRemainingBytes > 0)
+		{
+			memcpy(state->buffer, (void*)data, uRemainingBytes);
+		}
+
+		state->uBufferBytes = uRemainingBytes;
+	}
+	else
+	{
+		memcpy(state->buffer + state->uBufferBytes, (void*)data, uByteLength);
+		state->uBufferBytes += uByteLength;
+	}
+
+	return SUCCESS;
+}
+
+HashReturn final_echo(hashState_echo *state, BitSequence *hashval)
+{
+	__m128i remainingbits;
+
+	// Add remaining bytes in the buffer
+	state->processed_bits += state->uBufferBytes * 8;
+
+	remainingbits = _mm_set_epi32(0, 0, 0, state->uBufferBytes * 8);
+
+	// Pad with 0x80
+	state->buffer[state->uBufferBytes++] = 0x80;
+	
+	// Enough buffer space for padding in this block?
+	if((state->uBlockLength - state->uBufferBytes) >= 18)
+	{
+		// Pad with zeros
+		memset(state->buffer + state->uBufferBytes, 0, state->uBlockLength - (state->uBufferBytes + 18));
+
+		// Hash size
+		*((unsigned short*)(state->buffer + state->uBlockLength - 18)) = state->uHashSize;
+
+		// Processed bits
+		*((DataLength*)(state->buffer + state->uBlockLength - 16)) = state->processed_bits;
+		*((DataLength*)(state->buffer + state->uBlockLength - 8)) = 0;
+
+		// Last block contains message bits?
+		if(state->uBufferBytes == 1)
+		{
+			state->k = _mm_xor_si128(state->k, state->k);
+			state->k = _mm_sub_epi64(state->k, state->const1536);
+		}
+		else
+		{
+			state->k = _mm_add_epi64(state->k, remainingbits);
+			state->k = _mm_sub_epi64(state->k, state->const1536);
+		}
+
+		// Compress
+		Compress(state, state->buffer, 1);
+	}
+	else
+	{
+		// Fill with zero and compress
+		memset(state->buffer + state->uBufferBytes, 0, state->uBlockLength - state->uBufferBytes);
+		state->k = _mm_add_epi64(state->k, remainingbits);
+		state->k = _mm_sub_epi64(state->k, state->const1536);
+		Compress(state, state->buffer, 1);
+
+		// Last block
+		memset(state->buffer, 0, state->uBlockLength - 18);
+
+		// Hash size
+		*((unsigned short*)(state->buffer + state->uBlockLength - 18)) = state->uHashSize;
+
+		// Processed bits
+		*((DataLength*)(state->buffer + state->uBlockLength - 16)) = state->processed_bits;
+		*((DataLength*)(state->buffer + state->uBlockLength - 8)) = 0;
+
+		// Compress the last block
+		state->k = _mm_xor_si128(state->k, state->k);
+		state->k = _mm_sub_epi64(state->k, state->const1536);
+		Compress(state, state->buffer, 1);
+	}
+
+	// Store the hash value
+	_mm_storeu_si128((__m128i*)hashval + 0, state->state[0][0]);
+	_mm_storeu_si128((__m128i*)hashval + 1, state->state[1][0]);
+
+	if(state->uHashSize == 512)
+	{
+		_mm_storeu_si128((__m128i*)hashval + 2, state->state[2][0]);
+		_mm_storeu_si128((__m128i*)hashval + 3, state->state[3][0]);
+	}
+
+	return SUCCESS;
+}
+
+HashReturn update_final_echo( hashState_echo *state, BitSequence *hashval,
+                              const BitSequence *data, DataLength databitlen )
+{
+   unsigned int uByteLength, uBlockCount, uRemainingBytes;
+
+   uByteLength = (unsigned int)(databitlen / 8);
+
+/*   
+   if( (state->uBufferBytes + uByteLength) >= state->uBlockLength )
+   {
+printf("full block\n");
+      if( state->uBufferBytes != 0 )
+        {
+           // Fill the buffer
+           memcpy( state->buffer + state->uBufferBytes,
+                   (void*)data, state->uBlockLength - state->uBufferBytes );
+
+           // Process buffer
+           Compress( state, state->buffer, 1 );
+           state->processed_bits += state->uBlockLength * 8;
+
+           data += state->uBlockLength - state->uBufferBytes;
+           uByteLength -= state->uBlockLength - state->uBufferBytes;
+        }
+
+        // buffer now does not contain any unprocessed bytes
+
+        uBlockCount = uByteLength / state->uBlockLength;
+        uRemainingBytes = uByteLength % state->uBlockLength;
+
+        if( uBlockCount > 0 )
+        {
+           Compress( state, data, uBlockCount );
+           state->processed_bits += uBlockCount * state->uBlockLength * 8;
+           data += uBlockCount * state->uBlockLength;
+        }
+
+        if( uRemainingBytes > 0 )
+        memcpy(state->buffer, (void*)data, uRemainingBytes);
+
+        state->uBufferBytes = uRemainingBytes;
+   }
+   else
+   {
+*/
+   memcpy( state->buffer + state->uBufferBytes, (void*)data, uByteLength );
+        state->uBufferBytes += uByteLength;
+//   }
+
+   __m128i remainingbits;
+
+   // Add remaining bytes in the buffer
+   state->processed_bits += state->uBufferBytes * 8;
+
+   remainingbits = _mm_set_epi32( 0, 0, 0, state->uBufferBytes * 8 );
+
+   // Pad with 0x80
+   state->buffer[state->uBufferBytes++] = 0x80;
+
+   // Enough buffer space for padding in this block?
+
+//   if( (state->uBlockLength - state->uBufferBytes) >= 18 )
+//   {
+        // Pad with zeros
+
+        memset( state->buffer + state->uBufferBytes, 0, state->uBlockLength - (state->uBufferBytes + 18) );
+
+        // Hash size
+        *( (unsigned short*)(state->buffer + state->uBlockLength - 18) ) = state->uHashSize;
+
+        // Processed bits
+        *( (DataLength*)(state->buffer + state->uBlockLength - 16) ) =
+                   state->processed_bits;
+        *( (DataLength*)(state->buffer + state->uBlockLength - 8) ) = 0;
+
+
+        // Last block contains message bits?
+        if( state->uBufferBytes == 1 )
+        {
+           state->k = _mm_xor_si128( state->k, state->k );
+           state->k = _mm_sub_epi64( state->k, state->const1536 );
+        }
+        else
+        {
+           state->k = _mm_add_epi64( state->k, remainingbits );
+           state->k = _mm_sub_epi64( state->k, state->const1536 );
+        }
+
+uint64_t *b = (uint64_t*)&state->k;
+/*
+printf("Sk: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
+b = (uint64_t*)state->buffer;
+printf("Sb: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
+printf("Sb: %016lx %016lx %016lx %016lx\n",b[4],b[5],b[6],b[7]);
+printf("Sb: %016lx %016lx %016lx %016lx\n",b[8],b[9],b[10],b[11]);
+printf("Sb: %016lx %016lx %016lx %016lx\n",b[12],b[13],b[14],b[15]);
+
+b = (uint64_t*)state->state;
+printf("Ss1: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
+printf("Ss1: %016lx %016lx %016lx %016lx\n",b[4],b[5],b[6],b[7]);
+printf("Ss1: %016lx %016lx %016lx %016lx\n",b[8],b[9],b[10],b[11]);
+printf("Ss1: %016lx %016lx %016lx %016lx\n",b[12],b[13],b[14],b[15]);
+*/        
+        // Compress
+        Compress( state, state->buffer, 1 );
+
+//printf("Ss2: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
+
+        
+/*
+   }
+   else
+   {
+        // Fill with zero and compress
+        memset( state->buffer + state->uBufferBytes, 0,
+                state->uBlockLength - state->uBufferBytes );
+        state->k = _mm_add_epi64( state->k, remainingbits );
+        state->k = _mm_sub_epi64( state->k, state->const1536 );
+        Compress( state, state->buffer, 1 );
+
+        // Last block
+        memset( state->buffer, 0, state->uBlockLength - 18 );
+
+        // Hash size
+        *( (unsigned short*)(state->buffer + state->uBlockLength - 18) ) =
+                 state->uHashSize;
+
+        // Processed bits
+        *( (DataLength*)(state->buffer + state->uBlockLength - 16) ) =
+                   state->processed_bits;
+        *( (DataLength*)(state->buffer + state->uBlockLength - 8) ) = 0;
+        // Compress the last block
+        state->k = _mm_xor_si128( state->k, state->k );
+        state->k = _mm_sub_epi64( state->k, state->const1536 );
+        Compress( state, state->buffer, 1) ;
+   }
+*/
+
+   // Store the hash value
+   _mm_storeu_si128( (__m128i*)hashval + 0, state->state[0][0] );
+   _mm_storeu_si128( (__m128i*)hashval + 1, state->state[1][0] );
+
+   if( state->uHashSize == 512 )
+   {
+        _mm_storeu_si128( (__m128i*)hashval + 2, state->state[2][0] );
+        _mm_storeu_si128( (__m128i*)hashval + 3, state->state[3][0] );
+
+   }
+   return SUCCESS;
+}
+
+
+HashReturn hash_echo(int hashbitlen, const BitSequence *data, DataLength databitlen, BitSequence *hashval)
+{
+	HashReturn hRet;
+	hashState_echo hs;
+
+	/////
+	/*
+	__m128i a, b, c, d, t[4], u[4], v[4];
+
+	a = _mm_set_epi32(0x0f0e0d0c, 0x0b0a0908, 0x07060504, 0x03020100);
+	b = _mm_set_epi32(0x1f1e1d1c, 0x1b1a1918, 0x17161514, 0x13121110);
+	c = _mm_set_epi32(0x2f2e2d2c, 0x2b2a2928, 0x27262524, 0x23222120);
+	d = _mm_set_epi32(0x3f3e3d3c, 0x3b3a3938, 0x37363534, 0x33323130);
+
+	t[0] = _mm_unpacklo_epi8(a, b);
+	t[1] = _mm_unpackhi_epi8(a, b);
+	t[2] = _mm_unpacklo_epi8(c, d);
+	t[3] = _mm_unpackhi_epi8(c, d);
+
+	u[0] = _mm_unpacklo_epi16(t[0], t[2]);
+	u[1] = _mm_unpackhi_epi16(t[0], t[2]);
+	u[2] = _mm_unpacklo_epi16(t[1], t[3]);
+	u[3] = _mm_unpackhi_epi16(t[1], t[3]);
+
+
+	t[0] = _mm_unpacklo_epi16(u[0], u[1]);
+	t[1] = _mm_unpackhi_epi16(u[0], u[1]);
+	t[2] = _mm_unpacklo_epi16(u[2], u[3]);
+	t[3] = _mm_unpackhi_epi16(u[2], u[3]);
+
+	u[0] = _mm_unpacklo_epi8(t[0], t[1]);
+	u[1] = _mm_unpackhi_epi8(t[0], t[1]);
+	u[2] = _mm_unpacklo_epi8(t[2], t[3]);
+	u[3] = _mm_unpackhi_epi8(t[2], t[3]);
+
+	a = _mm_unpacklo_epi8(u[0], u[1]);
+	b = _mm_unpackhi_epi8(u[0], u[1]);
+	c = _mm_unpacklo_epi8(u[2], u[3]);
+	d = _mm_unpackhi_epi8(u[2], u[3]);
+	*/
+	/////
+
+	hRet = init_echo(&hs, hashbitlen);
+	if(hRet != SUCCESS)
+		return hRet;
+
+	hRet = update_echo(&hs, data, databitlen);
+	if(hRet != SUCCESS)
+		return hRet;
+
+	hRet = final_echo(&hs, hashval);
+	if(hRet != SUCCESS)
+		return hRet;
+
+	return SUCCESS;
+}
+
+#endif
--- a/algo/echo/echo-hash-4way.c
+++ b/algo/echo/echo-hash-4way.c
@@ -0,0 +1,316 @@
+//#if 0
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#include "simd-utils.h"
+#include "echo-hash-4way.h"
+
+/*
+static const unsigned int mul2ipt[] __attribute__ ((aligned (64))) =
+{  
+   0x728efc00, 0x6894e61a, 0x3fc3b14d, 0x25d9ab57,
+   0xfd5ba600, 0x2a8c71d7, 0x1eb845e3, 0xc96f9234
+};
+*/
+// do these need to be reversed?
+
+#define mul2mask \
+   _mm512_set4_epi32( 0, 0, 0, 0x00001b00 ) 
+//   _mm512_set4_epi32( 0x00001b00, 0, 0, 0 )  
+
+#define lsbmask    m512_const1_32( 0x01010101 ) 
+
+#define ECHO_SUBBYTES( state, i, j ) \
+	state[i][j] = _mm512_aesenc_epi128( state[i][j], k1 ); \
+	state[i][j] = _mm512_aesenc_epi128( state[i][j], m512_zero ); \
+	k1 = _mm512_add_epi32( k1, m512_one_128 );
+
+#define ECHO_MIXBYTES( state1, state2, j, t1, t2, s2 ) do \
+{ \
+   const int j1 = ( (j)+1 ) & 3; \
+   const int j2 = ( (j)+2 ) & 3; \
+   const int j3 = ( (j)+3 ) & 3; \
+   s2 = _mm512_add_epi8( state1[ 0 ] [j ], state1[ 0 ][ j ] ); \
+	t1 = _mm512_srli_epi16( state1[ 0 ][ j ], 7 ); \
+	t1 = _mm512_and_si512( t1, lsbmask );\
+	t2 = _mm512_shuffle_epi8( mul2mask, t1 ); \
+	s2 = _mm512_xor_si512( s2, t2 ); \
+	state2[ 0 ] [j ] = s2; \
+	state2[ 1 ] [j ] = state1[ 0 ][ j ]; \
+	state2[ 2 ] [j ] = state1[ 0 ][ j ]; \
+	state2[ 3 ] [j ] = _mm512_xor_si512( s2, state1[ 0 ][ j ] );\
+	s2 = _mm512_add_epi8( state1[ 1 ][ j1 ], state1[ 1 ][ j1 ] ); \
+	t1 = _mm512_srli_epi16( state1[ 1 ][ j1 ], 7 ); \
+	t1 = _mm512_and_si512( t1, lsbmask ); \
+	t2 = _mm512_shuffle_epi8( mul2mask, t1 ); \
+	s2 = _mm512_xor_si512( s2, t2 );\
+	state2[ 0 ][ j ] = _mm512_xor_si512( state2[ 0 ][ j ], \
+                            _mm512_xor_si512( s2, state1[ 1 ][ j1 ] ) ); \
+	state2[ 1 ][ j ] = _mm512_xor_si512( state2[ 1 ][ j ], s2 ); \
+	state2[ 2 ][ j ] = _mm512_xor_si512( state2[ 2 ][ j ], state1[ 1 ][ j1 ] ); \
+	state2[ 3 ][ j ] = _mm512_xor_si512( state2[ 3 ][ j ], state1[ 1 ][ j1 ] ); \
+	s2 = _mm512_add_epi8( state1[ 2 ][ j2 ], state1[ 2 ][ j2 ] ); \
+	t1 = _mm512_srli_epi16( state1[ 2 ][ j2 ], 7 ); \
+	t1 = _mm512_and_si512( t1, lsbmask ); \
+	t2 = _mm512_shuffle_epi8( mul2mask, t1 ); \
+	s2 = _mm512_xor_si512( s2, t2 ); \
+	state2[ 0 ][ j ] = _mm512_xor_si512( state2[ 0 ][ j ], state1[ 2 ][ j2 ] ); \
+	state2[ 1 ][ j ] = _mm512_xor_si512( state2[ 1 ][ j ], \
+                            _mm512_xor_si512( s2, state1[ 2 ][ j2 ] ) ); \
+	state2[ 2 ][ j ] = _mm512_xor_si512( state2[ 2 ][ j ], s2 ); \
+	state2[ 3 ][ j ] = _mm512_xor_si512( state2[ 3][ j ], state1[ 2 ][ j2 ] ); \
+	s2 = _mm512_add_epi8( state1[ 3 ][ j3 ], state1[ 3 ][ j3 ] ); \
+	t1 = _mm512_srli_epi16( state1[ 3 ][ j3 ], 7 ); \
+	t1 = _mm512_and_si512( t1, lsbmask ); \
+	t2 = _mm512_shuffle_epi8( mul2mask, t1 ); \
+	s2 = _mm512_xor_si512( s2, t2 ); \
+	state2[ 0 ][ j ] = _mm512_xor_si512( state2[ 0 ][ j ], state1[ 3 ][ j3 ] ); \
+	state2[ 1 ][ j ] = _mm512_xor_si512( state2[ 1 ][ j ], state1[ 3 ][ j3 ] ); \
+	state2[ 2 ][ j ] = _mm512_xor_si512( state2[ 2 ][ j ], \
+                            _mm512_xor_si512( s2, state1[ 3 ][ j3] ) ); \
+	state2[ 3 ][ j ] = _mm512_xor_si512( state2[ 3 ][ j ], s2 ); \
+} while(0)
+
+#define ECHO_ROUND_UNROLL2 \
+	ECHO_SUBBYTES(_state, 0, 0);\
+   ECHO_SUBBYTES(_state, 1, 0);\
+	ECHO_SUBBYTES(_state, 2, 0);\
+	ECHO_SUBBYTES(_state, 3, 0);\
+	ECHO_SUBBYTES(_state, 0, 1);\
+	ECHO_SUBBYTES(_state, 1, 1);\
+	ECHO_SUBBYTES(_state, 2, 1);\
+	ECHO_SUBBYTES(_state, 3, 1);\
+	ECHO_SUBBYTES(_state, 0, 2);\
+	ECHO_SUBBYTES(_state, 1, 2);\
+	ECHO_SUBBYTES(_state, 2, 2);\
+	ECHO_SUBBYTES(_state, 3, 2);\
+	ECHO_SUBBYTES(_state, 0, 3);\
+	ECHO_SUBBYTES(_state, 1, 3);\
+	ECHO_SUBBYTES(_state, 2, 3);\
+	ECHO_SUBBYTES(_state, 3, 3);\
+	ECHO_MIXBYTES(_state, _state2, 0, t1, t2, s2);\
+	ECHO_MIXBYTES(_state, _state2, 1, t1, t2, s2);\
+	ECHO_MIXBYTES(_state, _state2, 2, t1, t2, s2);\
+	ECHO_MIXBYTES(_state, _state2, 3, t1, t2, s2);\
+	ECHO_SUBBYTES(_state2, 0, 0);\
+	ECHO_SUBBYTES(_state2, 1, 0);\
+	ECHO_SUBBYTES(_state2, 2, 0);\
+	ECHO_SUBBYTES(_state2, 3, 0);\
+	ECHO_SUBBYTES(_state2, 0, 1);\
+	ECHO_SUBBYTES(_state2, 1, 1);\
+	ECHO_SUBBYTES(_state2, 2, 1);\
+	ECHO_SUBBYTES(_state2, 3, 1);\
+	ECHO_SUBBYTES(_state2, 0, 2);\
+	ECHO_SUBBYTES(_state2, 1, 2);\
+	ECHO_SUBBYTES(_state2, 2, 2);\
+	ECHO_SUBBYTES(_state2, 3, 2);\
+	ECHO_SUBBYTES(_state2, 0, 3);\
+	ECHO_SUBBYTES(_state2, 1, 3);\
+	ECHO_SUBBYTES(_state2, 2, 3);\
+	ECHO_SUBBYTES(_state2, 3, 3);\
+	ECHO_MIXBYTES(_state2, _state, 0, t1, t2, s2);\
+	ECHO_MIXBYTES(_state2, _state, 1, t1, t2, s2);\
+	ECHO_MIXBYTES(_state2, _state, 2, t1, t2, s2);\
+	ECHO_MIXBYTES(_state2, _state, 3, t1, t2, s2)
+
+#define SAVESTATE(dst, src)\
+	dst[0][0] = src[0][0];\
+	dst[0][1] = src[0][1];\
+	dst[0][2] = src[0][2];\
+	dst[0][3] = src[0][3];\
+	dst[1][0] = src[1][0];\
+	dst[1][1] = src[1][1];\
+	dst[1][2] = src[1][2];\
+	dst[1][3] = src[1][3];\
+	dst[2][0] = src[2][0];\
+	dst[2][1] = src[2][1];\
+	dst[2][2] = src[2][2];\
+	dst[2][3] = src[2][3];\
+	dst[3][0] = src[3][0];\
+	dst[3][1] = src[3][1];\
+	dst[3][2] = src[3][2];\
+	dst[3][3] = src[3][3]
+
+// blockcount always 1
+void echo_4way_compress( echo_4way_context *ctx, const __m512i *pmsg,
+               unsigned int uBlockCount )
+{
+  unsigned int r, b, i, j;
+  __m512i t1, t2, s2, k1;
+  __m512i _state[4][4], _state2[4][4], _statebackup[4][4]; 
+
+  _state[ 0 ][ 0 ] = ctx->state[ 0 ][ 0 ];
+  _state[ 0 ][ 1 ] = ctx->state[ 0 ][ 1 ];
+  _state[ 0 ][ 2 ] = ctx->state[ 0 ][ 2 ];
+  _state[ 0 ][ 3 ] = ctx->state[ 0 ][ 3 ];
+  _state[ 1 ][ 0 ] = ctx->state[ 1 ][ 0 ];
+  _state[ 1 ][ 1 ] = ctx->state[ 1 ][ 1 ];
+  _state[ 1 ][ 2 ] = ctx->state[ 1 ][ 2 ];
+  _state[ 1 ][ 3 ] = ctx->state[ 1 ][ 3 ];
+  _state[ 2 ][ 0 ] = ctx->state[ 2 ][ 0 ];
+  _state[ 2 ][ 1 ] = ctx->state[ 2 ][ 1 ];
+  _state[ 2 ][ 2 ] = ctx->state[ 2 ][ 2 ];
+  _state[ 2 ][ 3 ] = ctx->state[ 2 ][ 3 ];
+  _state[ 3 ][ 0 ] = ctx->state[ 3 ][ 0 ];
+  _state[ 3 ][ 1 ] = ctx->state[ 3 ][ 1 ];
+  _state[ 3 ][ 2 ] = ctx->state[ 3 ][ 2 ];
+  _state[ 3 ][ 3 ] = ctx->state[ 3 ][ 3 ];
+
+  for ( b = 0; b < uBlockCount; b++ )
+  {
+    ctx->k = _mm512_add_epi64( ctx->k, ctx->const1536 );
+
+    for( j = ctx->uHashSize / 256; j < 4; j++ )
+    {
+      for ( i = 0; i < 4; i++ )
+	   {
+        _state[ i ][ j ] = _mm512_load_si512( 
+                     pmsg + 4 * (j - (ctx->uHashSize / 256)) + i );
+	   }
+	 }
+    
+    // save state
+	 SAVESTATE( _statebackup, _state );
+
+	 k1 = ctx->k;
+
+	 for ( r = 0; r < ctx->uRounds / 2; r++ )
+	 {
+		ECHO_ROUND_UNROLL2;
+	 }
+		
+	 if ( ctx->uHashSize == 256 )
+	 {
+	   for ( i = 0; i < 4; i++ )
+	   {
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _state[ i ][ 1 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _state[ i ][ 2 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _state[ i ][ 3 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _statebackup[ i ][ 0 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _statebackup[ i ][ 1 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _statebackup[ i ][ 2 ] ) ;
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _statebackup[ i ][ 3 ] );
+	   }
+	 }
+	 else
+	 {
+	   for ( i = 0; i < 4; i++ )
+	   {
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _state[ i ][ 2 ] );
+		   _state[ i ][ 1 ] = _mm512_xor_si512( _state[ i ][ 1 ],
+                                              _state[ i ][ 3 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ][ 0 ],
+                                              _statebackup[ i ][ 0 ] );
+		   _state[ i ][ 0 ] = _mm512_xor_si512( _state[ i ] [0 ],
+                                              _statebackup[ i ][ 2 ] );
+		   _state[ i ][ 1 ] = _mm512_xor_si512( _state[ i ][ 1 ],
+                                              _statebackup[ i ][ 1 ] );
+		   _state[ i ][ 1 ] = _mm512_xor_si512( _state[ i ][ 1 ],
+                                              _statebackup[ i ][ 3 ] );
+      }
+	 }
+    pmsg += ctx->uBlockLength;
+  }
+  SAVESTATE(ctx->state, _state);
+
+}
+
+int echo_4way_init( echo_4way_context *ctx, int nHashSize )
+{
+	int i, j;
+
+   ctx->k = m512_zero; 
+	ctx->processed_bits = 0;
+	ctx->uBufferBytes = 0;
+
+	switch( nHashSize )
+	{
+		case 256:
+			ctx->uHashSize = 256;
+			ctx->uBlockLength = 192;
+			ctx->uRounds = 8;
+			ctx->hashsize = _mm512_set4_epi32( 0, 0, 0, 0x100 );
+			ctx->const1536 = _mm512_set4_epi32( 0, 0, 0, 0x600 );
+			break;
+
+		case 512:
+			ctx->uHashSize = 512;
+			ctx->uBlockLength = 128;
+			ctx->uRounds = 10;
+			ctx->hashsize = _mm512_set4_epi32( 0, 0, 0, 0x200 );
+			ctx->const1536 = _mm512_set4_epi32( 0, 0, 0, 0x400);
+			break;
+
+		default:
+			return 1;
+	}
+
+	for( i = 0; i < 4; i++ )
+		for( j = 0; j < nHashSize / 256; j++ )
+			ctx->state[ i ][ j ] = ctx->hashsize;
+
+	for( i = 0; i < 4; i++ )
+		for( j = nHashSize / 256; j < 4; j++ )
+			ctx->state[ i ][ j ] = m512_zero;
+
+	return 0;
+}
+
+int echo_4way_update_close( echo_4way_context *state, void *hashval,
+                              const void *data, int databitlen )
+{
+// bytelen is either 32 (maybe), 64 or 80 or 128!
+// all are less than full block.
+
+   int vlen = databitlen / 128;  // * 4 lanes / 128 bits per lane
+   const int vblen = state->uBlockLength / 16; //  16 bytes per lane
+   __m512i remainingbits;
+
+   if ( databitlen == 1024 )
+   {
+      echo_4way_compress( state, data, 1 );
+      state->processed_bits = 1024;
+      remainingbits = m512_const2_64( 0, -1024 );
+      vlen = 0;
+   }
+   else
+   {
+      vlen = databitlen / 128;  // * 4 lanes / 128 bits per lane
+      memcpy_512( state->buffer, data, vlen );
+      state->processed_bits += (unsigned int)( databitlen );
+      remainingbits = _mm512_set4_epi32( 0, 0, 0, databitlen );
+
+   }
+
+   state->buffer[ vlen ] = _mm512_set4_epi32( 0, 0, 0, 0x80 );
+   memset_zero_512( state->buffer + vlen + 1, vblen - vlen - 2 );
+   state->buffer[ vblen-2 ] =
+                _mm512_set4_epi32( (uint32_t)state->uHashSize << 16, 0, 0, 0 );
+   state->buffer[ vblen-1 ] =
+                   _mm512_set4_epi64( 0, state->processed_bits,
+                                      0, state->processed_bits );  
+
+   state->k = _mm512_add_epi64( state->k, remainingbits );
+   state->k = _mm512_sub_epi64( state->k, state->const1536 );
+
+   echo_4way_compress( state, state->buffer, 1 );
+
+   _mm512_store_si512( (__m512i*)hashval + 0, state->state[ 0 ][ 0] );
+   _mm512_store_si512( (__m512i*)hashval + 1, state->state[ 1 ][ 0] );
+
+   if ( state->uHashSize == 512 )
+   {
+      _mm512_store_si512( (__m512i*)hashval + 2, state->state[ 2 ][ 0 ] );
+      _mm512_store_si512( (__m512i*)hashval + 3, state->state[ 3 ][ 0 ] );
+   }
+   return 0;
+}
+
+#endif
--- a/algo/echo/echo-hash-4way.h
+++ b/algo/echo/echo-hash-4way.h
@@ -0,0 +1,36 @@
+#if !defined(ECHO_HASH_4WAY_H__)
+#define ECHO_HASH_4WAY_H__ 1
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#include "simd-utils.h"
+
+typedef struct
+{
+   __m512i    state[4][4];
+   __m512i    buffer[ 4 * 192 / 16 ];  // 4x128 interleaved 192 bytes
+   __m512i    k;
+   __m512i    hashsize;
+   __m512i    const1536;
+
+   unsigned int   uRounds;
+   unsigned int   uHashSize;
+   unsigned int   uBlockLength;
+   unsigned int   uBufferBytes;
+   unsigned int   processed_bits;
+
+} echo_4way_context __attribute__ ((aligned (64)));
+
+int echo_4way_init( echo_4way_context *state, int hashbitlen );
+
+
+int echo_4way_update( echo_4way_context *state, const void *data,
+    unsigned int databitlen);
+
+int echo_close( echo_4way_context *state, void *hashval );
+
+int echo_4way_update_close( echo_4way_context *state, void *hashval,
+                              const void *data, int databitlen );
+
+#endif 
+#endif
--- a/algo/groestl/aes_ni/groestl-intr-aes.h
+++ b/algo/groestl/aes_ni/groestl-intr-aes.h
@@ -73,7 +73,7 @@ __m128i ALL_FF;
  b5 = a7;\
  a6 = _mm_xor_si128(a6, a7);\
  a7 = _mm_xor_si128(a7, b6);\
-  \
+   \
  /* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
  b0 = _mm_xor_si128(b0, a4);\
  b6 = _mm_xor_si128(b6, a4);\
@@ -195,7 +195,7 @@ __m128i ALL_FF;
  for(round_counter = 0; round_counter < 14; round_counter+=2) {\
    /* AddRoundConstant P1024 */\
    xmm8 = _mm_xor_si128(xmm8, (ROUND_CONST_P[round_counter]));\
-    /* ShiftBytes P1024 + pre-AESENCLAST */\
+     /* ShiftBytes P1024 + pre-AESENCLAST */\
    xmm8  = _mm_shuffle_epi8(xmm8,  (SUBSH_MASK[0]));\
    xmm9  = _mm_shuffle_epi8(xmm9,  (SUBSH_MASK[1]));\
    xmm10 = _mm_shuffle_epi8(xmm10, (SUBSH_MASK[2]));\
@@ -209,7 +209,6 @@ __m128i ALL_FF;
    \
    /* AddRoundConstant P1024 */\
    xmm0 = _mm_xor_si128(xmm0, (ROUND_CONST_P[round_counter+1]));\
-    /* ShiftBytes P1024 + pre-AESENCLAST */\
    xmm0 = _mm_shuffle_epi8(xmm0, (SUBSH_MASK[0]));\
    xmm1 = _mm_shuffle_epi8(xmm1, (SUBSH_MASK[1]));\
    xmm2 = _mm_shuffle_epi8(xmm2, (SUBSH_MASK[2]));\
@@ -218,7 +217,6 @@ __m128i ALL_FF;
    xmm5 = _mm_shuffle_epi8(xmm5, (SUBSH_MASK[5]));\
    xmm6 = _mm_shuffle_epi8(xmm6, (SUBSH_MASK[6]));\
    xmm7 = _mm_shuffle_epi8(xmm7, (SUBSH_MASK[7]));\
-    /* SubBytes + MixBytes */\
    SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
  }\
 }
--- a/algo/groestl/aes_ni/groestl-version.h
+++ b/algo/groestl/aes_ni/groestl-version.h
@@ -9,6 +9,7 @@

 //#ifndef NO_AES_NI

+// Not to be confused with AVX512VAES
 #define VAES
 // #define VAVX
 // #define VVPERM
--- a/algo/groestl/aes_ni/hash-groestl.c
+++ b/algo/groestl/aes_ni/hash-groestl.c
@@ -230,6 +230,7 @@ HashReturn_gr update_and_final_groestl( hashState_groestl* ctx, void* output,

   // digest final padding block and do output transform
   TF1024( ctx->chaining, ctx->buffer );
+
   OF1024( ctx->chaining );

   // store hash result in output 
--- a/algo/groestl/groestl-4way.c
+++ b/algo/groestl/groestl-4way.c
@@ -0,0 +1,64 @@
+#include "groestl-gate.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+
+#if defined(GROESTL_4WAY_VAES)
+
+#include "groestl512-hash-4way.h"
+
+void groestl_4way_hash( void *output, const void *input )
+{
+     uint32_t hash[16*4] __attribute__ ((aligned (128)));
+     groestl512_4way_context ctx;
+
+     groestl512_4way_init( &ctx, 64 );
+     groestl512_4way_update_close( &ctx, hash, input, 640 );
+
+     groestl512_4way_init( &ctx, 64 );
+     groestl512_4way_update_close( &ctx, hash, hash, 512 );
+
+     dintrlv_4x128( output, output+32, output+64, output+96, hash, 256 );
+ }
+
+int scanhash_groestl_4way( struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
+{
+     uint32_t hash[8*4] __attribute__ ((aligned (128)));
+     uint32_t vdata[24*4] __attribute__ ((aligned (64)));
+     uint32_t *pdata = work->data;
+     uint32_t *ptarget = work->target;
+     uint32_t n = pdata[19];
+     const uint32_t first_nonce = pdata[19];
+     const uint32_t last_nonce = max_nonce - 4;
+     uint32_t *noncep = vdata + 64+3;   // 4*16 + 3
+     int thr_id = mythr->id;
+     const uint32_t Htarg = ptarget[7];
+
+     mm512_bswap32_intrlv80_4x128( vdata, pdata );
+
+     do
+     {
+        be32enc( noncep,    n   );
+        be32enc( noncep+ 4, n+1 );
+        be32enc( noncep+ 8, n+2 );
+        be32enc( noncep+12, n+3 );
+
+        groestl_4way_hash( hash, vdata );
+        pdata[19] = n;
+
+        for ( int lane = 0; lane < 4; lane++ )
+        if ( ( hash+(lane<<3) )[7] < Htarg )
+        if ( fulltest( hash+(lane<<3), ptarget) && !opt_benchmark )
+        {
+           pdata[19] = n + lane;
+           submit_lane_solution( work, hash+(lane<<3), mythr, lane );
+        }
+        n += 4;
+     } while ( ( n < last_nonce ) && !work_restart[thr_id].restart );
+     *hashes_done = n - first_nonce;
+     return 0;
+}
+
+#endif
--- a/algo/groestl/groestl-gate.c
+++ b/algo/groestl/groestl-gate.c
@@ -0,0 +1,23 @@
+#include "groestl-gate.h"
+
+bool register_dmd_gr_algo( algo_gate_t *gate )
+{
+#if defined (GROESTL_4WAY_VAES)
+  gate->scanhash  = (void*)&scanhash_groestl_4way;
+  gate->hash      = (void*)&groestl_4way_hash;
+#else
+  init_groestl_ctx();
+  gate->scanhash  = (void*)&scanhash_groestl;
+  gate->hash      = (void*)&groestlhash;
+#endif
+  gate->optimizations = AES_OPT | VAES_OPT;
+  return true;
+};
+
+bool register_groestl_algo( algo_gate_t* gate )
+{
+    register_dmd_gr_algo( gate );
+    gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
+    return true;
+};
+
--- a/algo/groestl/groestl-gate.h
+++ b/algo/groestl/groestl-gate.h
@@ -0,0 +1,31 @@
+#ifndef GROESTL_GATE_H__
+#define GROESTL_GATE_H__ 1
+
+#include "algo-gate-api.h"
+#include <stdint.h>
+
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define GROESTL_4WAY_VAES 1
+#endif
+
+bool register_dmd_gr_algo( algo_gate_t* gate );
+
+bool register_groestl_algo( algo_gate_t* gate );
+
+#if defined(GROESTL_4WAY_VAES)
+
+void groestl_4way_hash( void *state, const void *input );
+int scanhash_groestl_4way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+#else
+
+void groestlhash( void *state, const void *input );
+int scanhash_groestl( struct work *work, uint32_t max_nonce,
+                    uint64_t *hashes_done, struct thr_info *mythr );
+void init_groestl_ctx();
+
+#endif
+
+#endif
+
--- a/algo/groestl/groestl.c
+++ b/algo/groestl/groestl.c
@@ -1,5 +1,4 @@
-#include "algo-gate-api.h"
-
+#include "groestl-gate.h"
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
@@ -78,15 +77,12 @@ int scanhash_groestl( struct work *work, uint32_t max_nonce,
 		groestlhash(hash, endiandata);

 		if (hash[7] <= Htarg )
-                   if ( fulltest(hash, ptarget))
-                   {
+      if ( fulltest(hash, ptarget) && !opt_benchmark )
+      {
 			pdata[19] = nonce;
-			*hashes_done = pdata[19] - first_nonce;
-			return 1;
-	           }
-         
+         submit_solution( work, hash, mythr );
+	   }
 		nonce++;
-
 	} while (nonce < max_nonce && !work_restart[thr_id].restart);

 	pdata[19] = nonce;
@@ -94,26 +90,3 @@ int scanhash_groestl( struct work *work, uint32_t max_nonce,
 	return 0;
 }

-void groestl_set_target( struct work* work, double job_diff )
-{
- work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
-}
-
-bool register_dmd_gr_algo( algo_gate_t* gate )
-{
-    init_groestl_ctx();
-    gate->optimizations   = SSE2_OPT | AES_OPT;
-    gate->scanhash        = (void*)&scanhash_groestl;
-    gate->hash            = (void*)&groestlhash;
-    gate->set_target      = (void*)&groestl_set_target;
-    gate->get_max64       = (void*)&get_max64_0x3ffff;
-    return true;
-};
-
-bool register_groestl_algo( algo_gate_t* gate )
-{
-    register_dmd_gr_algo( gate );
-    gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
-    return true;
-};
-
--- a/algo/groestl/groestl256-hash-4way.c
+++ b/algo/groestl/groestl256-hash-4way.c
@@ -0,0 +1,280 @@
+/* hash.c     Aug 2011
+ *
+ * Groestl implementation for different versions.
+ * Author: Krystian Matusiewicz, Günther A. Roland, Martin Schläffer
+ *
+ * This code is placed in the public domain
+ */
+
+#include <memory.h>
+#include "hash-groestl256.h"
+#include "miner.h"
+#include "simd-utils.h"
+
+#ifndef NO_AES_NI
+
+#include "groestl-version.h"
+
+#ifdef TASM
+  #ifdef VAES
+    #include "groestl256-asm-aes.h"
+  #else
+    #ifdef VAVX
+      #include "groestl256-asm-avx.h"
+    #else
+      #ifdef VVPERM
+        #include "groestl256-asm-vperm.h"
+      #else
+        #error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
+      #endif
+    #endif
+  #endif
+#else
+  #ifdef TINTR
+    #ifdef VAES
+      #include "groestl256-intr-aes.h"
+    #else
+      #ifdef VAVX
+        #include "groestl256-intr-avx.h"
+      #else
+        #ifdef VVPERM
+          #include "groestl256-intr-vperm.h"
+        #else
+          #error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
+        #endif
+      #endif
+    #endif
+  #else
+    #error NO TYPE SPECIFIED (-DT[ASM/INTR])
+  #endif
+#endif
+
+/* initialise context */
+HashReturn_gr init_groestl256( hashState_groestl256* ctx, int hashlen )
+{
+  int i;
+
+  ctx->hashlen = hashlen;
+  SET_CONSTANTS();
+
+  if (ctx->chaining == NULL || ctx->buffer == NULL)
+    return FAIL_GR;
+
+  for ( i = 0; i < SIZE256; i++ )
+  {
+     ctx->chaining[i] = _mm_setzero_si128();
+     ctx->buffer[i]   = _mm_setzero_si128();
+  }
+  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
+  INIT256( ctx->chaining );
+  ctx->buf_ptr = 0;
+  ctx->rem_ptr = 0;
+
+  return SUCCESS_GR;
+}
+
+
+HashReturn_gr reinit_groestl256(hashState_groestl256* ctx)
+ {
+  int i;
+
+  if (ctx->chaining == NULL || ctx->buffer == NULL)
+    return FAIL_GR;
+
+  for ( i = 0; i < SIZE256; i++ )
+  {
+     ctx->chaining[i] = _mm_setzero_si128();
+     ctx->buffer[i]   = _mm_setzero_si128();
+  }
+  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
+  INIT256(ctx->chaining);
+  ctx->buf_ptr = 0;
+  ctx->rem_ptr = 0;
+
+  return SUCCESS_GR;
+}
+
+// Use this only for midstate and never for cryptonight
+HashReturn_gr update_groestl256( hashState_groestl256* ctx, const void* input,
+                                 DataLength_gr databitlen )
+{
+   __m128i* in = (__m128i*)input;
+   const int len = (int)databitlen / 128;  // bits to __m128i
+   const int blocks = len / SIZE256;    // __M128i to blocks
+   int rem = ctx->rem_ptr;
+   int i;
+
+   ctx->blk_count = blocks;
+   ctx->databitlen = databitlen;
+
+   // digest any full blocks 
+   for ( i = 0; i < blocks; i++ )
+       TF512( ctx->chaining, &in[ i * SIZE256 ] );
+   // adjust buf_ptr to last block
+   ctx->buf_ptr = blocks * SIZE256;
+
+   // Copy any remainder to buffer
+   for ( i = 0; i < len % SIZE256; i++ )
+       ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
+   // adjust rem_ptr for new data
+   ctx->rem_ptr += i;
+
+   return SUCCESS_GR;
+}
+
+// don't use this at all
+HashReturn_gr final_groestl256( hashState_groestl256* ctx, void* output )
+{
+   const int len = (int)ctx->databitlen / 128;  // bits to __m128i 
+   const int blocks = ctx->blk_count + 1;       // adjust for final block
+   const int rem_ptr = ctx->rem_ptr;      // end of data start of padding
+   const int hashlen_m128i = ctx->hashlen / 16;  // bytes to __m128i
+   const int hash_offset = SIZE256 - hashlen_m128i;  // where in buffer
+   int i;
+
+   // first pad byte = 0x80, last pad byte = block count
+   // everything in between is zero
+
+   if ( rem_ptr == len - 1 )
+   {
+       // all padding at once
+       ctx->buffer[rem_ptr] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
+                                                  0,0,0,0, 0,0,0,0x80 );
+   }
+   else
+   {
+       // add first padding
+       ctx->buffer[rem_ptr] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
+                                            0,0,0,0, 0,0,0,0x80 );
+       // add zero padding
+       for ( i = rem_ptr + 1; i < SIZE256 - 1; i++ )
+           ctx->buffer[i] = _mm_setzero_si128();
+       // add length padding
+       // cheat since we know the block count is trivial, good if block < 256
+       ctx->buffer[i] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
+                                           0,0,0,0, 0,0,0,0 );
+   }
+
+   // digest final padding block and do output transform
+   TF512( ctx->chaining, ctx->buffer );
+   OF512( ctx->chaining );
+
+   // store hash result in output 
+   for ( i = 0; i < hashlen_m128i; i++ )
+      casti_m128i( output, i ) = ctx->chaining[ hash_offset + i];
+
+   return SUCCESS_GR;
+}
+
+HashReturn_gr update_and_final_groestl256( hashState_groestl256* ctx,
+                   void* output, const void* input, DataLength_gr databitlen )
+{
+   const int len = (int)databitlen / 128;
+   const int hashlen_m128i = ctx->hashlen / 16;   // bytes to __m128i
+   const int hash_offset = SIZE256 - hashlen_m128i;
+   int rem = ctx->rem_ptr;
+   int blocks = len / SIZE256;
+   __m128i* in = (__m128i*)input;
+   int i;
+
+   // --- update ---
+
+   // digest any full blocks, process directly from input 
+   for ( i = 0; i < blocks; i++ )
+      TF512( ctx->chaining, &in[ i * SIZE256 ] );
+   ctx->buf_ptr = blocks * SIZE256;
+
+   // cryptonight has 200 byte input, an odd number of __m128i
+   // remainder is only 8 bytes, ie u64.
+   if ( databitlen % 128 !=0 )
+   {
+      // must be cryptonight, copy 64 bits of data
+      *(uint64_t*)(ctx->buffer) = *(uint64_t*)(&in[ ctx->buf_ptr ] );
+      i = -1; // signal for odd length
+   }
+   else   
+   { 
+      // Copy any remaining data to buffer for final transform
+      for ( i = 0; i < len % SIZE256; i++ )
+          ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
+      i += rem;   // use i as rem_ptr in final
+   }
+
+   //--- final ---
+
+   // adjust for final block
+   blocks++;
+
+   if ( i == len - 1 )
+   {
+       // all padding at once
+       ctx->buffer[i] = _mm_set_epi8( blocks,blocks>>8,0,0, 0,0,0,0,
+                                           0,        0,0,0, 0,0,0,0x80 );
+   }
+   else
+   {
+      if ( i == -1 )
+      {
+         // cryptonight odd length
+         ((uint64_t*)ctx->buffer)[ 1 ] = 0x80ull;
+         // finish the block with zero and length padding as normal
+         i = 0;
+       }
+       else
+       {
+          // add first padding
+          ctx->buffer[i] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
+                                         0,0,0,0, 0,0,0,0x80 );
+       }
+       // add zero padding
+       for ( i += 1; i < SIZE256 - 1; i++ )
+           ctx->buffer[i] = _mm_setzero_si128();
+       // add length padding
+       // cheat since we know the block count is trivial, good if block < 256
+       ctx->buffer[i] = _mm_set_epi8( blocks,blocks>>8,0,0, 0,0,0,0,
+                                           0,        0,0,0, 0,0,0,0 );
+   }
+
+   // digest final padding block and do output transform
+   TF512( ctx->chaining, ctx->buffer );
+   OF512( ctx->chaining );
+
+   // store hash result in output 
+   for ( i = 0; i < hashlen_m128i; i++ )
+      casti_m128i( output, i ) = ctx->chaining[ hash_offset + i ];
+
+   return SUCCESS_GR;
+}
+
+/* hash bit sequence */
+HashReturn_gr hash_groestl256(int hashbitlen,
+                const BitSequence_gr* data,
+                DataLength_gr databitlen,
+                BitSequence_gr* hashval) {
+  HashReturn_gr ret;
+  hashState_groestl256 context;
+
+  /* initialise */
+  if ((ret = init_groestl256(&context, hashbitlen/8)) != SUCCESS_GR)
+    return ret;
+
+  /* process message */
+  if ((ret = update_groestl256(&context, data, databitlen)) != SUCCESS_GR)
+    return ret;
+
+  /* finalise */
+  ret = final_groestl256(&context, hashval);
+
+  return ret;
+}
+
+/* eBash API */
+//#ifdef crypto_hash_BYTES
+//int crypto_hash(unsigned char *out, const unsigned char *in, unsigned long long inlen)
+//{
+//  if (hash_groestl(crypto_hash_BYTES * 8, in, inlen * 8,out) == SUCCESS_GR) return 0;
+//  return -1;
+//}
+//#endif
+
+#endif
--- a/algo/groestl/groestl256-hash-4way.h
+++ b/algo/groestl/groestl256-hash-4way.h
@@ -0,0 +1,121 @@
+/* hash.h     Aug 2011
+ *
+ * Groestl implementation for different versions.
+ * Author: Krystian Matusiewicz, Günther A. Roland, Martin Schläffer
+ *
+ * This code is placed in the public domain
+ */
+
+#ifndef __hash_h
+#define __hash_h
+
+#include <immintrin.h>
+#include <stdio.h>
+#if defined(_WIN64) || defined(__WINDOWS__)
+#include <windows.h>
+#endif
+#include <stdlib.h>
+
+/* eBash API begin */
+/*
+#include "crypto_hash.h"
+#ifdef crypto_hash_BYTES
+
+#include <crypto_uint8.h>
+#include <crypto_uint32.h>
+#include <crypto_uint64.h>
+typedef crypto_uint8 u8;
+typedef crypto_uint32 u32;
+typedef crypto_uint64 u64;
+#endif
+ */
+/* eBash API end */
+
+//#define LENGTH (512)
+
+#include "brg_endian.h"
+#define NEED_UINT_64T
+#include "algo/sha/brg_types.h"
+
+#ifdef IACA_TRACE
+  #include IACA_MARKS
+#endif
+
+#define LENGTH (256)
+
+/* some sizes (number of bytes) */
+#define ROWS (8)
+#define LENGTHFIELDLEN (ROWS)
+#define COLS512 (8)
+//#define COLS1024 (16)
+#define SIZE_512 ((ROWS)*(COLS512))
+//#define SIZE1024 ((ROWS)*(COLS1024))
+#define ROUNDS512 (10)
+//#define ROUNDS1024 (14)
+
+//#if LENGTH<=256
+#define COLS (COLS512)
+//#define SIZE (SIZE512)
+#define ROUNDS (ROUNDS512)
+//#else
+//#define COLS (COLS1024)
+//#define SIZE (SIZE1024)
+//#define ROUNDS (ROUNDS1024)
+//#endif
+
+#define ROTL64(a,n) ((((a)<<(n))|((a)>>(64-(n))))&li_64(ffffffffffffffff))
+
+#if (PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN)
+#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*(7-(n)))))
+#define U64BIG(a) (a)
+#endif /* IS_BIG_ENDIAN */
+
+#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
+#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*n)))
+#define U64BIG(a) \
+  ((ROTL64(a, 8) & li_64(000000FF000000FF)) | \
+   (ROTL64(a,24) & li_64(0000FF000000FF00)) | \
+   (ROTL64(a,40) & li_64(00FF000000FF0000)) | \
+   (ROTL64(a,56) & li_64(FF000000FF000000)))
+#endif /* IS_LITTLE_ENDIAN */
+
+typedef unsigned char BitSequence_gr;
+typedef unsigned long long DataLength_gr;
+typedef enum
+{
+    SUCCESS_GR = 0,
+    FAIL_GR = 1,
+    BAD_HASHBITLEN_GR = 2
+} HashReturn_gr;
+
+#define SIZE256 (SIZE_512/16)
+
+typedef struct {
+  __attribute__ ((aligned (32))) __m128i chaining[SIZE256];
+  __attribute__ ((aligned (32))) __m128i buffer[SIZE256];
+//  __attribute__ ((aligned (32))) u64 chaining[SIZE/8];      /* actual state */
+//  __attribute__ ((aligned (32))) BitSequence_gr buffer[SIZE];  /* data buffer */
+//  u64 block_counter;        /* message block counter */
+  int hashlen;              // bytes
+  int blk_count;
+  int buf_ptr;              /* data buffer pointer */
+  int rem_ptr;
+  int databitlen;
+} hashState_groestl256;
+
+HashReturn_gr init_groestl256( hashState_groestl256*, int );
+
+HashReturn_gr reinit_groestl256( hashState_groestl256* );
+
+HashReturn_gr update_groestl256( hashState_groestl256*, const void*,
+                              DataLength_gr );
+
+HashReturn_gr final_groestl256( hashState_groestl256*, void* );
+
+HashReturn_gr hash_groestli256( int, const BitSequence_gr*, DataLength_gr,
+                            BitSequence_gr* );
+
+HashReturn_gr update_and_final_groestl256( hashState_groestl256*, void*,
+                                           const void*, DataLength_gr );
+
+#endif /* __hash_h */
--- a/algo/groestl/groestl256-intr-4way.h
+++ b/algo/groestl/groestl256-intr-4way.h
@@ -0,0 +1,492 @@
+/* groestl-intr-aes.h     Aug 2011
+ *
+ * Groestl implementation with intrinsics using ssse3, sse4.1, and aes
+ * instructions.
+ * Author: Günther A. Roland, Martin Schläffer, Krystian Matusiewicz
+ *
+ * This code is placed in the public domain
+ */
+
+#include <smmintrin.h>
+#include <wmmintrin.h>
+#include "hash-groestl256.h"
+
+/* global constants  */
+__m128i ROUND_CONST_Lx;
+__m128i ROUND_CONST_L0[ROUNDS512];
+__m128i ROUND_CONST_L7[ROUNDS512];
+//__m128i ROUND_CONST_P[ROUNDS1024];
+//__m128i ROUND_CONST_Q[ROUNDS1024];
+__m128i TRANSP_MASK;
+__m128i SUBSH_MASK[8];
+__m128i ALL_1B;
+__m128i ALL_FF;
+
+
+#define tos(a)    #a
+#define tostr(a)  tos(a)
+
+
+/* xmm[i] will be multiplied by 2
+ * xmm[j] will be lost
+ * xmm[k] has to be all 0x1b */
+#define MUL2(i, j, k){\
+  j = _mm_xor_si128(j, j);\
+  j = _mm_cmpgt_epi8(j, i);\
+  i = _mm_add_epi8(i, i);\
+  j = _mm_and_si128(j, k);\
+  i = _mm_xor_si128(i, j);\
+} 
+
+ /**/
+
+/* Yet another implementation of MixBytes.
+   This time we use the formulae (3) from the paper "Byte Slicing Groestl".
+   Input: a0, ..., a7
+   Output: b0, ..., b7 = MixBytes(a0,...,a7).
+   but we use the relations:
+   t_i = a_i + a_{i+3}
+   x_i = t_i + t_{i+3}
+   y_i = t_i + t+{i+2} + a_{i+6}
+   z_i = 2*x_i
+   w_i = z_i + y_{i+4}
+   v_i = 2*w_i
+   b_i = v_{i+3} + y_{i+4}
+   We keep building b_i in registers xmm8..xmm15 by first building y_{i+4} there
+   and then adding v_i computed in the meantime in registers xmm0..xmm7.
+   We almost fit into 16 registers, need only 3 spills to memory.
+   This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
+   K. Matusiewicz, 2011/05/29 */
+#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
+  /* t_i = a_i + a_{i+1} */\
+  b6 = a0;\
+  b7 = a1;\
+  a0 = _mm_xor_si128(a0, a1);\
+  b0 = a2;\
+  a1 = _mm_xor_si128(a1, a2);\
+  b1 = a3;\
+  a2 = _mm_xor_si128(a2, a3);\
+  b2 = a4;\
+  a3 = _mm_xor_si128(a3, a4);\
+  b3 = a5;\
+  a4 = _mm_xor_si128(a4, a5);\
+  b4 = a6;\
+  a5 = _mm_xor_si128(a5, a6);\
+  b5 = a7;\
+  a6 = _mm_xor_si128(a6, a7);\
+  a7 = _mm_xor_si128(a7, b6);\
+  \
+  /* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
+  b0 = _mm_xor_si128(b0, a4);\
+  b6 = _mm_xor_si128(b6, a4);\
+  b1 = _mm_xor_si128(b1, a5);\
+  b7 = _mm_xor_si128(b7, a5);\
+  b2 = _mm_xor_si128(b2, a6);\
+  b0 = _mm_xor_si128(b0, a6);\
+  /* spill values y_4, y_5 to memory */\
+  TEMP0 = b0;\
+  b3 = _mm_xor_si128(b3, a7);\
+  b1 = _mm_xor_si128(b1, a7);\
+  TEMP1 = b1;\
+  b4 = _mm_xor_si128(b4, a0);\
+  b2 = _mm_xor_si128(b2, a0);\
+  /* save values t0, t1, t2 to xmm8, xmm9 and memory */\
+  b0 = a0;\
+  b5 = _mm_xor_si128(b5, a1);\
+  b3 = _mm_xor_si128(b3, a1);\
+  b1 = a1;\
+  b6 = _mm_xor_si128(b6, a2);\
+  b4 = _mm_xor_si128(b4, a2);\
+  TEMP2 = a2;\
+  b7 = _mm_xor_si128(b7, a3);\
+  b5 = _mm_xor_si128(b5, a3);\
+  \
+  /* compute x_i = t_i + t_{i+3} */\
+  a0 = _mm_xor_si128(a0, a3);\
+  a1 = _mm_xor_si128(a1, a4);\
+  a2 = _mm_xor_si128(a2, a5);\
+  a3 = _mm_xor_si128(a3, a6);\
+  a4 = _mm_xor_si128(a4, a7);\
+  a5 = _mm_xor_si128(a5, b0);\
+  a6 = _mm_xor_si128(a6, b1);\
+  a7 = _mm_xor_si128(a7, TEMP2);\
+  \
+  /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
+  /* compute w_i : add y_{i+4} */\
+  b1 = ALL_1B;\
+  MUL2(a0, b0, b1);\
+  a0 = _mm_xor_si128(a0, TEMP0);\
+  MUL2(a1, b0, b1);\
+  a1 = _mm_xor_si128(a1, TEMP1);\
+  MUL2(a2, b0, b1);\
+  a2 = _mm_xor_si128(a2, b2);\
+  MUL2(a3, b0, b1);\
+  a3 = _mm_xor_si128(a3, b3);\
+  MUL2(a4, b0, b1);\
+  a4 = _mm_xor_si128(a4, b4);\
+  MUL2(a5, b0, b1);\
+  a5 = _mm_xor_si128(a5, b5);\
+  MUL2(a6, b0, b1);\
+  a6 = _mm_xor_si128(a6, b6);\
+  MUL2(a7, b0, b1);\
+  a7 = _mm_xor_si128(a7, b7);\
+  \
+  /* compute v_i : double w_i      */\
+  /* add to y_4 y_5 .. v3, v4, ... */\
+  MUL2(a0, b0, b1);\
+  b5 = _mm_xor_si128(b5, a0);\
+  MUL2(a1, b0, b1);\
+  b6 = _mm_xor_si128(b6, a1);\
+  MUL2(a2, b0, b1);\
+  b7 = _mm_xor_si128(b7, a2);\
+  MUL2(a5, b0, b1);\
+  b2 = _mm_xor_si128(b2, a5);\
+  MUL2(a6, b0, b1);\
+  b3 = _mm_xor_si128(b3, a6);\
+  MUL2(a7, b0, b1);\
+  b4 = _mm_xor_si128(b4, a7);\
+  MUL2(a3, b0, b1);\
+  MUL2(a4, b0, b1);\
+  b0 = TEMP0;\
+  b1 = TEMP1;\
+  b0 = _mm_xor_si128(b0, a3);\
+  b1 = _mm_xor_si128(b1, a4);\
+}/*MixBytes*/
+
+#define SET_CONSTANTS(){\
+   ALL_1B = _mm_set_epi32(0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b);\
+  TRANSP_MASK = _mm_set_epi32(0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800);\
+  SUBSH_MASK[0] = _mm_set_epi32(0x03060a0d, 0x08020509, 0x0c0f0104, 0x070b0e00);\
+  SUBSH_MASK[1] = _mm_set_epi32(0x04070c0f, 0x0a03060b, 0x0e090205, 0x000d0801);\
+  SUBSH_MASK[2] = _mm_set_epi32(0x05000e09, 0x0c04070d, 0x080b0306, 0x010f0a02);\
+  SUBSH_MASK[3] = _mm_set_epi32(0x0601080b, 0x0e05000f, 0x0a0d0407, 0x02090c03);\
+  SUBSH_MASK[4] = _mm_set_epi32(0x0702090c, 0x0f060108, 0x0b0e0500, 0x030a0d04);\
+  SUBSH_MASK[5] = _mm_set_epi32(0x00030b0e, 0x0907020a, 0x0d080601, 0x040c0f05);\
+  SUBSH_MASK[6] = _mm_set_epi32(0x01040d08, 0x0b00030c, 0x0f0a0702, 0x050e0906);\
+  SUBSH_MASK[7] = _mm_set_epi32(0x02050f0a, 0x0d01040e, 0x090c0003, 0x06080b07);\
+  for(i = 0; i < ROUNDS512; i++)\
+  {\
+    ROUND_CONST_L0[i] = _mm_set_epi32(0xffffffff, 0xffffffff, 0x70605040 ^ (i * 0x01010101), 0x30201000 ^ (i * 0x01010101));\
+    ROUND_CONST_L7[i] = _mm_set_epi32(0x8f9fafbf ^ (i * 0x01010101), 0xcfdfefff ^ (i * 0x01010101), 0x00000000, 0x00000000);\
+  }\
+  ROUND_CONST_Lx = _mm_set_epi32(0xffffffff, 0xffffffff, 0x00000000, 0x00000000);\
+}while(0); \
+
+/* one round
+ * i = round number
+ * a0-a7 = input rows
+ * b0-b7 = output rows
+ */
+#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
+  /* AddRoundConstant */\
+  b1 = ROUND_CONST_Lx;\
+  a0 = _mm_xor_si128(a0, (ROUND_CONST_L0[i]));\
+  a1 = _mm_xor_si128(a1, b1);\
+  a2 = _mm_xor_si128(a2, b1);\
+  a3 = _mm_xor_si128(a3, b1);\
+  a4 = _mm_xor_si128(a4, b1);\
+  a5 = _mm_xor_si128(a5, b1);\
+  a6 = _mm_xor_si128(a6, b1);\
+  a7 = _mm_xor_si128(a7, (ROUND_CONST_L7[i]));\
+  \
+  /* ShiftBytes + SubBytes (interleaved) */\
+  b0 = _mm_xor_si128(b0,  b0);\
+  a0 = _mm_shuffle_epi8(a0, (SUBSH_MASK[0]));\
+  a0 = _mm_aesenclast_si128(a0, b0);\
+  a1 = _mm_shuffle_epi8(a1, (SUBSH_MASK[1]));\
+  a1 = _mm_aesenclast_si128(a1, b0);\
+  a2 = _mm_shuffle_epi8(a2, (SUBSH_MASK[2]));\
+  a2 = _mm_aesenclast_si128(a2, b0);\
+  a3 = _mm_shuffle_epi8(a3, (SUBSH_MASK[3]));\
+  a3 = _mm_aesenclast_si128(a3, b0);\
+  a4 = _mm_shuffle_epi8(a4, (SUBSH_MASK[4]));\
+  a4 = _mm_aesenclast_si128(a4, b0);\
+  a5 = _mm_shuffle_epi8(a5, (SUBSH_MASK[5]));\
+  a5 = _mm_aesenclast_si128(a5, b0);\
+  a6 = _mm_shuffle_epi8(a6, (SUBSH_MASK[6]));\
+  a6 = _mm_aesenclast_si128(a6, b0);\
+  a7 = _mm_shuffle_epi8(a7, (SUBSH_MASK[7]));\
+  a7 = _mm_aesenclast_si128(a7, b0);\
+  \
+  /* MixBytes */\
+  MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
+\
+}
+
+/* 10 rounds, P and Q in parallel */
+#define ROUNDS_P_Q(){\
+  ROUND(0, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+  ROUND(1, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+  ROUND(2, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+  ROUND(3, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+  ROUND(4, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+  ROUND(5, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+  ROUND(6, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+  ROUND(7, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+  ROUND(8, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+  ROUND(9, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+}
+
+/* Matrix Transpose Step 1
+ * input is a 512-bit state with two columns in one xmm
+ * output is a 512-bit state with two rows in one xmm
+ * inputs: i0-i3
+ * outputs: i0, o1-o3
+ * clobbers: t0
+ */
+#define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
+  t0 = TRANSP_MASK;\
+  \
+  i0 = _mm_shuffle_epi8(i0, t0);\
+  i1 = _mm_shuffle_epi8(i1, t0);\
+  i2 = _mm_shuffle_epi8(i2, t0);\
+  i3 = _mm_shuffle_epi8(i3, t0);\
+  \
+  o1 = i0;\
+  t0 = i2;\
+  \
+  i0 = _mm_unpacklo_epi16(i0, i1);\
+  o1 = _mm_unpackhi_epi16(o1, i1);\
+  i2 = _mm_unpacklo_epi16(i2, i3);\
+  t0 = _mm_unpackhi_epi16(t0, i3);\
+  \
+  i0 = _mm_shuffle_epi32(i0, 216);\
+  o1 = _mm_shuffle_epi32(o1, 216);\
+  i2 = _mm_shuffle_epi32(i2, 216);\
+  t0 = _mm_shuffle_epi32(t0, 216);\
+  \
+  o2 = i0;\
+  o3 = o1;\
+  \
+  i0 = _mm_unpacklo_epi32(i0, i2);\
+  o1 = _mm_unpacklo_epi32(o1, t0);\
+  o2 = _mm_unpackhi_epi32(o2, i2);\
+  o3 = _mm_unpackhi_epi32(o3, t0);\
+}/**/
+
+/* Matrix Transpose Step 2
+ * input are two 512-bit states with two rows in one xmm
+ * output are two 512-bit states with one row of each state in one xmm
+ * inputs: i0-i3 = P, i4-i7 = Q
+ * outputs: (i0, o1-o7) = (P|Q)
+ * possible reassignments: (output reg = input reg)
+ * * i1 -> o3-7
+ * * i2 -> o5-7
+ * * i3 -> o7
+ * * i4 -> o3-7
+ * * i5 -> o6-7
+ */
+#define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
+  o1 = i0;\
+  o2 = i1;\
+  i0 = _mm_unpacklo_epi64(i0, i4);\
+  o1 = _mm_unpackhi_epi64(o1, i4);\
+  o3 = i1;\
+  o4 = i2;\
+  o2 = _mm_unpacklo_epi64(o2, i5);\
+  o3 = _mm_unpackhi_epi64(o3, i5);\
+  o5 = i2;\
+  o6 = i3;\
+  o4 = _mm_unpacklo_epi64(o4, i6);\
+  o5 = _mm_unpackhi_epi64(o5, i6);\
+  o7 = i3;\
+  o6 = _mm_unpacklo_epi64(o6, i7);\
+  o7 = _mm_unpackhi_epi64(o7, i7);\
+}/**/
+
+/* Matrix Transpose Inverse Step 2
+ * input are two 512-bit states with one row of each state in one xmm
+ * output are two 512-bit states with two rows in one xmm
+ * inputs: i0-i7 = (P|Q)
+ * outputs: (i0, i2, i4, i6) = P, (o0-o3) = Q
+ */
+#define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
+  o0 = i0;\
+  i0 = _mm_unpacklo_epi64(i0, i1);\
+  o0 = _mm_unpackhi_epi64(o0, i1);\
+  o1 = i2;\
+  i2 = _mm_unpacklo_epi64(i2, i3);\
+  o1 = _mm_unpackhi_epi64(o1, i3);\
+  o2 = i4;\
+  i4 = _mm_unpacklo_epi64(i4, i5);\
+  o2 = _mm_unpackhi_epi64(o2, i5);\
+  o3 = i6;\
+  i6 = _mm_unpacklo_epi64(i6, i7);\
+  o3 = _mm_unpackhi_epi64(o3, i7);\
+}/**/
+
+/* Matrix Transpose Output Step 2
+ * input is one 512-bit state with two rows in one xmm
+ * output is one 512-bit state with one row in the low 64-bits of one xmm
+ * inputs: i0,i2,i4,i6 = S
+ * outputs: (i0-7) = (0|S)
+ */
+#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
+  t0 = _mm_xor_si128(t0, t0);\
+  i1 = i0;\
+  i3 = i2;\
+  i5 = i4;\
+  i7 = i6;\
+  i0 = _mm_unpacklo_epi64(i0, t0);\
+  i1 = _mm_unpackhi_epi64(i1, t0);\
+  i2 = _mm_unpacklo_epi64(i2, t0);\
+  i3 = _mm_unpackhi_epi64(i3, t0);\
+  i4 = _mm_unpacklo_epi64(i4, t0);\
+  i5 = _mm_unpackhi_epi64(i5, t0);\
+  i6 = _mm_unpacklo_epi64(i6, t0);\
+  i7 = _mm_unpackhi_epi64(i7, t0);\
+}/**/
+
+/* Matrix Transpose Output Inverse Step 2
+ * input is one 512-bit state with one row in the low 64-bits of one xmm
+ * output is one 512-bit state with two rows in one xmm
+ * inputs: i0-i7 = (0|S)
+ * outputs: (i0, i2, i4, i6) = S
+ */
+#define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
+  i0 = _mm_unpacklo_epi64(i0, i1);\
+  i2 = _mm_unpacklo_epi64(i2, i3);\
+  i4 = _mm_unpacklo_epi64(i4, i5);\
+  i6 = _mm_unpacklo_epi64(i6, i7);\
+}/**/
+
+
+void INIT256( __m128i* chaining )
+{
+  static __m128i xmm0, /*xmm1,*/ xmm2, /*xmm3, xmm4, xmm5,*/ xmm6, xmm7;
+  static __m128i /*xmm8, xmm9, xmm10, xmm11,*/ xmm12, xmm13, xmm14, xmm15;
+
+  /* load IV into registers xmm12 - xmm15 */
+  xmm12 = chaining[0];
+  xmm13 = chaining[1];
+  xmm14 = chaining[2];
+  xmm15 = chaining[3];
+
+  /* transform chaining value from column ordering into row ordering */
+  /* we put two rows (64 bit) of the IV into one 128-bit XMM register */
+  Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
+
+  /* store transposed IV */
+  chaining[0] = xmm12;
+  chaining[1] = xmm2;
+  chaining[2] = xmm6;
+  chaining[3] = xmm7;
+}
+
+void TF512( __m128i* chaining, __m128i* message )
+{
+  static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  static __m128i TEMP0;
+  static __m128i TEMP1;
+  static __m128i TEMP2;
+
+#ifdef IACA_TRACE
+  IACA_START;
+#endif
+
+  /* load message into registers xmm12 - xmm15 */
+  xmm12 = message[0];
+  xmm13 = message[1];
+  xmm14 = message[2];
+  xmm15 = message[3];
+
+  /* transform message M from column ordering into row ordering */
+  /* we first put two rows (64 bit) of the message into one 128-bit xmm register */
+  Matrix_Transpose_A(xmm12, xmm13, xmm14, xmm15, xmm2, xmm6, xmm7, xmm0);
+
+  /* load previous chaining value */
+  /* we first put two rows (64 bit) of the CV into one 128-bit xmm register */
+  xmm8 = chaining[0];
+  xmm0 = chaining[1];
+  xmm4 = chaining[2];
+  xmm5 = chaining[3];
+
+  /* xor message to CV get input of P */
+  /* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
+  xmm8 = _mm_xor_si128(xmm8, xmm12);
+  xmm0 = _mm_xor_si128(xmm0, xmm2);
+  xmm4 = _mm_xor_si128(xmm4, xmm6);
+  xmm5 = _mm_xor_si128(xmm5, xmm7);
+
+  /* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
+  /* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
+  /* result: the 8 rows of P and Q in xmm8 - xmm12 */
+  Matrix_Transpose_B(xmm8, xmm0, xmm4, xmm5, xmm12, xmm2, xmm6, xmm7, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);
+
+  /* compute the two permutations P and Q in parallel */
+  ROUNDS_P_Q();
+
+  /* unpack again to get two rows of P or two rows of Q in one xmm register */
+  Matrix_Transpose_B_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3);
+
+  /* xor output of P and Q */
+  /* result: P(CV+M)+Q(M) in xmm0...xmm3 */
+  xmm0 = _mm_xor_si128(xmm0, xmm8);
+  xmm1 = _mm_xor_si128(xmm1, xmm10);
+  xmm2 = _mm_xor_si128(xmm2, xmm12);
+  xmm3 = _mm_xor_si128(xmm3, xmm14);
+
+  /* xor CV (feed-forward) */
+  /* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
+  xmm0 = _mm_xor_si128(xmm0, (chaining[0]));
+  xmm1 = _mm_xor_si128(xmm1, (chaining[1]));
+  xmm2 = _mm_xor_si128(xmm2, (chaining[2]));
+  xmm3 = _mm_xor_si128(xmm3, (chaining[3]));
+
+  /* store CV */
+  chaining[0] = xmm0;
+  chaining[1] = xmm1;
+  chaining[2] = xmm2;
+  chaining[3] = xmm3;
+
+#ifdef IACA_TRACE
+  IACA_END;
+#endif
+  return;
+}
+
+void OF512( __m128i* chaining )
+{
+  static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  static __m128i TEMP0;
+  static __m128i TEMP1;
+  static __m128i TEMP2;
+
+  /* load CV into registers xmm8, xmm10, xmm12, xmm14 */
+  xmm8 = chaining[0];
+  xmm10 = chaining[1];
+  xmm12 = chaining[2];
+  xmm14 = chaining[3];
+
+  /* there are now 2 rows of the CV in one xmm register */
+  /* unpack to get 1 row of P (64 bit) into one half of an xmm register */
+  /* result: the 8 input rows of P in xmm8 - xmm15 */
+  Matrix_Transpose_O_B(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0);
+
+  /* compute the permutation P */
+  /* result: the output of P(CV) in xmm8 - xmm15 */
+  ROUNDS_P_Q();
+
+  /* unpack again to get two rows of P in one xmm register */
+  /* result: P(CV) in xmm8, xmm10, xmm12, xmm14 */
+  Matrix_Transpose_O_B_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);
+
+  /* xor CV to P output (feed-forward) */
+  /* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
+  xmm8 = _mm_xor_si128(xmm8,  (chaining[0]));
+  xmm10 = _mm_xor_si128(xmm10, (chaining[1]));
+  xmm12 = _mm_xor_si128(xmm12, (chaining[2]));
+  xmm14 = _mm_xor_si128(xmm14, (chaining[3]));
+
+  /* transform state back from row ordering into column ordering */
+  /* result: final hash value in xmm9, xmm11 */
+  Matrix_Transpose_A(xmm8, xmm10, xmm12, xmm14, xmm4, xmm9, xmm11, xmm0);
+
+  /* we only need to return the truncated half of the state */
+  chaining[2] = xmm9;
+  chaining[3] = xmm11;
+}
+
+
--- a/algo/groestl/groestl512-hash-4way.c
+++ b/algo/groestl/groestl512-hash-4way.c
@@ -0,0 +1,114 @@
+/* hash.c     Aug 2011
+ * groestl512-hash-4way https://github.com/JayDDee/cpuminer-opt  2019-12.
+ *
+ * Groestl implementation for different versions.
+ * Author: Krystian Matusiewicz, Günther A. Roland, Martin Schläffer
+ *
+ * This code is placed in the public domain
+ */
+
+// Optimized for hash and data length that are integrals of __m128i 
+
+
+#include <memory.h>
+#include "groestl512-intr-4way.h"
+#include "miner.h"
+#include "simd-utils.h"
+
+#if defined(__VAES__)
+
+#define ROTL64(a,n) \
+   ( ( ( (a)<<(n) ) | ( (a) >> (64-(n)) ) ) & 0xffffffffffffffff )
+     
+#define U64BIG(a) \
+  ( ( ROTL64(a, 8) & 0x000000FF000000FF ) | \
+    ( ROTL64(a,24) & 0x0000FF000000FF00 ) | \
+    ( ROTL64(a,40) & 0x00FF000000FF0000 ) | \
+    ( ROTL64(a,56) & 0xFF000000FF000000 ) )
+
+int groestl512_4way_init( groestl512_4way_context* ctx, uint64_t hashlen )
+{
+  int i;
+
+  ctx->hashlen = hashlen;
+  SET_CONSTANTS();
+
+  if (ctx->chaining == NULL || ctx->buffer == NULL)
+    return 1;
+
+  for ( i = 0; i < SIZE512; i++ )
+  {
+     ctx->chaining[i] = m512_zero;
+     ctx->buffer[i]   = m512_zero;
+  }
+
+  uint64_t len = U64BIG((uint64_t)LENGTH);
+  ctx->chaining[ COLS/2 -1 ] = _mm512_set4_epi64( len, 0, len, 0 );
+  INIT_4way(ctx->chaining);
+  ctx->buf_ptr = 0;
+  ctx->rem_ptr = 0;
+
+  return 0;
+}
+
+int groestl512_4way_update_close( groestl512_4way_context* ctx, void* output,
+                                const void* input, uint64_t databitlen )
+{
+   const int len = (int)databitlen / 128;
+   const int hashlen_m128i = ctx->hashlen / 16;   // bytes to __m128i
+   const int hash_offset = SIZE512 - hashlen_m128i;
+   int rem = ctx->rem_ptr;
+   int blocks = len / SIZE512;
+   __m512i* in = (__m512i*)input;
+   int i;
+
+   // --- update ---
+
+   // digest any full blocks, process directly from input 
+   for ( i = 0; i < blocks; i++ )
+      TF1024_4way( ctx->chaining, &in[ i * SIZE512 ] );
+   ctx->buf_ptr = blocks * SIZE512;
+
+   // copy any remaining data to buffer, it may already contain data
+   // from a previous update for a midstate precalc
+   for ( i = 0; i < len % SIZE512; i++ )
+       ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
+   i += rem;    // use i as rem_ptr in final
+
+   //--- final ---
+
+   blocks++;      // adjust for final block
+
+   if ( i == SIZE512 - 1 )
+   {        
+       // only 1 vector left in buffer, all padding at once
+       ctx->buffer[i] = m512_const1_128( _mm_set_epi8(
+                      blocks, blocks>>8,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80 ) );
+   }   
+   else
+   {
+       // add first padding
+       ctx->buffer[i] = m512_const4_64( 0, 0x80, 0, 0x80 );
+       // add zero padding
+       for ( i += 1; i < SIZE512 - 1; i++ )
+           ctx->buffer[i] = m512_zero;
+
+       // add length padding, second last byte is zero unless blocks > 255
+       ctx->buffer[i] = m512_const1_128( _mm_set_epi8(
+                   blocks, blocks>>8, 0,0, 0,0, 0,0, 0,0, 0,0, 0,0, 0,0 ) );
+   }
+
+// digest final padding block and do output transform
+   TF1024_4way( ctx->chaining, ctx->buffer );
+
+   OF1024_4way( ctx->chaining );
+
+   // store hash result in output 
+   for ( i = 0; i < hashlen_m128i; i++ )
+      casti_m512i( output, i ) = ctx->chaining[ hash_offset + i ];
+
+   return 0;
+}
+
+#endif   // VAES
+
--- a/algo/groestl/groestl512-hash-4way.h
+++ b/algo/groestl/groestl512-hash-4way.h
@@ -0,0 +1,94 @@
+/* hash.h     Aug 2011
+ *
+ * Groestl implementation for different versions.
+ * Author: Krystian Matusiewicz, Günther A. Roland, Martin Schläffer
+ *
+ * This code is placed in the public domain
+ */
+
+#if !defined(GROESTL512_HASH_4WAY_H__)
+#define GROESTL512_HASH_4WAY_H__ 1
+
+#include "simd-utils.h"
+#include <immintrin.h>
+#include <stdint.h>
+#include <stdio.h>
+#if defined(_WIN64) || defined(__WINDOWS__)
+#include <windows.h>
+#endif
+#include <stdlib.h>
+
+#define LENGTH (512)
+
+//#include "brg_endian.h"
+//#define NEED_UINT_64T
+//#include "algo/sha/brg_types.h"
+
+/* some sizes (number of bytes) */
+#define ROWS (8)
+#define LENGTHFIELDLEN (ROWS)
+//#define COLS512 (8)
+#define COLS1024 (16)
+//#define SIZE512 ((ROWS)*(COLS512))
+#define SIZE_1024 ((ROWS)*(COLS1024))
+//#define ROUNDS512 (10)
+#define ROUNDS1024 (14)
+
+//#if LENGTH<=256
+//#define COLS (COLS512)
+//#define SIZE (SIZE512)
+//#define ROUNDS (ROUNDS512)
+//#else
+#define COLS (COLS1024)
+//#define SIZE (SIZE1024)
+#define ROUNDS (ROUNDS1024)
+//#endif
+
+/*
+#define ROTL64(a,n) ((((a)<<(n))|((a)>>(64-(n))))&li_64(ffffffffffffffff))
+
+#if (PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN)
+#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*(7-(n)))))
+#define U64BIG(a) (a)
+#endif // IS_BIG_ENDIAN 
+
+#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
+#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*n)))
+#define U64BIG(a) \
+  ((ROTL64(a, 8) & li_64(000000FF000000FF)) | \
+   (ROTL64(a,24) & li_64(0000FF000000FF00)) | \
+   (ROTL64(a,40) & li_64(00FF000000FF0000)) | \
+   (ROTL64(a,56) & li_64(FF000000FF000000)))
+#endif // IS_LITTLE_ENDIAN 
+
+typedef unsigned char BitSequence_gr;
+typedef unsigned long long DataLength_gr;
+typedef enum { SUCCESS_GR = 0, FAIL_GR = 1, BAD_HASHBITLEN_GR = 2} HashReturn_gr;
+*/
+
+#define SIZE512 (SIZE_1024/16)
+
+typedef struct {
+  __attribute__ ((aligned (128))) __m512i chaining[SIZE512];
+  __attribute__ ((aligned (64))) __m512i buffer[SIZE512];
+  int hashlen;       // byte
+  int blk_count;     // SIZE_m128i
+  int buf_ptr;       // __m128i offset
+  int rem_ptr;
+  int databitlen;    // bits
+} groestl512_4way_context;
+
+
+int groestl512_4way_init( groestl512_4way_context*, uint64_t );
+
+//int reinit_groestl( hashState_groestl* );
+
+int groestl512_4way_update( groestl512_4way_context*, const void*,
+                              uint64_t );
+
+int groestl512_4way_close( groestl512_4way_context*, void* );
+
+int groestl512_4way_update_close( groestl512_4way_context*,  void*,
+                                        const void*, uint64_t );
+
+#endif /* __hash_h */
--- a/algo/groestl/groestl512-intr-4way.h
+++ b/algo/groestl/groestl512-intr-4way.h
@@ -0,0 +1,654 @@
+/* groestl-intr-aes.h     Aug 2011
+ *
+ * Groestl implementation with intrinsics using ssse3, sse4.1, and aes
+ * instructions.
+ * Author: Günther A. Roland, Martin Schläffer, Krystian Matusiewicz
+ *
+ * This code is placed in the public domain
+ */
+
+
+#if !defined(GROESTL512_INTR_4WAY_H__)
+#define GROESTL512_INTR_4WAY_H__ 1
+      
+#include "groestl512-hash-4way.h"
+
+#if defined(__VAES__)
+
+/* global constants  */
+__m512i ROUND_CONST_Lx;
+//__m128i ROUND_CONST_L0[ROUNDS512];
+//__m128i ROUND_CONST_L7[ROUNDS512];
+__m512i ROUND_CONST_P[ROUNDS1024];
+__m512i ROUND_CONST_Q[ROUNDS1024];
+__m512i TRANSP_MASK;
+__m512i SUBSH_MASK[8];
+__m512i ALL_1B;
+__m512i ALL_FF;
+
+#define tos(a)    #a
+#define tostr(a)  tos(a)
+
+/* xmm[i] will be multiplied by 2
+ * xmm[j] will be lost
+ * xmm[k] has to be all 0x1b */
+#define MUL2(i, j, k){\
+  j = _mm512_xor_si512(j, j);\
+  j = _mm512_movm_epi8( _mm512_cmpgt_epi8_mask(j, i) );\
+  i = _mm512_add_epi8(i, i);\
+  j = _mm512_and_si512(j, k);\
+  i = _mm512_xor_si512(i, j);\
+} 
+
+ /**/
+
+/* Yet another implementation of MixBytes.
+   This time we use the formulae (3) from the paper "Byte Slicing Groestl".
+   Input: a0, ..., a7
+   Output: b0, ..., b7 = MixBytes(a0,...,a7).
+   but we use the relations:
+   t_i = a_i + a_{i+3}
+   x_i = t_i + t_{i+3}
+   y_i = t_i + t+{i+2} + a_{i+6}
+   z_i = 2*x_i
+   w_i = z_i + y_{i+4}
+   v_i = 2*w_i
+   b_i = v_{i+3} + y_{i+4}
+   We keep building b_i in registers xmm8..xmm15 by first building y_{i+4} there
+   and then adding v_i computed in the meantime in registers xmm0..xmm7.
+   We almost fit into 16 registers, need only 3 spills to memory.
+   This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
+   K. Matusiewicz, 2011/05/29 */
+#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
+  /* t_i = a_i + a_{i+1} */\
+  b6 = a0;\
+  b7 = a1;\
+  a0 = _mm512_xor_si512(a0, a1);\
+  b0 = a2;\
+  a1 = _mm512_xor_si512(a1, a2);\
+  b1 = a3;\
+  a2 = _mm512_xor_si512(a2, a3);\
+  b2 = a4;\
+  a3 = _mm512_xor_si512(a3, a4);\
+  b3 = a5;\
+  a4 = _mm512_xor_si512(a4, a5);\
+  b4 = a6;\
+  a5 = _mm512_xor_si512(a5, a6);\
+  b5 = a7;\
+  a6 = _mm512_xor_si512(a6, a7);\
+  a7 = _mm512_xor_si512(a7, b6);\
+  \
+  /* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
+  b0 = _mm512_xor_si512(b0, a4);\
+  b6 = _mm512_xor_si512(b6, a4);\
+  b1 = _mm512_xor_si512(b1, a5);\
+  b7 = _mm512_xor_si512(b7, a5);\
+  b2 = _mm512_xor_si512(b2, a6);\
+  b0 = _mm512_xor_si512(b0, a6);\
+  /* spill values y_4, y_5 to memory */\
+  TEMP0 = b0;\
+  b3 = _mm512_xor_si512(b3, a7);\
+  b1 = _mm512_xor_si512(b1, a7);\
+  TEMP1 = b1;\
+  b4 = _mm512_xor_si512(b4, a0);\
+  b2 = _mm512_xor_si512(b2, a0);\
+  /* save values t0, t1, t2 to xmm8, xmm9 and memory */\
+  b0 = a0;\
+  b5 = _mm512_xor_si512(b5, a1);\
+  b3 = _mm512_xor_si512(b3, a1);\
+  b1 = a1;\
+  b6 = _mm512_xor_si512(b6, a2);\
+  b4 = _mm512_xor_si512(b4, a2);\
+  TEMP2 = a2;\
+  b7 = _mm512_xor_si512(b7, a3);\
+  b5 = _mm512_xor_si512(b5, a3);\
+  \
+  /* compute x_i = t_i + t_{i+3} */\
+  a0 = _mm512_xor_si512(a0, a3);\
+  a1 = _mm512_xor_si512(a1, a4);\
+  a2 = _mm512_xor_si512(a2, a5);\
+  a3 = _mm512_xor_si512(a3, a6);\
+  a4 = _mm512_xor_si512(a4, a7);\
+  a5 = _mm512_xor_si512(a5, b0);\
+  a6 = _mm512_xor_si512(a6, b1);\
+  a7 = _mm512_xor_si512(a7, TEMP2);\
+  \
+  /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
+  /* compute w_i : add y_{i+4} */\
+  b1 = ALL_1B;\
+  MUL2(a0, b0, b1);\
+  a0 = _mm512_xor_si512(a0, TEMP0);\
+  MUL2(a1, b0, b1);\
+  a1 = _mm512_xor_si512(a1, TEMP1);\
+  MUL2(a2, b0, b1);\
+  a2 = _mm512_xor_si512(a2, b2);\
+  MUL2(a3, b0, b1);\
+  a3 = _mm512_xor_si512(a3, b3);\
+  MUL2(a4, b0, b1);\
+  a4 = _mm512_xor_si512(a4, b4);\
+  MUL2(a5, b0, b1);\
+  a5 = _mm512_xor_si512(a5, b5);\
+  MUL2(a6, b0, b1);\
+  a6 = _mm512_xor_si512(a6, b6);\
+  MUL2(a7, b0, b1);\
+  a7 = _mm512_xor_si512(a7, b7);\
+  \
+  /* compute v_i : double w_i      */\
+  /* add to y_4 y_5 .. v3, v4, ... */\
+  MUL2(a0, b0, b1);\
+  b5 = _mm512_xor_si512(b5, a0);\
+  MUL2(a1, b0, b1);\
+  b6 = _mm512_xor_si512(b6, a1);\
+  MUL2(a2, b0, b1);\
+  b7 = _mm512_xor_si512(b7, a2);\
+  MUL2(a5, b0, b1);\
+  b2 = _mm512_xor_si512(b2, a5);\
+  MUL2(a6, b0, b1);\
+  b3 = _mm512_xor_si512(b3, a6);\
+  MUL2(a7, b0, b1);\
+  b4 = _mm512_xor_si512(b4, a7);\
+  MUL2(a3, b0, b1);\
+  MUL2(a4, b0, b1);\
+  b0 = TEMP0;\
+  b1 = TEMP1;\
+  b0 = _mm512_xor_si512(b0, a3);\
+  b1 = _mm512_xor_si512(b1, a4);\
+}/*MixBytes*/
+
+// calculate the round constants seperately and load at startup
+
+#define SET_CONSTANTS(){\
+  ALL_FF = _mm512_set1_epi32( 0xffffffff );\
+  ALL_1B = _mm512_set1_epi32( 0x1b1b1b1b );\
+  TRANSP_MASK   = _mm512_set_epi32( \
+                         0x3f373b33, 0x3e363a32, 0x3d353931, 0x3c343830, \
+                         0x2f272b23, 0x2e262a22, 0x2d252921, 0x2c242820, \
+                         0x1f171b13, 0x1e161a12, 0x1d151911, 0x1c141810, \
+                         0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800 ); \
+  SUBSH_MASK[0] = _mm512_set_epi32( \
+                         0x3336393c, 0x3f323538, 0x3b3e3134, 0x373a3d30, \
+                         0x2326292c, 0x2f222528, 0x2b2e2124, 0x272a2d20, \
+                         0x1316191c, 0x1f121518, 0x1b1e1114, 0x171a1d10, \
+                         0x0306090c, 0x0f020508, 0x0b0e0104, 0x070a0d00 ); \
+  SUBSH_MASK[1] = _mm512_set_epi32( \
+                         0x34373a3d, 0x30333639, 0x3c3f3235, 0x383b3e31, \
+                         0x24272a2d, 0x20232629, 0x2c2f2225, 0x282b2e21, \
+                         0x14171a1d, 0x10131619, 0x1c1f1215, 0x181b1e11, \
+                         0x04070a0d, 0x00030609, 0x0c0f0205, 0x080b0e01 ); \
+  SUBSH_MASK[2] = _mm512_set_epi32( \
+                         0x35383b3e, 0x3134373a, 0x3d303336, 0x393c3f32, \
+                         0x25282b2e, 0x2124272a, 0x2d202326, 0x292c2f22, \
+                         0x15181b1e, 0x1114171a, 0x1d101316, 0x191c1f12, \
+                         0x05080b0e, 0x0104070a, 0x0d000306, 0x090c0f02 ); \
+  SUBSH_MASK[3] = _mm512_set_epi32( \
+                         0x36393c3f, 0x3235383b, 0x3e313437, 0x3a3d3033, \
+                         0x26292c2f, 0x2225282b, 0x2e212427, 0x2a2d2023, \
+                         0x16191c1f, 0x1215181b, 0x1e111417, 0x1a1d1013, \
+                         0x06090c0f, 0x0205080b, 0x0e010407, 0x0a0d0003 ); \
+  SUBSH_MASK[4] = _mm512_set_epi32( \
+                         0x373a3d30, 0x3336393c, 0x3f323538, 0x3b3e3134, \
+                         0x272a2d20, 0x2326292c, 0x2f222528, 0x2b2e2124, \
+                         0x171a1d10, 0x1316191c, 0x1f121518, 0x1b1e1114, \
+                         0x070a0d00, 0x0306090c, 0x0f020508, 0x0b0e0104 ); \
+  SUBSH_MASK[5] = _mm512_set_epi32( \
+                         0x383b3e31, 0x34373a3d, 0x30333639, 0x3c3f3235, \
+                         0x282b2e21, 0x24272a2d, 0x20232629, 0x2c2f2225, \
+                         0x181b1e11, 0x14171a1d, 0x10131619, 0x1c1f1215, \
+                         0x080b0e01, 0x04070a0d, 0x00030609, 0x0c0f0205 ); \
+  SUBSH_MASK[6] = _mm512_set_epi32( \
+                         0x393c3f32, 0x35383b3e, 0x3134373a, 0x3d303336, \
+                         0x292c2f22, 0x25282b2e, 0x2124272a, 0x2d202326, \
+                         0x191c1f12, 0x15181b1e, 0x1114171a, 0x1d101316, \
+                         0x090c0f02, 0x05080b0e, 0x0104070a, 0x0d000306 ); \
+  SUBSH_MASK[7] = _mm512_set_epi32( \
+                         0x3e313437, 0x3a3d3033, 0x36393c3f, 0x3235383b, \
+                         0x2e212427, 0x2a2d2023, 0x26292c2f, 0x2225282b, \
+                         0x1e111417, 0x1a1d1013, 0x16191c1f, 0x1215181b, \
+                         0x0e010407, 0x0a0d0003, 0x06090c0f, 0x0205080b ); \
+  for( i = 0; i < ROUNDS1024; i++ ) \
+  { \
+    ROUND_CONST_P[i] = _mm512_set4_epi32( 0xf0e0d0c0 ^ (i * 0x01010101), \
+                                          0xb0a09080 ^ (i * 0x01010101), \
+                                          0x70605040 ^ (i * 0x01010101), \
+                                          0x30201000 ^ (i * 0x01010101) ); \
+    ROUND_CONST_Q[i] = _mm512_set4_epi32( 0x0f1f2f3f ^ (i * 0x01010101), \
+                                          0x4f5f6f7f ^ (i * 0x01010101), \
+                                          0x8f9fafbf ^ (i * 0x01010101), \
+                                          0xcfdfefff ^ (i * 0x01010101));\
+  } \
+}while(0);\
+
+/* one round
+ * a0-a7 = input rows
+ * b0-b7 = output rows
+ */
+#define SUBMIX(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
+  /* SubBytes */\
+  b0 = _mm512_xor_si512( b0, b0 );\
+  a0 = _mm512_aesenclast_epi128( a0, b0 );\
+  a1 = _mm512_aesenclast_epi128( a1, b0 );\
+  a2 = _mm512_aesenclast_epi128( a2, b0 );\
+  a3 = _mm512_aesenclast_epi128( a3, b0 );\
+  a4 = _mm512_aesenclast_epi128( a4, b0 );\
+  a5 = _mm512_aesenclast_epi128( a5, b0 );\
+  a6 = _mm512_aesenclast_epi128( a6, b0 );\
+  a7 = _mm512_aesenclast_epi128( a7, b0 );\
+  /* MixBytes */\
+  MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
+}
+
+#define ROUNDS_P(){\
+  uint8_t round_counter = 0;\
+  for ( round_counter = 0; round_counter < 14; round_counter += 2 ) \
+  { \
+    /* AddRoundConstant P1024 */\
+    xmm8 = _mm512_xor_si512( xmm8, ( ROUND_CONST_P[ round_counter ] ) );\
+    /* ShiftBytes P1024 + pre-AESENCLAST */\
+    xmm8  = _mm512_shuffle_epi8( xmm8,  ( SUBSH_MASK[0] ) );\
+    xmm9  = _mm512_shuffle_epi8( xmm9,  ( SUBSH_MASK[1] ) );\
+    xmm10 = _mm512_shuffle_epi8( xmm10, ( SUBSH_MASK[2] ) );\
+    xmm11 = _mm512_shuffle_epi8( xmm11, ( SUBSH_MASK[3] ) );\
+    xmm12 = _mm512_shuffle_epi8( xmm12, ( SUBSH_MASK[4] ) );\
+    xmm13 = _mm512_shuffle_epi8( xmm13, ( SUBSH_MASK[5] ) );\
+    xmm14 = _mm512_shuffle_epi8( xmm14, ( SUBSH_MASK[6] ) );\
+    xmm15 = _mm512_shuffle_epi8( xmm15, ( SUBSH_MASK[7] ) );\
+    /* SubBytes + MixBytes */\
+    SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+    \
+     /* AddRoundConstant P1024 */\
+    xmm0 = _mm512_xor_si512( xmm0, ( ROUND_CONST_P[ round_counter+1 ] ) );\
+    /* ShiftBytes P1024 + pre-AESENCLAST */\
+    xmm0 = _mm512_shuffle_epi8( xmm0, ( SUBSH_MASK[0] ) );\
+    xmm1 = _mm512_shuffle_epi8( xmm1, ( SUBSH_MASK[1] ) );\
+    xmm2 = _mm512_shuffle_epi8( xmm2, ( SUBSH_MASK[2] ) );\
+    xmm3 = _mm512_shuffle_epi8( xmm3, ( SUBSH_MASK[3] ) );\
+    xmm4 = _mm512_shuffle_epi8( xmm4, ( SUBSH_MASK[4] ) );\
+    xmm5 = _mm512_shuffle_epi8( xmm5, ( SUBSH_MASK[5] ) );\
+    xmm6 = _mm512_shuffle_epi8( xmm6, ( SUBSH_MASK[6] ) );\
+    xmm7 = _mm512_shuffle_epi8( xmm7, ( SUBSH_MASK[7] ) );\
+    /* SubBytes + MixBytes */\
+     SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+  }\
+}
+
+#define ROUNDS_Q(){\
+  uint8_t round_counter = 0;\
+  for ( round_counter = 0; round_counter < 14; round_counter += 2) \
+  { \
+    /* AddRoundConstant Q1024 */\
+    xmm1 = ALL_FF;\
+    xmm8  = _mm512_xor_si512( xmm8,  xmm1 );\
+    xmm9  = _mm512_xor_si512( xmm9,  xmm1 );\
+    xmm10 = _mm512_xor_si512( xmm10, xmm1 );\
+    xmm11 = _mm512_xor_si512( xmm11, xmm1 );\
+    xmm12 = _mm512_xor_si512( xmm12, xmm1 );\
+    xmm13 = _mm512_xor_si512( xmm13, xmm1 );\
+    xmm14 = _mm512_xor_si512( xmm14, xmm1 );\
+    xmm15 = _mm512_xor_si512( xmm15, ( ROUND_CONST_Q[ round_counter ] ) );\
+    /* ShiftBytes Q1024 + pre-AESENCLAST */\
+    xmm8  = _mm512_shuffle_epi8( xmm8,  ( SUBSH_MASK[1] ) );\
+    xmm9  = _mm512_shuffle_epi8( xmm9,  ( SUBSH_MASK[3] ) );\
+    xmm10 = _mm512_shuffle_epi8( xmm10, ( SUBSH_MASK[5] ) );\
+    xmm11 = _mm512_shuffle_epi8( xmm11, ( SUBSH_MASK[7] ) );\
+    xmm12 = _mm512_shuffle_epi8( xmm12, ( SUBSH_MASK[0] ) );\
+    xmm13 = _mm512_shuffle_epi8( xmm13, ( SUBSH_MASK[2] ) );\
+    xmm14 = _mm512_shuffle_epi8( xmm14, ( SUBSH_MASK[4] ) );\
+    xmm15 = _mm512_shuffle_epi8( xmm15, ( SUBSH_MASK[6] ) );\
+    /* SubBytes + MixBytes */\
+    SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
+    \
+    /* AddRoundConstant Q1024 */\
+    xmm9 = ALL_FF;\
+    xmm0 = _mm512_xor_si512( xmm0, xmm9 );\
+    xmm1 = _mm512_xor_si512( xmm1, xmm9 );\
+    xmm2 = _mm512_xor_si512( xmm2, xmm9 );\
+    xmm3 = _mm512_xor_si512( xmm3, xmm9 );\
+    xmm4 = _mm512_xor_si512( xmm4, xmm9 );\
+    xmm5 = _mm512_xor_si512( xmm5, xmm9 );\
+    xmm6 = _mm512_xor_si512( xmm6, xmm9 );\
+    xmm7 = _mm512_xor_si512( xmm7, ( ROUND_CONST_Q[ round_counter+1 ] ) );\
+    /* ShiftBytes Q1024 + pre-AESENCLAST */\
+    xmm0 = _mm512_shuffle_epi8( xmm0, ( SUBSH_MASK[1] ) );\
+    xmm1 = _mm512_shuffle_epi8( xmm1, ( SUBSH_MASK[3] ) );\
+    xmm2 = _mm512_shuffle_epi8( xmm2, ( SUBSH_MASK[5] ) );\
+    xmm3 = _mm512_shuffle_epi8( xmm3, ( SUBSH_MASK[7] ) );\
+    xmm4 = _mm512_shuffle_epi8( xmm4, ( SUBSH_MASK[0] ) );\
+    xmm5 = _mm512_shuffle_epi8( xmm5, ( SUBSH_MASK[2] ) );\
+    xmm6 = _mm512_shuffle_epi8( xmm6, ( SUBSH_MASK[4] ) );\
+    xmm7 = _mm512_shuffle_epi8( xmm7, ( SUBSH_MASK[6] ) );\
+    /* SubBytes + MixBytes */\
+    SUBMIX(xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15);\
+  }\
+}
+
+/* Matrix Transpose
+ * input is a 1024-bit state with two columns in one xmm
+ * output is a 1024-bit state with two rows in one xmm
+ * inputs: i0-i7
+ * outputs: i0-i7
+ * clobbers: t0-t7
+ */
+#define Matrix_Transpose(i0, i1, i2, i3, i4, i5, i6, i7, t0, t1, t2, t3, t4, t5, t6, t7){\
+  t0 = TRANSP_MASK;\
+\
+  i6 = _mm512_shuffle_epi8(i6, t0);\
+  i0 = _mm512_shuffle_epi8(i0, t0);\
+  i1 = _mm512_shuffle_epi8(i1, t0);\
+  i2 = _mm512_shuffle_epi8(i2, t0);\
+  i3 = _mm512_shuffle_epi8(i3, t0);\
+  t1 = i2;\
+  i4 = _mm512_shuffle_epi8(i4, t0);\
+  i5 = _mm512_shuffle_epi8(i5, t0);\
+  t2 = i4;\
+  t3 = i6;\
+  i7 = _mm512_shuffle_epi8(i7, t0);\
+\
+  /* continue with unpack using 4 temp registers */\
+  t0 = i0;\
+  t2 = _mm512_unpackhi_epi16(t2, i5);\
+  i4 = _mm512_unpacklo_epi16(i4, i5);\
+  t3 = _mm512_unpackhi_epi16(t3, i7);\
+  i6 = _mm512_unpacklo_epi16(i6, i7);\
+  t0 = _mm512_unpackhi_epi16(t0, i1);\
+  t1 = _mm512_unpackhi_epi16(t1, i3);\
+  i2 = _mm512_unpacklo_epi16(i2, i3);\
+  i0 = _mm512_unpacklo_epi16(i0, i1);\
+\
+  /* shuffle with immediate */\
+  t0 = _mm512_shuffle_epi32(t0, 216);\
+  t1 = _mm512_shuffle_epi32(t1, 216);\
+  t2 = _mm512_shuffle_epi32(t2, 216);\
+  t3 = _mm512_shuffle_epi32(t3, 216);\
+  i0 = _mm512_shuffle_epi32(i0, 216);\
+  i2 = _mm512_shuffle_epi32(i2, 216);\
+  i4 = _mm512_shuffle_epi32(i4, 216);\
+  i6 = _mm512_shuffle_epi32(i6, 216);\
+\
+  /* continue with unpack */\
+  t4 = i0;\
+  i0 = _mm512_unpacklo_epi32(i0, i2);\
+  t4 = _mm512_unpackhi_epi32(t4, i2);\
+  t5 = t0;\
+  t0 = _mm512_unpacklo_epi32(t0, t1);\
+  t5 = _mm512_unpackhi_epi32(t5, t1);\
+  t6 = i4;\
+  i4 = _mm512_unpacklo_epi32(i4, i6);\
+  t7 = t2;\
+  t6 = _mm512_unpackhi_epi32(t6, i6);\
+  i2 = t0;\
+  t2 = _mm512_unpacklo_epi32(t2, t3);\
+  i3 = t0;\
+  t7 = _mm512_unpackhi_epi32(t7, t3);\
+\
+  /* there are now 2 rows in each xmm */\
+  /* unpack to get 1 row of CV in each xmm */\
+  i1 = i0;\
+  i1 = _mm512_unpackhi_epi64(i1, i4);\
+  i0 = _mm512_unpacklo_epi64(i0, i4);\
+  i4 = t4;\
+  i3 = _mm512_unpackhi_epi64(i3, t2);\
+  i5 = t4;\
+  i2 = _mm512_unpacklo_epi64(i2, t2);\
+  i6 = t5;\
+  i5 = _mm512_unpackhi_epi64(i5, t6);\
+  i7 = t5;\
+  i4 = _mm512_unpacklo_epi64(i4, t6);\
+  i7 = _mm512_unpackhi_epi64(i7, t7);\
+  i6 = _mm512_unpacklo_epi64(i6, t7);\
+  /* transpose done */\
+}/**/
+
+/* Matrix Transpose Inverse
+ * input is a 1024-bit state with two rows in one xmm
+ * output is a 1024-bit state with two columns in one xmm
+ * inputs: i0-i7
+ * outputs: (i0, o0, i1, i3, o1, o2, i5, i7)
+ * clobbers: t0-t4
+ */
+#define Matrix_Transpose_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, t0, t1, t2, t3, t4){\
+  /*  transpose matrix to get output format */\
+  o1 = i0;\
+  i0 = _mm512_unpacklo_epi64(i0, i1);\
+  o1 = _mm512_unpackhi_epi64(o1, i1);\
+  t0 = i2;\
+  i2 = _mm512_unpacklo_epi64(i2, i3);\
+  t0 = _mm512_unpackhi_epi64(t0, i3);\
+  t1 = i4;\
+  i4 = _mm512_unpacklo_epi64(i4, i5);\
+  t1 = _mm512_unpackhi_epi64(t1, i5);\
+  t2 = i6;\
+  o0 = TRANSP_MASK;\
+  i6 = _mm512_unpacklo_epi64(i6, i7);\
+  t2 = _mm512_unpackhi_epi64(t2, i7);\
+  /* load transpose mask into a register, because it will be used 8 times */\
+  i0 = _mm512_shuffle_epi8(i0, o0);\
+  i2 = _mm512_shuffle_epi8(i2, o0);\
+  i4 = _mm512_shuffle_epi8(i4, o0);\
+  i6 = _mm512_shuffle_epi8(i6, o0);\
+  o1 = _mm512_shuffle_epi8(o1, o0);\
+  t0 = _mm512_shuffle_epi8(t0, o0);\
+  t1 = _mm512_shuffle_epi8(t1, o0);\
+  t2 = _mm512_shuffle_epi8(t2, o0);\
+  /* continue with unpack using 4 temp registers */\
+  t3 = i4;\
+  o2 = o1;\
+  o0 = i0;\
+  t4 = t1;\
+  \
+  t3 = _mm512_unpackhi_epi16(t3, i6);\
+  i4 = _mm512_unpacklo_epi16(i4, i6);\
+  o0 = _mm512_unpackhi_epi16(o0, i2);\
+  i0 = _mm512_unpacklo_epi16(i0, i2);\
+  o2 = _mm512_unpackhi_epi16(o2, t0);\
+  o1 = _mm512_unpacklo_epi16(o1, t0);\
+  t4 = _mm512_unpackhi_epi16(t4, t2);\
+  t1 = _mm512_unpacklo_epi16(t1, t2);\
+  /* shuffle with immediate */\
+  i4 = _mm512_shuffle_epi32(i4, 216);\
+  t3 = _mm512_shuffle_epi32(t3, 216);\
+  o1 = _mm512_shuffle_epi32(o1, 216);\
+  o2 = _mm512_shuffle_epi32(o2, 216);\
+  i0 = _mm512_shuffle_epi32(i0, 216);\
+  o0 = _mm512_shuffle_epi32(o0, 216);\
+  t1 = _mm512_shuffle_epi32(t1, 216);\
+  t4 = _mm512_shuffle_epi32(t4, 216);\
+  /* continue with unpack */\
+  i1 = i0;\
+  i3 = o0;\
+  i5 = o1;\
+  i7 = o2;\
+  i0 = _mm512_unpacklo_epi32(i0, i4);\
+  i1 = _mm512_unpackhi_epi32(i1, i4);\
+  o0 = _mm512_unpacklo_epi32(o0, t3);\
+  i3 = _mm512_unpackhi_epi32(i3, t3);\
+  o1 = _mm512_unpacklo_epi32(o1, t1);\
+  i5 = _mm512_unpackhi_epi32(i5, t1);\
+  o2 = _mm512_unpacklo_epi32(o2, t4);\
+  i7 = _mm512_unpackhi_epi32(i7, t4);\
+  /* transpose done */\
+}/**/
+
+
+void INIT_4way( __m512i* chaining )
+{
+  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+
+  /* load IV into registers xmm8 - xmm15 */
+  xmm8 = chaining[0];
+  xmm9 = chaining[1];
+  xmm10 = chaining[2];
+  xmm11 = chaining[3];
+  xmm12 = chaining[4];
+  xmm13 = chaining[5];
+  xmm14 = chaining[6];
+  xmm15 = chaining[7];
+
+  /* transform chaining value from column ordering into row ordering */
+  Matrix_Transpose(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);
+
+  /* store transposed IV */
+  chaining[0] = xmm8;
+  chaining[1] = xmm9;
+  chaining[2] = xmm10;
+  chaining[3] = xmm11;
+  chaining[4] = xmm12;
+  chaining[5] = xmm13;
+  chaining[6] = xmm14;
+  chaining[7] = xmm15;
+}
+
+void TF1024_4way( __m512i* chaining, const __m512i* message )
+{
+  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  static __m512i QTEMP[8];
+  static __m512i TEMP0;
+  static __m512i TEMP1;
+  static __m512i TEMP2;
+
+  /* load message into registers xmm8 - xmm15 (Q = message) */
+  xmm8 = message[0];
+  xmm9 = message[1];
+  xmm10 = message[2];
+  xmm11 = message[3];
+  xmm12 = message[4];
+  xmm13 = message[5];
+  xmm14 = message[6];
+  xmm15 = message[7];
+
+  /* transform message M from column ordering into row ordering */
+  Matrix_Transpose(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);
+
+  /* store message M (Q input) for later */
+  QTEMP[0] = xmm8;
+  QTEMP[1] = xmm9;
+  QTEMP[2] = xmm10;
+  QTEMP[3] = xmm11;
+  QTEMP[4] = xmm12;
+  QTEMP[5] = xmm13;
+  QTEMP[6] = xmm14;
+  QTEMP[7] = xmm15;
+
+  /* xor CV to message to get P input */
+  /* result: CV+M in xmm8...xmm15 */
+  xmm8 = _mm512_xor_si512( xmm8,  (chaining[0]) );
+  xmm9 = _mm512_xor_si512( xmm9,  (chaining[1]) );
+  xmm10 = _mm512_xor_si512( xmm10, (chaining[2]) );
+  xmm11 = _mm512_xor_si512( xmm11, (chaining[3]) );
+  xmm12 = _mm512_xor_si512( xmm12, (chaining[4]) );
+  xmm13 = _mm512_xor_si512( xmm13, (chaining[5]) );
+  xmm14 = _mm512_xor_si512( xmm14, (chaining[6]) );
+  xmm15 = _mm512_xor_si512( xmm15, (chaining[7]) );
+
+  /* compute permutation P */
+  /* result: P(CV+M) in xmm8...xmm15 */
+  ROUNDS_P();
+
+  /* xor CV to P output (feed-forward) */
+  /* result: P(CV+M)+CV in xmm8...xmm15 */
+  xmm8 = _mm512_xor_si512( xmm8,  (chaining[0]) );
+  xmm9 = _mm512_xor_si512( xmm9,  (chaining[1]) );
+  xmm10 = _mm512_xor_si512( xmm10, (chaining[2]) );
+  xmm11 = _mm512_xor_si512( xmm11, (chaining[3]) );
+  xmm12 = _mm512_xor_si512( xmm12, (chaining[4]) );
+  xmm13 = _mm512_xor_si512( xmm13, (chaining[5]) );
+  xmm14 = _mm512_xor_si512( xmm14, (chaining[6]) );
+  xmm15 = _mm512_xor_si512( xmm15, (chaining[7]) );
+
+  /* store P(CV+M)+CV */
+  chaining[0] = xmm8;
+  chaining[1] = xmm9;
+  chaining[2] = xmm10;
+  chaining[3] = xmm11;
+  chaining[4] = xmm12;
+  chaining[5] = xmm13;
+  chaining[6] = xmm14;
+  chaining[7] = xmm15;
+
+  /* load message M (Q input) into xmm8-15 */
+  xmm8 = QTEMP[0];
+  xmm9 = QTEMP[1];
+  xmm10 = QTEMP[2];
+  xmm11 = QTEMP[3];
+  xmm12 = QTEMP[4];
+  xmm13 = QTEMP[5];
+  xmm14 = QTEMP[6];
+  xmm15 = QTEMP[7];
+
+  /* compute permutation Q */
+  /* result: Q(M) in xmm8...xmm15 */
+  ROUNDS_Q();
+
+  /* xor Q output */
+  /* result: P(CV+M)+CV+Q(M) in xmm8...xmm15 */
+  xmm8 = _mm512_xor_si512( xmm8,  (chaining[0]) );
+  xmm9 = _mm512_xor_si512( xmm9,  (chaining[1]) );
+  xmm10 = _mm512_xor_si512( xmm10, (chaining[2]) );
+  xmm11 = _mm512_xor_si512( xmm11, (chaining[3]) );
+  xmm12 = _mm512_xor_si512( xmm12, (chaining[4]) );
+  xmm13 = _mm512_xor_si512( xmm13, (chaining[5]) );
+  xmm14 = _mm512_xor_si512( xmm14, (chaining[6]) );
+  xmm15 = _mm512_xor_si512( xmm15, (chaining[7]) );
+
+  /* store CV */
+  chaining[0] = xmm8;
+  chaining[1] = xmm9;
+  chaining[2] = xmm10;
+  chaining[3] = xmm11;
+  chaining[4] = xmm12;
+  chaining[5] = xmm13;
+  chaining[6] = xmm14;
+  chaining[7] = xmm15;
+
+  return;
+}
+
+void OF1024_4way( __m512i* chaining )
+{
+  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  static __m512i TEMP0;
+  static __m512i TEMP1;
+  static __m512i TEMP2;
+
+  /* load CV into registers xmm8 - xmm15 */
+  xmm8 = chaining[0];
+  xmm9 = chaining[1];
+  xmm10 = chaining[2];
+  xmm11 = chaining[3];
+  xmm12 = chaining[4];
+  xmm13 = chaining[5];
+  xmm14 = chaining[6];
+  xmm15 = chaining[7];
+
+  /* compute permutation P */
+  /* result: P(CV) in xmm8...xmm15 */
+  ROUNDS_P();
+
+  /* xor CV to P output (feed-forward) */
+  /* result: P(CV)+CV in xmm8...xmm15 */
+  xmm8 = _mm512_xor_si512( xmm8,  (chaining[0]) );
+  xmm9 = _mm512_xor_si512( xmm9,  (chaining[1]) );
+  xmm10 = _mm512_xor_si512( xmm10, (chaining[2]) );
+  xmm11 = _mm512_xor_si512( xmm11, (chaining[3]) );
+  xmm12 = _mm512_xor_si512( xmm12, (chaining[4]) );
+  xmm13 = _mm512_xor_si512( xmm13, (chaining[5]) );
+  xmm14 = _mm512_xor_si512( xmm14, (chaining[6]) );
+  xmm15 = _mm512_xor_si512( xmm15, (chaining[7]) );
+
+  /* transpose CV back from row ordering to column ordering */
+  /* result: final hash value in xmm0, xmm6, xmm13, xmm15 */
+  Matrix_Transpose_INV(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm4, xmm0, xmm6, xmm1, xmm2, xmm3, xmm5, xmm7);
+
+  /* we only need to return the truncated half of the state */
+  chaining[4] = xmm0;
+  chaining[5] = xmm6;
+  chaining[6] = xmm13;
+  chaining[7] = xmm15;
+
+  return;
+}
+
+#endif  // VAES
+#endif  // GROESTL512_INTR_4WAY_H__
--- a/algo/groestl/myr-groestl.c
+++ b/algo/groestl/myr-groestl.c
@@ -10,7 +10,7 @@
 #else
  #include "aes_ni/hash-groestl.h"
 #endif
-#include "algo/sha/sph_sha2.h"
+#include <openssl/sha.h>

 typedef struct {
 #ifdef NO_AES_NI
@@ -18,7 +18,7 @@ typedef struct {
 #else
    hashState_groestl       groestl;
 #endif
-    sph_sha256_context sha;
+    SHA256_CTX              sha;
 } myrgr_ctx_holder;

 myrgr_ctx_holder myrgr_ctx;
@@ -28,15 +28,15 @@ void init_myrgr_ctx()
 #ifdef NO_AES_NI
     sph_groestl512_init( &myrgr_ctx.groestl );
 #else
-     init_groestl (&myrgr_ctx.groestl, 64 );
+     init_groestl ( &myrgr_ctx.groestl, 64 );
 #endif
-     sph_sha256_init(&myrgr_ctx.sha);
+     SHA256_Init( &myrgr_ctx.sha );
 }

 void myriad_hash(void *output, const void *input)
 {
-        myrgr_ctx_holder ctx;
-        memcpy( &ctx, &myrgr_ctx, sizeof(myrgr_ctx) );
+   myrgr_ctx_holder ctx;
+   memcpy( &ctx, &myrgr_ctx, sizeof(myrgr_ctx) );

 	uint32_t _ALIGN(32) hash[16];

@@ -44,23 +44,22 @@ void myriad_hash(void *output, const void *input)
 	sph_groestl512(&ctx.groestl, input, 80);
 	sph_groestl512_close(&ctx.groestl, hash);
 #else
-        update_groestl( &ctx.groestl, (char*)input, 640 );
-        final_groestl( &ctx.groestl, (char*)hash);
+   update_groestl( &ctx.groestl, (char*)input, 640 );
+   final_groestl( &ctx.groestl, (char*)hash);
 #endif

-	sph_sha256(&ctx.sha, hash, 64);
-	sph_sha256_close(&ctx.sha, hash);
+   SHA256_Update( &ctx.sha, (unsigned char*)hash, 64 );
+   SHA256_Final( (unsigned char*)hash, &ctx.sha );

 	memcpy(output, hash, 32);
 }

-int scanhash_myriad( struct work *work,
-	uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr)
+int scanhash_myriad( struct work *work, uint32_t max_nonce,
+                     uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
-
 	uint32_t _ALIGN(64) endiandata[20];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
 	const uint32_t first_nonce = pdata[19];
 	uint32_t nonce = first_nonce;
   int thr_id = mythr->id;  // thr_id arg is deprecated
@@ -89,15 +88,3 @@ int scanhash_myriad( struct work *work,
 	*hashes_done = pdata[19] - first_nonce + 1;
 	return 0;
 }
-/*
-bool register_myriad_algo( algo_gate_t* gate )
-{
-    gate->optimizations = SSE2_OPT | AES_OPT;
-    init_myrgr_ctx();
-    gate->scanhash = (void*)&scanhash_myriad;
-    gate->hash     = (void*)&myriadhash;
-//    gate->hash_alt = (void*)&myriadhash;
-    gate->get_max64 = (void*)&get_max64_0x3ffff;
-    return true;
-};
-*/
--- a/algo/groestl/myrgr-4way.c
+++ b/algo/groestl/myrgr-4way.c
@@ -1,14 +1,159 @@
 #include "myrgr-gate.h"
-
-#if defined(MYRGR_4WAY)
-
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
 #include <string.h>
-
 #include "aes_ni/hash-groestl.h"
-#include "algo/sha/sha2-hash-4way.h"
+#include "algo/sha/sha-hash-4way.h"
+#if defined(__VAES__)
+  #include "groestl512-hash-4way.h"
+#endif
+
+#if defined(MYRGR_8WAY)
+
+typedef struct {
+#if defined(__VAES__)
+   groestl512_4way_context groestl;
+#else
+   hashState_groestl       groestl;
+#endif
+   sha256_8way_context     sha;
+} myrgr_8way_ctx_holder;
+
+myrgr_8way_ctx_holder myrgr_8way_ctx;
+
+void init_myrgr_8way_ctx()
+{
+#if defined(__VAES__)
+     groestl512_4way_init( &myrgr_8way_ctx.groestl, 64 );
+#else
+     init_groestl( &myrgr_8way_ctx.groestl, 64 );
+#endif
+     sha256_8way_init( &myrgr_8way_ctx.sha );
+}
+
+void myriad_8way_hash( void *output, const void *input )
+{
+     uint32_t vhash[16*8] __attribute__ ((aligned (128)));
+     uint32_t vhashA[20*8] __attribute__ ((aligned (64)));
+     uint32_t vhashB[20*8] __attribute__ ((aligned (64)));
+     myrgr_8way_ctx_holder ctx;
+     memcpy( &ctx, &myrgr_8way_ctx, sizeof(myrgr_8way_ctx) );
+
+#if defined(__VAES__)
+
+     rintrlv_8x64_4x128( vhashA, vhashB, input, 640 );
+     groestl512_4way_update_close( &ctx.groestl, vhashA, vhashA, 640 );
+     groestl512_4way_update_close( &ctx.groestl, vhashB, vhashB, 640 );
+
+     uint32_t hash0[20] __attribute__ ((aligned (64)));
+     uint32_t hash1[20] __attribute__ ((aligned (64)));
+     uint32_t hash2[20] __attribute__ ((aligned (64)));
+     uint32_t hash3[20] __attribute__ ((aligned (64)));
+     uint32_t hash4[20] __attribute__ ((aligned (64)));
+     uint32_t hash5[20] __attribute__ ((aligned (64)));
+     uint32_t hash6[20] __attribute__ ((aligned (64)));
+     uint32_t hash7[20] __attribute__ ((aligned (64)));
+
+//     rintrlv_4x128_8x32( vhash, vhashA, vhashB, 512 );
+     dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhashA );
+     dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhashB );
+     intrlv_8x32_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5,
+                       hash6, hash7 );
+
+#else
+
+     uint32_t hash0[20] __attribute__ ((aligned (64)));
+     uint32_t hash1[20] __attribute__ ((aligned (64)));
+     uint32_t hash2[20] __attribute__ ((aligned (64)));
+     uint32_t hash3[20] __attribute__ ((aligned (64)));
+     uint32_t hash4[20] __attribute__ ((aligned (64)));
+     uint32_t hash5[20] __attribute__ ((aligned (64)));  
+     uint32_t hash6[20] __attribute__ ((aligned (64)));
+     uint32_t hash7[20] __attribute__ ((aligned (64)));
+
+     dintrlv_8x64( hash0, hash1, hash2, hash3,
+                   hash4, hash5, hash6, hash7, input, 640 );
+
+     update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash4, (char*)hash4, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash5, (char*)hash5, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash6, (char*)hash6, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+     update_and_final_groestl( &ctx.groestl, (char*)hash7, (char*)hash7, 640 );
+     memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
+
+     intrlv_8x32( vhash, hash0, hash1, hash2, hash3,
+                         hash4, hash5, hash6, hash7, 512 );
+
+#endif
+
+     sha256_8way_update( &ctx.sha, vhash, 64 );
+     sha256_8way_close( &ctx.sha, output );
+}
+
+int scanhash_myriad_8way( struct work *work, uint32_t max_nonce,
+                          uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t hash[8*8] __attribute__ ((aligned (128)));
+   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[7<<3]);
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce - 8;
+   uint32_t n = first_nonce;
+   uint32_t *noncep = vdata + 64+3;   // 4*16 + 3
+   int thr_id = mythr->id;  // thr_id arg is deprecated
+
+   if ( opt_benchmark )
+      ( (uint32_t*)ptarget )[7] = 0x0000ff;
+
+   mm512_bswap32_intrlv80_4x128( vdata, pdata );
+
+   do
+   {
+      be32enc( noncep,    n   );
+      be32enc( noncep+ 8, n+1 );
+      be32enc( noncep+16, n+2 );
+      be32enc( noncep+24, n+3 );
+      be32enc( noncep+32, n+4 );
+      be32enc( noncep+40, n+5 );
+      be32enc( noncep+48, n+6 );
+      be32enc( noncep+64, n+7 );
+
+      myriad_8way_hash( hash, vdata );
+      pdata[19] = n;
+
+      for ( int lane = 0; lane < 8; lane++ )
+      if ( hash7[ lane ] <= Htarg )
+      {
+         extr_lane_8x32( lane_hash, hash, lane, 256 );
+         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+         {
+            pdata[19] = n + lane;
+            submit_lane_solution( work, lane_hash, mythr, lane );
+         }
+      }
+      n += 8;
+   } while ( (n < last_nonce) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+
+#elif defined(MYRGR_4WAY)

 typedef struct {
    hashState_groestl       groestl;
@@ -45,12 +190,8 @@ void myriad_4way_hash( void *output, const void *input )

     intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );

-     sha256_4way( &ctx.sha, vhash, 64 );
+     sha256_4way_update( &ctx.sha, vhash, 64 );
     sha256_4way_close( &ctx.sha, output );
-
-//     sha256_4way_close( &ctx.sha, vhash );
-//     mm128_dintrlv_4x32( output, output+32, output+64, output+96,
-//                           vhash, 256 );
 }

 int scanhash_myriad_4way( struct work *work, uint32_t max_nonce,
@@ -68,18 +209,10 @@ int scanhash_myriad_4way( struct work *work, uint32_t max_nonce,
   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
   int thr_id = mythr->id;  // thr_id arg is deprecated

-/*
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
-
-	uint32_t _ALIGN(64) endiandata[20];
-	const uint32_t first_nonce = pdata[19];
-	uint32_t nonce = first_nonce;
-*/
   if ( opt_benchmark )
      ( (uint32_t*)ptarget )[7] = 0x0000ff;

-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
   do {
      *noncev = mm128_bswap_32( _mm_set_epi32( n+3,n+2,n+1,n ) );

--- a/algo/groestl/myrgr-gate.c
+++ b/algo/groestl/myrgr-gate.c
@@ -2,17 +2,22 @@

 bool register_myriad_algo( algo_gate_t* gate )
 {
-#if defined (MYRGR_4WAY)
+#if defined (MYRGR_8WAY)
+  init_myrgr_8way_ctx();
+  gate->scanhash  = (void*)&scanhash_myriad_8way;
+  gate->hash      = (void*)&myriad_8way_hash;
+  gate->optimizations = AES_OPT | AVX2_OPT | VAES_OPT;
+#elif defined (MYRGR_4WAY)
  init_myrgr_4way_ctx();
  gate->scanhash  = (void*)&scanhash_myriad_4way;
  gate->hash      = (void*)&myriad_4way_hash;
+  gate->optimizations = AES_OPT | SSE2_OPT | AVX2_OPT | VAES_OPT;
 #else
  init_myrgr_ctx();
  gate->scanhash  = (void*)&scanhash_myriad;
  gate->hash      = (void*)&myriad_hash;
+  gate->optimizations = AES_OPT | SSE2_OPT | AVX2_OPT | SHA_OPT | VAES_OPT;
 #endif
-  gate->optimizations = AES_OPT | AVX2_OPT;
-  gate->get_max64 = (void*)&get_max64_0x3ffff;
  return true;
 };

--- a/algo/groestl/myrgr-gate.h
+++ b/algo/groestl/myrgr-gate.h
@@ -1,30 +1,35 @@
 #ifndef MYRGR_GATE_H__
-#define MYRGR_GATE_H__
+#define MYRGR_GATE_H__ 1

 #include "algo-gate-api.h"
 #include <stdint.h>

-#if defined(__AVX2__) && defined(__AES__)
-  #define MYRGR_4WAY
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define MYRGR_8WAY 1
+#elif defined(__AVX2__) && defined(__AES__) && !defined(__SHA__)
+  #define MYRGR_4WAY 1
 #endif

-#if defined(MYRGR_4WAY)
+#if defined(MYRGR_8WAY)
+
+void myriad_8way_hash( void *state, const void *input );
+int scanhash_myriad_8way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+void init_myrgr_8way_ctx();
+
+#elif defined(MYRGR_4WAY)

 void myriad_4way_hash( void *state, const void *input );
-
 int scanhash_myriad_4way( struct work *work, uint32_t max_nonce,
                         uint64_t *hashes_done, struct thr_info *mythr );
-
 void init_myrgr_4way_ctx();

-#endif
+#else

 void myriad_hash( void *state, const void *input );
-
 int scanhash_myriad( struct work *work, uint32_t max_nonce,
                    uint64_t *hashes_done, struct thr_info *mythr );
-
 void init_myrgr_ctx();

 #endif
-
+#endif
--- a/algo/hamsi/hamsi-hash-4way.c
+++ b/algo/hamsi/hamsi-hash-4way.c
@@ -32,8 +32,6 @@

 #include <stddef.h>
 #include <string.h>
-
-//#include "miner.h"
 #include "hamsi-hash-4way.h"

 #if defined(__AVX2__)
@@ -100,7 +98,7 @@ extern "C"{
 #endif

 //#include "hamsi-helper-4way.c"
-
+/*
 static const sph_u32 IV512[] = {
 	SPH_C32(0x73746565), SPH_C32(0x6c706172), SPH_C32(0x6b204172),
 	SPH_C32(0x656e6265), SPH_C32(0x72672031), SPH_C32(0x302c2062),
@@ -109,7 +107,7 @@ static const sph_u32 IV512[] = {
 	SPH_C32(0x65766572), SPH_C32(0x6c65652c), SPH_C32(0x2042656c),
 	SPH_C32(0x6769756d)
 };
-
+*/
 static const sph_u32 alpha_n[] = {
 	SPH_C32(0xff00f0f0), SPH_C32(0xccccaaaa), SPH_C32(0xf0f0cccc),
 	SPH_C32(0xff00aaaa), SPH_C32(0xccccaaaa), SPH_C32(0xf0f0ff00),
@@ -138,6 +136,7 @@ static const sph_u32 alpha_f[] = {
 	SPH_C32(0xcaf9f9c0), SPH_C32(0x0ff0639c)
 };

+
 // imported from hamsi helper

 /* Note: this table lists bits within each byte from least
@@ -529,48 +528,374 @@ static const sph_u32 T512[64][16] = {
 	  SPH_C32(0xe7e00a94) }
 };

+#define s0   m0
+#define s1   c0
+#define s2   m1
+#define s3   c1
+#define s4   c2
+#define s5   m2
+#define s6   c3
+#define s7   m3
+#define s8   m4
+#define s9   c4
+#define sA   m5
+#define sB   c5
+#define sC   c6
+#define sD   m6
+#define sE   c7
+#define sF   m7
+
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// Hamsi 8 way 
+
+#define INPUT_BIG8 \
+do { \
+  __m512i db = *buf; \
+  const uint64_t *tp = (uint64_t*)&T512[0][0];  \
+  m0 = m1 = m2 = m3 = m4 = m5 = m6 = m7 = m512_zero; \
+  for ( int u = 0; u < 64; u++ ) \
+  { \
+     __m512i dm = _mm512_and_si512( db, m512_one_64 ) ; \
+     dm = mm512_negate_32( _mm512_or_si512( dm, \
+                                          _mm512_slli_epi64( dm, 32 ) ) ); \
+     m0 = _mm512_xor_si512( m0, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[0] ) ) ); \
+     m1 = _mm512_xor_si512( m1, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[1] ) ) ); \
+     m2 = _mm512_xor_si512( m2, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[2] ) ) ); \
+     m3 = _mm512_xor_si512( m3, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[3] ) ) ); \
+     m4 = _mm512_xor_si512( m4, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[4] ) ) ); \
+     m5 = _mm512_xor_si512( m5, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[5] ) ) ); \
+     m6 = _mm512_xor_si512( m6, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[6] ) ) ); \
+     m7 = _mm512_xor_si512( m7, _mm512_and_si512( dm, \
+                                          m512_const1_64( tp[7] ) ) ); \
+     tp += 8; \
+     db = _mm512_srli_epi64( db, 1 ); \
+  } \
+} while (0)
+
+#define SBOX8( a, b, c, d ) \
+do { \
+  __m512i t; \
+  t = a; \
+  a = _mm512_and_si512( a, c ); \
+  a = _mm512_xor_si512( a, d ); \
+  c = _mm512_xor_si512( c, b ); \
+  c = _mm512_xor_si512( c, a ); \
+  d = _mm512_or_si512( d, t ); \
+  d = _mm512_xor_si512( d, b ); \
+  t = _mm512_xor_si512( t, c ); \
+  b = d; \
+  d = _mm512_or_si512( d, t ); \
+  d = _mm512_xor_si512( d, a ); \
+  a = _mm512_and_si512( a, b ); \
+  t = _mm512_xor_si512( t, a ); \
+  b = _mm512_xor_si512( b, d ); \
+  b = _mm512_xor_si512( b, t ); \
+  a = c; \
+  c = b; \
+  b = d; \
+  d = mm512_not( t ); \
+} while (0)
+
+#define L8( a, b, c, d ) \
+do { \
+   a = mm512_rol_32( a, 13 ); \
+   c = mm512_rol_32( c,  3 ); \
+   b = _mm512_xor_si512( b, _mm512_xor_si512( a, c ) ); \
+   d = _mm512_xor_si512( d, _mm512_xor_si512( c, \
+                                              _mm512_slli_epi32( a, 3 ) ) ); \
+   b = mm512_rol_32( b, 1 ); \
+   d = mm512_rol_32( d, 7 ); \
+   a = _mm512_xor_si512( a, _mm512_xor_si512( b, d ) ); \
+   c = _mm512_xor_si512( c, _mm512_xor_si512( d, \
+                                              _mm512_slli_epi32( b, 7 ) ) ); \
+   a = mm512_rol_32( a,  5 ); \
+   c = mm512_rol_32( c, 22 ); \
+} while (0)
+
+#define DECL_STATE_BIG8 \
+   __m512i c0, c1, c2, c3, c4, c5, c6, c7; \
+
+#define READ_STATE_BIG8(sc) \
+do { \
+   c0 = sc->h[0x0]; \
+   c1 = sc->h[0x1]; \
+   c2 = sc->h[0x2]; \
+   c3 = sc->h[0x3]; \
+   c4 = sc->h[0x4]; \
+   c5 = sc->h[0x5]; \
+   c6 = sc->h[0x6]; \
+   c7 = sc->h[0x7]; \
+} while (0)
+
+#define WRITE_STATE_BIG8(sc) \
+do { \
+   sc->h[0x0] = c0; \
+   sc->h[0x1] = c1; \
+   sc->h[0x2] = c2; \
+   sc->h[0x3] = c3; \
+   sc->h[0x4] = c4; \
+   sc->h[0x5] = c5; \
+   sc->h[0x6] = c6; \
+   sc->h[0x7] = c7; \
+} while (0)
+
+
+#define ROUND_BIG8(rc, alpha) \
+do { \
+   __m512i t0, t1, t2, t3; \
+   s0 = _mm512_xor_si512( s0, m512_const1_64( \
+                   ( (uint64_t)(rc) << 32 ) ^ ( (uint64_t*)(alpha) )[ 0] ) ); \
+   s1 = _mm512_xor_si512( s1, m512_const1_64( ( (uint64_t*)(alpha) )[ 1] ) ); \
+   s2 = _mm512_xor_si512( s2, m512_const1_64( ( (uint64_t*)(alpha) )[ 2] ) ); \
+   s3 = _mm512_xor_si512( s3, m512_const1_64( ( (uint64_t*)(alpha) )[ 3] ) ); \
+   s4 = _mm512_xor_si512( s4, m512_const1_64( ( (uint64_t*)(alpha) )[ 4] ) ); \
+   s5 = _mm512_xor_si512( s5, m512_const1_64( ( (uint64_t*)(alpha) )[ 5] ) ); \
+   s6 = _mm512_xor_si512( s6, m512_const1_64( ( (uint64_t*)(alpha) )[ 6] ) ); \
+   s7 = _mm512_xor_si512( s7, m512_const1_64( ( (uint64_t*)(alpha) )[ 7] ) ); \
+   s8 = _mm512_xor_si512( s8, m512_const1_64( ( (uint64_t*)(alpha) )[ 8] ) ); \
+   s9 = _mm512_xor_si512( s9, m512_const1_64( ( (uint64_t*)(alpha) )[ 9] ) ); \
+   sA = _mm512_xor_si512( sA, m512_const1_64( ( (uint64_t*)(alpha) )[10] ) ); \
+   sB = _mm512_xor_si512( sB, m512_const1_64( ( (uint64_t*)(alpha) )[11] ) ); \
+   sC = _mm512_xor_si512( sC, m512_const1_64( ( (uint64_t*)(alpha) )[12] ) ); \
+   sD = _mm512_xor_si512( sD, m512_const1_64( ( (uint64_t*)(alpha) )[13] ) ); \
+   sE = _mm512_xor_si512( sE, m512_const1_64( ( (uint64_t*)(alpha) )[14] ) ); \
+   sF = _mm512_xor_si512( sF, m512_const1_64( ( (uint64_t*)(alpha) )[15] ) ); \
+\
+  SBOX8( s0, s4, s8, sC ); \
+  SBOX8( s1, s5, s9, sD ); \
+  SBOX8( s2, s6, sA, sE ); \
+  SBOX8( s3, s7, sB, sF ); \
+\
+  t1 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s4, 4 ), \
+                                        _mm512_bslli_epi128( s5, 4 ) ); \
+  t3 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( sD, 4 ), \
+                                        _mm512_bslli_epi128( sE, 4 ) ); \
+  L8( s0, t1, s9, t3 ); \
+  s4 = _mm512_mask_blend_epi32( 0xaaaa, s4, _mm512_bslli_epi128( t1, 4 ) ); \
+  s5 = _mm512_mask_blend_epi32( 0x5555, s5, _mm512_bsrli_epi128( t1, 4 ) ); \
+  sD = _mm512_mask_blend_epi32( 0xaaaa, sD, _mm512_bslli_epi128( t3, 4 ) ); \
+  sE = _mm512_mask_blend_epi32( 0x5555, sE, _mm512_bsrli_epi128( t3, 4 ) ); \
+\
+  t1 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s5, 4 ), \
+                                        _mm512_bslli_epi128( s6, 4 ) ); \
+  t3 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( sE, 4 ), \
+                                        _mm512_bslli_epi128( sF, 4 ) ); \
+  L8( s1, t1, sA, t3 ); \
+  s5 = _mm512_mask_blend_epi32( 0xaaaa, s5, _mm512_bslli_epi128( t1, 4 ) ); \
+  s6 = _mm512_mask_blend_epi32( 0x5555, s6, _mm512_bsrli_epi128( t1, 4 ) ); \
+  sE = _mm512_mask_blend_epi32( 0xaaaa, sE, _mm512_bslli_epi128( t3, 4 ) ); \
+  sF = _mm512_mask_blend_epi32( 0x5555, sF, _mm512_bsrli_epi128( t3, 4 ) ); \
+\
+  t1 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s6, 4 ), \
+                                        _mm512_bslli_epi128( s7, 4 ) ); \
+  t3 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( sF, 4 ), \
+                                        _mm512_bslli_epi128( sC, 4 ) ); \
+  L8( s2, t1, sB, t3 ); \
+  s6 = _mm512_mask_blend_epi32( 0xaaaa, s6, _mm512_bslli_epi128( t1, 4 ) ); \
+  s7 = _mm512_mask_blend_epi32( 0x5555, s7, _mm512_bsrli_epi128( t1, 4 ) ); \
+  sF = _mm512_mask_blend_epi32( 0xaaaa, sF, _mm512_bslli_epi128( t3, 4 ) ); \
+  sC = _mm512_mask_blend_epi32( 0x5555, sC, _mm512_bsrli_epi128( t3, 4 ) ); \
+\
+  t1 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s7, 4 ), \
+                                        _mm512_bslli_epi128( s4, 4 ) ); \
+  t3 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( sC, 4 ), \
+                                        _mm512_bslli_epi128( sD, 4 ) ); \
+  L8( s3, t1, s8, t3 ); \
+  s7 = _mm512_mask_blend_epi32( 0xaaaa, s7, _mm512_bslli_epi128( t1, 4 ) ); \
+  s4 = _mm512_mask_blend_epi32( 0x5555, s4, _mm512_bsrli_epi128( t1, 4 ) ); \
+  sC = _mm512_mask_blend_epi32( 0xaaaa, sC, _mm512_bslli_epi128( t3, 4 ) ); \
+  sD = _mm512_mask_blend_epi32( 0x5555, sD, _mm512_bsrli_epi128( t3, 4 ) ); \
+\
+  t0 = _mm512_mask_blend_epi32( 0xaaaa, s0, _mm512_bslli_epi128( s8, 4 ) ); \
+  t1 = _mm512_mask_blend_epi32( 0xaaaa, s1, s9 ); \
+  t2 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s2, 4 ), sA ); \
+  t3 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s3, 4 ), \
+                                        _mm512_bslli_epi128( sB, 4 ) ); \
+  L8( t0, t1, t2, t3 ); \
+  s0 = _mm512_mask_blend_epi32( 0x5555, s0, t0 ); \
+  s8 = _mm512_mask_blend_epi32( 0x5555, s8, _mm512_bsrli_epi128( t0, 4 ) ); \
+  s1 = _mm512_mask_blend_epi32( 0x5555, s1, t1 ); \
+  s9 = _mm512_mask_blend_epi32( 0xaaaa, s9, t1 ); \
+  s2 = _mm512_mask_blend_epi32( 0xaaaa, s2, _mm512_bslli_epi128( t2, 4 ) ); \
+  sA = _mm512_mask_blend_epi32( 0xaaaa, sA, t2 ); \
+  s3 = _mm512_mask_blend_epi32( 0xaaaa, s3, _mm512_bslli_epi128( t3, 4 ) ); \
+  sB = _mm512_mask_blend_epi32( 0x5555, sB, _mm512_bsrli_epi128( t3, 4 ) ); \
+\
+  t0 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s4, 4 ), sC ); \
+  t1 = _mm512_mask_blend_epi32( 0xaaaa, _mm512_bsrli_epi128( s5, 4 ), \
+                                        _mm512_bslli_epi128( sD, 4 ) ); \
+  t2 = _mm512_mask_blend_epi32( 0xaaaa, s6, _mm512_bslli_epi128( sE, 4 ) ); \
+  t3 = _mm512_mask_blend_epi32( 0xaaaa, s7, sF ); \
+  L8( t0, t1, t2, t3 ); \
+  s4 = _mm512_mask_blend_epi32( 0xaaaa, s4, _mm512_bslli_epi128( t0, 4 ) ); \
+  sC = _mm512_mask_blend_epi32( 0xaaaa, sC, t0 ); \
+  s5 = _mm512_mask_blend_epi32( 0xaaaa, s5, _mm512_bslli_epi128( t1, 4 ) ); \
+  sD = _mm512_mask_blend_epi32( 0x5555, sD, _mm512_bsrli_epi128( t1, 4 ) ); \
+  s6 = _mm512_mask_blend_epi32( 0x5555, s6, t2 ); \
+  sE = _mm512_mask_blend_epi32( 0x5555, sE, _mm512_bsrli_epi128( t2, 4 ) ); \
+  s7 = _mm512_mask_blend_epi32( 0x5555, s7, t3 ); \
+  sF = _mm512_mask_blend_epi32( 0xaaaa, sF, t3 ); \
+} while (0)
+
+#define P_BIG8 \
+do { \
+   ROUND_BIG8(0, alpha_n); \
+   ROUND_BIG8(1, alpha_n); \
+   ROUND_BIG8(2, alpha_n); \
+   ROUND_BIG8(3, alpha_n); \
+   ROUND_BIG8(4, alpha_n); \
+   ROUND_BIG8(5, alpha_n); \
+} while (0)
+
+#define PF_BIG8 \
+do { \
+   ROUND_BIG8( 0, alpha_f); \
+   ROUND_BIG8( 1, alpha_f); \
+   ROUND_BIG8( 2, alpha_f); \
+   ROUND_BIG8( 3, alpha_f); \
+   ROUND_BIG8( 4, alpha_f); \
+   ROUND_BIG8( 5, alpha_f); \
+   ROUND_BIG8( 6, alpha_f); \
+   ROUND_BIG8( 7, alpha_f); \
+   ROUND_BIG8( 8, alpha_f); \
+   ROUND_BIG8( 9, alpha_f); \
+   ROUND_BIG8(10, alpha_f); \
+   ROUND_BIG8(11, alpha_f); \
+} while (0)
+
+#define T_BIG8 \
+do { /* order is important */ \
+   c7 = sc->h[ 0x7 ] = _mm512_xor_si512( sc->h[ 0x7 ], sB ); \
+   c6 = sc->h[ 0x6 ] = _mm512_xor_si512( sc->h[ 0x6 ], sA ); \
+   c5 = sc->h[ 0x5 ] = _mm512_xor_si512( sc->h[ 0x5 ], s9 ); \
+   c4 = sc->h[ 0x4 ] = _mm512_xor_si512( sc->h[ 0x4 ], s8 ); \
+   c3 = sc->h[ 0x3 ] = _mm512_xor_si512( sc->h[ 0x3 ], s3 ); \
+   c2 = sc->h[ 0x2 ] = _mm512_xor_si512( sc->h[ 0x2 ], s2 ); \
+   c1 = sc->h[ 0x1 ] = _mm512_xor_si512( sc->h[ 0x1 ], s1 ); \
+   c0 = sc->h[ 0x0 ] = _mm512_xor_si512( sc->h[ 0x0 ], s0 ); \
+} while (0)
+
+void hamsi_8way_big( hamsi_8way_big_context *sc, __m512i *buf, size_t num )
+{
+   DECL_STATE_BIG8
+   uint32_t tmp = num << 6;
+
+   sc->count_low = SPH_T32( sc->count_low + tmp );
+   sc->count_high += (sph_u32)( (num >> 13) >> 13 );
+   if ( sc->count_low < tmp )
+      sc->count_high++;
+
+   READ_STATE_BIG8( sc );
+   while ( num-- > 0 )
+   {
+      __m512i m0, m1, m2, m3, m4, m5, m6, m7;
+
+      INPUT_BIG8;
+      P_BIG8;
+      T_BIG8;
+      buf++;
+   }
+   WRITE_STATE_BIG8( sc );
+}
+
+void hamsi_8way_big_final( hamsi_8way_big_context *sc, __m512i *buf )
+{
+   __m512i m0, m1, m2, m3, m4, m5, m6, m7;
+   DECL_STATE_BIG8
+   READ_STATE_BIG8( sc );
+   INPUT_BIG8;
+   PF_BIG8;
+   T_BIG8;
+   WRITE_STATE_BIG8( sc );
+}
+
+
+void hamsi512_8way_init( hamsi_8way_big_context *sc )
+{
+   sc->partial_len = 0;
+   sc->count_high = sc->count_low = 0;
+
+   sc->h[0] = m512_const1_64( 0x6c70617273746565 );
+   sc->h[1] = m512_const1_64( 0x656e62656b204172 );
+   sc->h[2] = m512_const1_64( 0x302c206272672031 );
+   sc->h[3] = m512_const1_64( 0x3434362c75732032 );
+   sc->h[4] = m512_const1_64( 0x3030312020422d33 );
+   sc->h[5] = m512_const1_64( 0x656e2d484c657576 );
+   sc->h[6] = m512_const1_64( 0x6c65652c65766572 );
+   sc->h[7] = m512_const1_64( 0x6769756d2042656c );
+}
+
+void hamsi512_8way_update( hamsi_8way_big_context *sc, const void *data,
+                           size_t len )
+{
+   __m512i *vdata = (__m512i*)data;
+
+   hamsi_8way_big( sc, vdata, len>>3 );
+   vdata += ( (len& ~(size_t)7) >> 3 );
+   len &= (size_t)7;
+   memcpy_512( sc->buf, vdata, len>>3 );
+   sc->partial_len = len;
+}
+
+void hamsi512_8way_close( hamsi_8way_big_context *sc, void *dst )
+{
+   __m512i pad[1];
+   int ch, cl;
+
+   sph_enc32be( &ch, sc->count_high );
+   sph_enc32be( &cl, sc->count_low + ( sc->partial_len << 3 ) );
+   pad[0] =  _mm512_set_epi32( cl, ch, cl, ch, cl, ch, cl, ch,
+                               cl, ch, cl, ch, cl, ch, cl, ch );
+//   pad[0] =  m512_const2_32( cl, ch );
+   sc->buf[0] = m512_const1_64( 0x80 );
+   hamsi_8way_big( sc, sc->buf, 1 );
+   hamsi_8way_big_final( sc, pad );
+
+   mm512_block_bswap_32( (__m512i*)dst, sc->h );
+}
+
+
+#endif // AVX512
+
+
+// Hamsi 4 way
+
 #define INPUT_BIG \
 do { \
  __m256i db = *buf; \
-  const sph_u32 *tp = &T512[0][0]; \
-  m0 = m256_zero; \
-  m1 = m256_zero; \
-  m2 = m256_zero; \
-  m3 = m256_zero; \
-  m4 = m256_zero; \
-  m5 = m256_zero; \
-  m6 = m256_zero; \
-  m7 = m256_zero; \
+  const uint64_t *tp = (uint64_t*)&T512[0][0];  \
+  m0 = m1 = m2 = m3 = m4 = m5 = m6 = m7 = m256_zero; \
  for ( int u = 0; u < 64; u++ ) \
  { \
     __m256i dm = _mm256_and_si256( db, m256_one_64 ) ; \
     dm = mm256_negate_32( _mm256_or_si256( dm, \
                         _mm256_slli_epi64( dm, 32 ) ) ); \
     m0 = _mm256_xor_si256( m0, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0x1], tp[0x0], tp[0x1], tp[0x0], \
-                                    tp[0x1], tp[0x0], tp[0x1], tp[0x0] ) ) ); \
+                                          m256_const1_64( tp[0] ) ) ); \
     m1 = _mm256_xor_si256( m1, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0x3], tp[0x2], tp[0x3], tp[0x2], \
-                                    tp[0x3], tp[0x2], tp[0x3], tp[0x2] ) ) ); \
+                                          m256_const1_64( tp[1] ) ) ); \
     m2 = _mm256_xor_si256( m2, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0x5], tp[0x4], tp[0x5], tp[0x4], \
-                                    tp[0x5], tp[0x4], tp[0x5], tp[0x4] ) ) ); \
+                                          m256_const1_64( tp[2] ) ) ); \
     m3 = _mm256_xor_si256( m3, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0x7], tp[0x6], tp[0x7], tp[0x6], \
-                                    tp[0x7], tp[0x6], tp[0x7], tp[0x6] ) ) ); \
+                                          m256_const1_64( tp[3] ) ) ); \
     m4 = _mm256_xor_si256( m4, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0x9], tp[0x8], tp[0x9], tp[0x8], \
-                                    tp[0x9], tp[0x8], tp[0x9], tp[0x8] ) ) ); \
+                                          m256_const1_64( tp[4] ) ) ); \
     m5 = _mm256_xor_si256( m5, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0xB], tp[0xA], tp[0xB], tp[0xA], \
-                                    tp[0xB], tp[0xA], tp[0xB], tp[0xA] ) ) ); \
+                                          m256_const1_64( tp[5] ) ) ); \
     m6 = _mm256_xor_si256( m6, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0xD], tp[0xC], tp[0xD], tp[0xC], \
-                                    tp[0xD], tp[0xC], tp[0xD], tp[0xC] ) ) ); \
+                                          m256_const1_64( tp[6] ) ) ); \
     m7 = _mm256_xor_si256( m7, _mm256_and_si256( dm, \
-                  _mm256_set_epi32( tp[0xF], tp[0xE], tp[0xF], tp[0xE], \
-                                    tp[0xF], tp[0xE], tp[0xF], tp[0xE] ) ) ); \
-     tp += 0x10; \
+                                          m256_const1_64( tp[7] ) ) ); \
+     tp += 8; \
     db = _mm256_srli_epi64( db, 1 ); \
  } \
 } while (0)
@@ -642,6 +967,7 @@ do { \
   sc->h[0x7] = c7; \
 } while (0)

+/*
 #define s0   m0
 #define s1   c0
 #define s2   m1
@@ -658,58 +984,28 @@ do { \
 #define sD   m6
 #define sE   c7
 #define sF   m7
+*/

 #define ROUND_BIG(rc, alpha) \
 do { \
-  __m256i t0, t1, t2, t3; \
-  s0 = _mm256_xor_si256( s0, _mm256_set_epi32( \
-        alpha[0x01] ^ (rc), alpha[0x00], alpha[0x01] ^ (rc), alpha[0x00], \
-        alpha[0x01] ^ (rc), alpha[0x00], alpha[0x01] ^ (rc), alpha[0x00] ) ); \
-  s1 = _mm256_xor_si256( s1, _mm256_set_epi32( \
-                     alpha[0x03], alpha[0x02], alpha[0x03], alpha[0x02], \
-                     alpha[0x03], alpha[0x02], alpha[0x03], alpha[0x02] ) ); \
-  s2 = _mm256_xor_si256( s2, _mm256_set_epi32( \
-                     alpha[0x05], alpha[0x04], alpha[0x05], alpha[0x04], \
-                     alpha[0x05], alpha[0x04], alpha[0x05], alpha[0x04] ) ); \
-  s3 = _mm256_xor_si256( s3, _mm256_set_epi32( \
-                     alpha[0x07], alpha[0x06], alpha[0x07], alpha[0x06], \
-                     alpha[0x07], alpha[0x06], alpha[0x07], alpha[0x06] ) ); \
-  s4 = _mm256_xor_si256( s4, _mm256_set_epi32( \
-                     alpha[0x09], alpha[0x08], alpha[0x09], alpha[0x08], \
-                     alpha[0x09], alpha[0x08], alpha[0x09], alpha[0x08] ) ); \
-  s5 = _mm256_xor_si256( s5, _mm256_set_epi32( \
-                     alpha[0x0B], alpha[0x0A], alpha[0x0B], alpha[0x0A], \
-                     alpha[0x0B], alpha[0x0A], alpha[0x0B], alpha[0x0A] ) ); \
-  s6 = _mm256_xor_si256( s6, _mm256_set_epi32( \
-                     alpha[0x0D], alpha[0x0C], alpha[0x0D], alpha[0x0C], \
-                     alpha[0x0D], alpha[0x0C], alpha[0x0D], alpha[0x0C] ) ); \
-  s7 = _mm256_xor_si256( s7, _mm256_set_epi32( \
-                     alpha[0x0F], alpha[0x0E], alpha[0x0F], alpha[0x0E], \
-                     alpha[0x0F], alpha[0x0E], alpha[0x0F], alpha[0x0E] ) ); \
-  s8 = _mm256_xor_si256( s8, _mm256_set_epi32( \
-                     alpha[0x11], alpha[0x10], alpha[0x11], alpha[0x10], \
-                     alpha[0x11], alpha[0x10], alpha[0x11], alpha[0x10] ) ); \
-  s9 = _mm256_xor_si256( s9, _mm256_set_epi32( \
-                     alpha[0x13], alpha[0x12], alpha[0x13], alpha[0x12], \
-                     alpha[0x13], alpha[0x12], alpha[0x13], alpha[0x12] ) ); \
-  sA = _mm256_xor_si256( sA, _mm256_set_epi32( \
-                     alpha[0x15], alpha[0x14], alpha[0x15], alpha[0x14], \
-                     alpha[0x15], alpha[0x14], alpha[0x15], alpha[0x14] ) ); \
-  sB = _mm256_xor_si256( sB, _mm256_set_epi32( \
-                     alpha[0x17], alpha[0x16], alpha[0x17], alpha[0x16], \
-                     alpha[0x17], alpha[0x16], alpha[0x17], alpha[0x16] ) ); \
-  sC = _mm256_xor_si256( sC, _mm256_set_epi32( \
-                     alpha[0x19], alpha[0x18], alpha[0x19], alpha[0x18], \
-                     alpha[0x19], alpha[0x18], alpha[0x19], alpha[0x18] ) ); \
-  sD = _mm256_xor_si256( sD, _mm256_set_epi32( \
-                     alpha[0x1B], alpha[0x1A], alpha[0x1B], alpha[0x1A], \
-                     alpha[0x1B], alpha[0x1A], alpha[0x1B], alpha[0x1A] ) ); \
-  sE = _mm256_xor_si256( sE, _mm256_set_epi32( \
-                     alpha[0x1D], alpha[0x1C], alpha[0x1D], alpha[0x1C], \
-                     alpha[0x1D], alpha[0x1C], alpha[0x1D], alpha[0x1C] ) ); \
-  sF = _mm256_xor_si256( sF, _mm256_set_epi32( \
-                     alpha[0x1F], alpha[0x1E], alpha[0x1F], alpha[0x1E], \
-                     alpha[0x1F], alpha[0x1E], alpha[0x1F], alpha[0x1E] ) ); \
+   __m256i t0, t1, t2, t3; \
+   s0 = _mm256_xor_si256( s0, m256_const1_64( \
+                   ( (uint64_t)(rc) << 32 ) ^ ( (uint64_t*)(alpha) )[ 0] ) ); \
+   s1 = _mm256_xor_si256( s1, m256_const1_64( ( (uint64_t*)(alpha) )[ 1] ) ); \
+   s2 = _mm256_xor_si256( s2, m256_const1_64( ( (uint64_t*)(alpha) )[ 2] ) ); \
+   s3 = _mm256_xor_si256( s3, m256_const1_64( ( (uint64_t*)(alpha) )[ 3] ) ); \
+   s4 = _mm256_xor_si256( s4, m256_const1_64( ( (uint64_t*)(alpha) )[ 4] ) ); \
+   s5 = _mm256_xor_si256( s5, m256_const1_64( ( (uint64_t*)(alpha) )[ 5] ) ); \
+   s6 = _mm256_xor_si256( s6, m256_const1_64( ( (uint64_t*)(alpha) )[ 6] ) ); \
+   s7 = _mm256_xor_si256( s7, m256_const1_64( ( (uint64_t*)(alpha) )[ 7] ) ); \
+   s8 = _mm256_xor_si256( s8, m256_const1_64( ( (uint64_t*)(alpha) )[ 8] ) ); \
+   s9 = _mm256_xor_si256( s9, m256_const1_64( ( (uint64_t*)(alpha) )[ 9] ) ); \
+   sA = _mm256_xor_si256( sA, m256_const1_64( ( (uint64_t*)(alpha) )[10] ) ); \
+   sB = _mm256_xor_si256( sB, m256_const1_64( ( (uint64_t*)(alpha) )[11] ) ); \
+   sC = _mm256_xor_si256( sC, m256_const1_64( ( (uint64_t*)(alpha) )[12] ) ); \
+   sD = _mm256_xor_si256( sD, m256_const1_64( ( (uint64_t*)(alpha) )[13] ) ); \
+   sE = _mm256_xor_si256( sE, m256_const1_64( ( (uint64_t*)(alpha) )[14] ) ); \
+   sF = _mm256_xor_si256( sF, m256_const1_64( ( (uint64_t*)(alpha) )[15] ) ); \
 \
  SBOX( s0, s4, s8, sC ); \
  SBOX( s1, s5, s9, sD ); \
@@ -863,47 +1159,23 @@ void hamsi_big_final( hamsi_4way_big_context *sc, __m256i *buf )
 void hamsi512_4way_init( hamsi_4way_big_context *sc )
 {
   sc->partial_len = 0;
-   sph_u32 lo, hi;
   sc->count_high = sc->count_low = 0;
-   for ( int i = 0; i < 8; i++ )
-   {
-      lo = 2*i;
-      hi = 2*i + 1;
-      sc->h[i] = _mm256_set_epi32( IV512[hi], IV512[lo], IV512[hi], IV512[lo],
-                                   IV512[hi], IV512[lo], IV512[hi], IV512[lo] );
-   }
+
+   sc->h[0] = m256_const1_64( 0x6c70617273746565 );
+   sc->h[1] = m256_const1_64( 0x656e62656b204172 );
+   sc->h[2] = m256_const1_64( 0x302c206272672031 );
+   sc->h[3] = m256_const1_64( 0x3434362c75732032 );
+   sc->h[4] = m256_const1_64( 0x3030312020422d33 );
+   sc->h[5] = m256_const1_64( 0x656e2d484c657576 );
+   sc->h[6] = m256_const1_64( 0x6c65652c65766572 );
+   sc->h[7] = m256_const1_64( 0x6769756d2042656c );
 }

-void hamsi512_4way( hamsi_4way_big_context *sc, const void *data, size_t len )
+void hamsi512_4way_update( hamsi_4way_big_context *sc, const void *data,
+      size_t len )
 {
   __m256i *vdata = (__m256i*)data;

-// It looks like the only way to get in here is if core was previously called
-// with a very small len
-// That's not likely even with 80 byte input so deprecate partial len
-/*
-   if ( sc->partial_len != 0 )
-   {
-      size_t mlen;
-
-      mlen = 8 - sc->partial_len;
-      if ( len < mlen )
-      {
-         memcpy_256( sc->partial + (sc->partial_len >> 3), data, len>>3 );
-         sc->partial_len += len;
-         return;
-      }
-      else
-      {
-         memcpy_256( sc->partial + (sc->partial_len >> 3), data, mlen>>3 );
-         len -= mlen;
-         vdata += mlen>>3;
-         hamsi_big( sc, sc->partial, 1 );
-         sc->partial_len = 0;
-      }
-   }
-*/
-
   hamsi_big( sc, vdata, len>>3 );
   vdata += ( (len& ~(size_t)7) >> 3 );
   len &= (size_t)7;
@@ -913,20 +1185,19 @@ void hamsi512_4way( hamsi_4way_big_context *sc, const void *data, size_t len )

 void hamsi512_4way_close( hamsi_4way_big_context *sc, void *dst )
 {
-   __m256i *out = (__m256i*)dst;
   __m256i pad[1];
-   size_t u;
   int ch, cl;

   sph_enc32be( &ch, sc->count_high );
   sph_enc32be( &cl, sc->count_low + ( sc->partial_len << 3 ) );
   pad[0] =  _mm256_set_epi32( cl, ch, cl, ch, cl, ch, cl, ch );
-   sc->buf[0] = _mm256_set_epi32( 0UL, 0x80UL, 0UL, 0x80UL,
-                                  0UL, 0x80UL, 0UL, 0x80UL );
+   sc->buf[0] = m256_const1_64( 0x80 );
+//      sc->buf[0] = _mm256_set_epi32( 0UL, 0x80UL, 0UL, 0x80UL,
+//                                  0UL, 0x80UL, 0UL, 0x80UL );
   hamsi_big( sc, sc->buf, 1 );
   hamsi_big_final( sc, pad );
-   for ( u = 0; u < 8; u ++ )
-      out[u] = mm256_bswap_32( sc->h[u] );
+
+   mm256_block_bswap_32( (__m256i*)dst, sc->h );
 }

 #ifdef __cplusplus
--- a/algo/hamsi/hamsi-hash-4way.h
+++ b/algo/hamsi/hamsi-hash-4way.h
@@ -60,9 +60,32 @@ typedef struct {
 typedef hamsi_4way_big_context hamsi512_4way_context;

 void hamsi512_4way_init( hamsi512_4way_context *sc );
-void hamsi512_4way( hamsi512_4way_context *sc, const void *data, size_t len );
+void hamsi512_4way_update( hamsi512_4way_context *sc, const void *data,
+      size_t len );
+//#define hamsi512_4way hamsi512_4way_update
 void hamsi512_4way_close( hamsi512_4way_context *sc, void *dst );

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+typedef struct {
+   __m512i h[8];
+   __m512i buf[1];
+   size_t partial_len;
+   sph_u32 count_high, count_low;
+} hamsi_8way_big_context;
+
+typedef hamsi_8way_big_context hamsi512_8way_context;
+
+void hamsi512_8way_init( hamsi512_8way_context *sc );
+void hamsi512_8way_update( hamsi512_8way_context *sc, const void *data,
+                           size_t len );
+void hamsi512_8way_close( hamsi512_8way_context *sc, void *dst );
+
+
+
+#endif
+
+
 #ifdef __cplusplus
 }
 #endif
--- a/algo/haval/haval-4way-helper.c
+++ b/algo/haval/haval-4way-helper.c
@@ -38,7 +38,7 @@
 #define SPH_XCAT_(a, b)   a ## b

 static void
-SPH_XCAT(SPH_XCAT(haval, PASSES), _4way)
+SPH_XCAT(SPH_XCAT(haval, PASSES), _4way_update)
 ( haval_4way_context *sc, const void *data, size_t len )
 {
   __m128i *vdata = (__m128i*)data;
--- a/algo/haval/haval-8way-helper.c
+++ b/algo/haval/haval-8way-helper.c
@@ -0,0 +1,115 @@
+/* $Id: haval_helper.c 218 2010-06-08 17:06:34Z tp $ */
+/*
+ * Helper code, included (three times !) by HAVAL implementation.
+ *
+ * TODO: try to merge this with md_helper.c.
+ *
+ * ==========================(LICENSE BEGIN)============================
+ *
+ * Copyright (c) 2007-2010  Projet RNRT SAPHIR
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+ * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * ===========================(LICENSE END)=============================
+ *
+ * @author   Thomas Pornin <thomas.pornin@cryptolog.com>
+ */
+
+#undef SPH_XCAT
+#define SPH_XCAT(a, b)    SPH_XCAT_(a, b)
+#undef SPH_XCAT_
+#define SPH_XCAT_(a, b)   a ## b
+
+static void
+SPH_XCAT(SPH_XCAT(haval, PASSES), _8way_update)
+( haval_8way_context *sc, const void *data, size_t len )
+{
+   __m256i *vdata = (__m256i*)data;
+   unsigned current;
+
+   current = (unsigned)sc->count_low & 127U;
+   while ( len > 0 )
+   {
+      unsigned clen;
+      uint32_t clow, clow2;
+
+      clen = 128U - current;
+      if ( clen > len )
+         clen = len;
+      memcpy_256( sc->buf + (current>>2), vdata, clen>>2 );
+      vdata += clen>>2;
+      current += clen;
+      len -= clen;
+      if ( current == 128U )
+      {
+         DSTATE_8W;
+         IN_PREPARE_8W(sc->buf);
+         RSTATE_8W;
+         SPH_XCAT(CORE_8W, PASSES)(INW_8W);
+         WSTATE_8W;
+         current = 0;
+      }
+      clow = sc->count_low;
+      clow2 = clow + clen;
+      sc->count_low = clow2;
+      if ( clow2 < clow )
+         sc->count_high ++;
+   }
+}
+
+static void
+SPH_XCAT(SPH_XCAT(haval, PASSES), _8way_close)( haval_8way_context *sc,
+                                                void *dst)
+{
+   unsigned current;
+   DSTATE_8W;
+
+   current = (unsigned)sc->count_low & 127UL;
+
+   sc->buf[ current>>2 ] = m256_one_32;
+   current += 4;   
+   RSTATE_8W;
+   if ( current > 116UL )
+   {
+      memset_zero_256( sc->buf + ( current>>2 ), (128UL-current) >> 2 );
+      do
+      {
+         IN_PREPARE_8W(sc->buf);
+         SPH_XCAT(CORE_8W, PASSES)(INW_8W);
+      } while (0);
+      current = 0;
+   }
+
+   uint32_t t1, t2;
+   memset_zero_256( sc->buf + ( current>>2 ), (116UL-current) >> 2 );
+   t1 = 0x01 | (PASSES << 3);
+   t2 = sc->olen << 3;
+   sc->buf[ 116>>2 ] = _mm256_set1_epi32( ( t1 << 16 ) | ( t2 << 24 ) );
+   sc->buf[ 120>>2 ] = _mm256_set1_epi32( sc->count_low << 3 );
+   sc->buf[ 124>>2 ] = _mm256_set1_epi32( (sc->count_high << 3)
+                                     | (sc->count_low >> 29) );
+   do
+   {
+      IN_PREPARE_8W(sc->buf);
+      SPH_XCAT(CORE_8W, PASSES)(INW_8W);
+   } while (0);
+   WSTATE_8W;
+   haval_8way_out( sc, dst );
+}
--- a/algo/haval/haval-hash-4way.c
+++ b/algo/haval/haval-hash-4way.c
@@ -40,7 +40,7 @@
 #include <string.h>
 #include "haval-hash-4way.h"

-// won't compile with sse4.2
+// won't compile with sse4.2, not a problem, it's only used with AVX2 4 way.
 //#if defined (__SSE4_2__)
 #if defined(__AVX__)

@@ -479,9 +479,9 @@ haval ## xxx ## _ ## y ## _4way_init(void *cc) \
 } \
 \
 void \
-haval ## xxx ## _ ## y ## _4way (void *cc, const void *data, size_t len) \
+haval ## xxx ## _ ## y ## _4way_update (void *cc, const void *data, size_t len) \
 { \
-	haval ## y ## _4way(cc, data, len); \
+	haval ## y ## _4way_update(cc, data, len); \
 } \
 \
 void \
@@ -518,6 +518,301 @@ do { \

 #define INMSG(i)   msg[i]

+#if defined(__AVX2__)
+
+// Haval-256 8 way 32 bit avx2
+
+#define F1_8W(x6, x5, x4, x3, x2, x1, x0) \
+   _mm256_xor_si256( x0, \
+       _mm256_xor_si256( _mm256_and_si256(_mm256_xor_si256( x0, x4 ), x1 ), \
+                      _mm256_xor_si256( _mm256_and_si256( x2, x5 ), \
+                                     _mm256_and_si256( x3, x6 ) ) ) ) \
+
+#define F2_8W(x6, x5, x4, x3, x2, x1, x0) \
+   _mm256_xor_si256( \
+      _mm256_and_si256( x2, \
+         _mm256_xor_si256( _mm256_andnot_si256( x3, x1 ), \
+                        _mm256_xor_si256( _mm256_and_si256( x4, x5 ), \
+                                       _mm256_xor_si256( x6, x0 ) ) ) ), \
+         _mm256_xor_si256( \
+             _mm256_and_si256( x4, _mm256_xor_si256( x1, x5 ) ), \
+             _mm256_xor_si256( _mm256_and_si256( x3, x5 ), x0 ) ) ) \
+
+#define F3_8W(x6, x5, x4, x3, x2, x1, x0) \
+  _mm256_xor_si256( \
+    _mm256_and_si256( x3, \
+      _mm256_xor_si256( _mm256_and_si256( x1, x2 ), \
+                     _mm256_xor_si256( x6, x0 ) ) ), \
+      _mm256_xor_si256( _mm256_xor_si256(_mm256_and_si256( x1, x4 ), \
+                                   _mm256_and_si256( x2, x5 ) ), x0 ) )
+
+#define F4_8W(x6, x5, x4, x3, x2, x1, x0) \
+  _mm256_xor_si256( \
+     _mm256_xor_si256( \
+        _mm256_and_si256( x3, \
+           _mm256_xor_si256( _mm256_xor_si256( _mm256_and_si256( x1, x2 ), \
+                                         _mm256_or_si256( x4, x6 ) ), x5 ) ), \
+        _mm256_and_si256( x4, \
+           _mm256_xor_si256( _mm256_xor_si256( _mm256_and_si256( mm256_not(x2), x5 ), \
+                          _mm256_xor_si256( x1, x6 ) ), x0 ) ) ), \
+     _mm256_xor_si256( _mm256_and_si256( x2, x6 ), x0 ) )
+
+
+#define F5_8W(x6, x5, x4, x3, x2, x1, x0) \
+   _mm256_xor_si256( \
+       _mm256_and_si256( x0, \
+            mm256_not( _mm256_xor_si256( \
+                    _mm256_and_si256( _mm256_and_si256( x1, x2 ), x3 ), x5 ) ) ), \
+      _mm256_xor_si256( _mm256_xor_si256( _mm256_and_si256( x1, x4 ), \
+                                    _mm256_and_si256( x2, x5 ) ), \
+                                    _mm256_and_si256( x3, x6 ) ) )
+
+#define FP3_1_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F1_8W(x1, x0, x3, x5, x6, x2, x4)
+#define FP3_2_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F2_8W(x4, x2, x1, x0, x5, x3, x6)
+#define FP3_3_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F3_8W(x6, x1, x2, x3, x4, x5, x0)
+
+#define FP4_1_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F1_8W(x2, x6, x1, x4, x5, x3, x0)
+#define FP4_2_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F2_8W(x3, x5, x2, x0, x1, x6, x4)
+#define FP4_3_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F3_8W(x1, x4, x3, x6, x0, x2, x5)
+#define FP4_4_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F4_8W(x6, x4, x0, x5, x2, x1, x3)
+
+#define FP5_1_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F1_8W(x3, x4, x1, x0, x5, x2, x6)
+#define FP5_2_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F2_8W(x6, x2, x1, x0, x3, x4, x5)
+#define FP5_3_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F3_8W(x2, x6, x0, x4, x3, x1, x5)
+#define FP5_4_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F4_8W(x1, x5, x3, x2, x0, x4, x6)
+#define FP5_5_8W(x6, x5, x4, x3, x2, x1, x0) \
+   F5_8W(x2, x5, x0, x6, x4, x3, x1)
+
+#define STEP_8W(n, p, x7, x6, x5, x4, x3, x2, x1, x0, w, c) \
+do { \
+   __m256i t = FP ## n ## _ ## p ## _8W(x6, x5, x4, x3, x2, x1, x0); \
+   x7 = _mm256_add_epi32( _mm256_add_epi32( mm256_ror_32( t, 7 ), \
+                                      mm256_ror_32( x7, 11 ) ), \
+                       _mm256_add_epi32( w, _mm256_set1_epi32( c ) ) ); \
+} while (0)
+
+#define PASS1_8W(n, in)   do { \
+      unsigned pass_count; \
+      for (pass_count = 0; pass_count < 32; pass_count += 8) { \
+         STEP_8W(n, 1, s7, s6, s5, s4, s3, s2, s1, s0, \
+            in(pass_count + 0), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s6, s5, s4, s3, s2, s1, s0, s7, \
+            in(pass_count + 1), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s5, s4, s3, s2, s1, s0, s7, s6, \
+            in(pass_count + 2), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s4, s3, s2, s1, s0, s7, s6, s5, \
+            in(pass_count + 3), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s3, s2, s1, s0, s7, s6, s5, s4, \
+            in(pass_count + 4), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s2, s1, s0, s7, s6, s5, s4, s3, \
+            in(pass_count + 5), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s1, s0, s7, s6, s5, s4, s3, s2, \
+            in(pass_count + 6), SPH_C32(0x00000000)); \
+         STEP_8W(n, 1, s0, s7, s6, s5, s4, s3, s2, s1, \
+            in(pass_count + 7), SPH_C32(0x00000000)); \
+         } \
+   } while (0)
+
+#define PASSG_8W(p, n, in)   do { \
+      unsigned pass_count; \
+      for (pass_count = 0; pass_count < 32; pass_count += 8) { \
+         STEP_8W(n, p, s7, s6, s5, s4, s3, s2, s1, s0, \
+            in(MP ## p[pass_count + 0]), \
+            RK ## p[pass_count + 0]); \
+         STEP_8W(n, p, s6, s5, s4, s3, s2, s1, s0, s7, \
+            in(MP ## p[pass_count + 1]), \
+            RK ## p[pass_count + 1]); \
+         STEP_8W(n, p, s5, s4, s3, s2, s1, s0, s7, s6, \
+            in(MP ## p[pass_count + 2]), \
+            RK ## p[pass_count + 2]); \
+         STEP_8W(n, p, s4, s3, s2, s1, s0, s7, s6, s5, \
+            in(MP ## p[pass_count + 3]), \
+            RK ## p[pass_count + 3]); \
+         STEP_8W(n, p, s3, s2, s1, s0, s7, s6, s5, s4, \
+            in(MP ## p[pass_count + 4]), \
+            RK ## p[pass_count + 4]); \
+         STEP_8W(n, p, s2, s1, s0, s7, s6, s5, s4, s3, \
+            in(MP ## p[pass_count + 5]), \
+            RK ## p[pass_count + 5]); \
+         STEP_8W(n, p, s1, s0, s7, s6, s5, s4, s3, s2, \
+            in(MP ## p[pass_count + 6]), \
+            RK ## p[pass_count + 6]); \
+         STEP_8W(n, p, s0, s7, s6, s5, s4, s3, s2, s1, \
+            in(MP ## p[pass_count + 7]), \
+            RK ## p[pass_count + 7]); \
+         } \
+   } while (0)
+
+#define PASS2_8W(n, in)    PASSG_8W(2, n, in)
+#define PASS3_8W(n, in)    PASSG_8W(3, n, in)
+#define PASS4_8W(n, in)    PASSG_8W(4, n, in)
+#define PASS5_8W(n, in)    PASSG_8W(5, n, in)
+
+#define SAVE_STATE_8W \
+   __m256i u0, u1, u2, u3, u4, u5, u6, u7; \
+   do { \
+      u0 = s0; \
+      u1 = s1; \
+      u2 = s2; \
+      u3 = s3; \
+      u4 = s4; \
+      u5 = s5; \
+      u6 = s6; \
+      u7 = s7; \
+   } while (0)
+
+#define UPDATE_STATE_8W \
+do { \
+   s0 = _mm256_add_epi32( s0, u0 ); \
+   s1 = _mm256_add_epi32( s1, u1 ); \
+   s2 = _mm256_add_epi32( s2, u2 ); \
+   s3 = _mm256_add_epi32( s3, u3 ); \
+   s4 = _mm256_add_epi32( s4, u4 ); \
+   s5 = _mm256_add_epi32( s5, u5 ); \
+   s6 = _mm256_add_epi32( s6, u6 ); \
+   s7 = _mm256_add_epi32( s7, u7 ); \
+} while (0)
+
+#define CORE_8W5(in)  do { \
+      SAVE_STATE_8W; \
+      PASS1_8W(5, in); \
+      PASS2_8W(5, in); \
+      PASS3_8W(5, in); \
+      PASS4_8W(5, in); \
+      PASS5_8W(5, in); \
+      UPDATE_STATE_8W; \
+   } while (0)
+
+#define DSTATE_8W   __m256i s0, s1, s2, s3, s4, s5, s6, s7
+
+#define RSTATE_8W \
+do { \
+   s0 = sc->s0; \
+   s1 = sc->s1; \
+   s2 = sc->s2; \
+   s3 = sc->s3; \
+   s4 = sc->s4; \
+   s5 = sc->s5; \
+   s6 = sc->s6; \
+   s7 = sc->s7; \
+} while (0)
+
+#define WSTATE_8W \
+do { \
+   sc->s0 = s0; \
+   sc->s1 = s1; \
+   sc->s2 = s2; \
+   sc->s3 = s3; \
+   sc->s4 = s4; \
+   sc->s5 = s5; \
+   sc->s6 = s6; \
+   sc->s7 = s7; \
+} while (0)
+
+static void
+haval_8way_init( haval_8way_context *sc, unsigned olen, unsigned passes )
+{
+   sc->s0 = m256_const1_32( 0x243F6A88UL );
+   sc->s1 = m256_const1_32( 0x85A308D3UL );
+   sc->s2 = m256_const1_32( 0x13198A2EUL );
+   sc->s3 = m256_const1_32( 0x03707344UL );
+   sc->s4 = m256_const1_32( 0xA4093822UL );
+   sc->s5 = m256_const1_32( 0x299F31D0UL );
+   sc->s6 = m256_const1_32( 0x082EFA98UL );
+   sc->s7 = m256_const1_32( 0xEC4E6C89UL );
+   sc->olen = olen;
+   sc->passes = passes;
+   sc->count_high = 0;
+   sc->count_low = 0;
+
+}
+#define IN_PREPARE_8W(indata) const __m256i *const load_ptr_8w = (indata)
+
+#define INW_8W(i)   load_ptr_8w[ i ] 
+
+static void
+haval_8way_out( haval_8way_context *sc, void *dst )
+{
+   __m256i *buf = (__m256i*)dst;
+   DSTATE_8W;
+   RSTATE_8W;
+
+   buf[0] = s0;
+   buf[1] = s1;
+   buf[2] = s2;
+   buf[3] = s3;
+   buf[4] = s4;
+   buf[5] = s5;
+   buf[6] = s6;
+   buf[7] = s7;
+}
+
+#undef PASSES
+#define PASSES   5
+#include "haval-8way-helper.c"
+
+#define API_8W(xxx, y) \
+void \
+haval ## xxx ## _ ## y ## _8way_init(void *cc) \
+{ \
+   haval_8way_init(cc, xxx >> 5, y); \
+} \
+ \
+void \
+haval ## xxx ## _ ## y ## _8way_update (void *cc, const void *data, size_t len) \
+{ \
+   haval ## y ## _8way_update(cc, data, len); \
+} \
+ \
+void \
+haval ## xxx ## _ ## y ## _8way_close(void *cc, void *dst) \
+{ \
+   haval ## y ## _8way_close(cc, dst); \
+} \
+
+API_8W(256, 5)
+
+#define RVAL_8W \
+do { \
+   s0 = val[0]; \
+   s1 = val[1]; \
+   s2 = val[2]; \
+   s3 = val[3]; \
+   s4 = val[4]; \
+   s5 = val[5]; \
+   s6 = val[6]; \
+   s7 = val[7]; \
+} while (0)
+
+#define WVAL_8W \
+do { \
+   val[0] = s0; \
+   val[1] = s1; \
+   val[2] = s2; \
+   val[3] = s3; \
+   val[4] = s4; \
+   val[5] = s5; \
+   val[6] = s6; \
+   val[7] = s7; \
+} while (0)
+
+#define INMSG_8W(i)   msg[i]
+
+
+
+#endif // AVX2
+
 #ifdef __cplusplus
 }
 #endif	
--- a/algo/haval/haval-hash-4way.h
+++ b/algo/haval/haval-hash-4way.h
@@ -59,7 +59,7 @@
 */

 #ifndef HAVAL_HASH_4WAY_H__
-#define HAVAL_HASH_4WAY_H__
+#define HAVAL_HASH_4WAY_H__ 1

 #if defined(__AVX__)

@@ -84,10 +84,30 @@ typedef haval_4way_context haval256_5_4way_context;

 void haval256_5_4way_init( void *cc );

-void haval256_5_4way( void *cc, const void *data, size_t len );
+void haval256_5_4way_update( void *cc, const void *data, size_t len );
+//#define haval256_5_4way haval256_5_4way_update

 void haval256_5_4way_close( void *cc, void *dst );

+#if defined(__AVX2__)
+
+typedef struct {
+   __m256i buf[32];
+   __m256i s0, s1, s2, s3, s4, s5, s6, s7;
+   unsigned olen, passes;
+   uint32_t count_high, count_low;
+} haval_8way_context __attribute__ ((aligned (64)));
+
+typedef haval_8way_context haval256_5_8way_context;
+
+void haval256_5_8way_init( void *cc );
+
+void haval256_5_8way_update( void *cc, const void *data, size_t len );
+
+void haval256_5_8way_close( void *cc, void *dst );
+
+#endif // AVX2
+
 #ifdef __cplusplus
 }
 #endif
--- a/algo/heavy/bastion.c
+++ b/algo/heavy/bastion.c
@@ -16,7 +16,6 @@
 #include "algo/echo/sph_echo.h"
 #include "algo/hamsi/sph_hamsi.h"
 #include "algo/luffa/luffa_for_sse2.h"
-#include "algo/skein/sse2/skein.c"

 #ifndef NO_AES_NI
  #include "algo/echo/aes_ni/hash_api.h"
@@ -35,12 +34,13 @@ void bastionhash(void *output, const void *input)
 	sph_fugue512_context ctx_fugue;
 	sph_whirlpool_context ctx_whirlpool;
 	sph_shabal512_context ctx_shabal;
-	sph_hamsi512_context ctx_hamsi;
+   sph_hamsi512_context ctx_hamsi;
+	sph_skein512_context ctx_skein;

-        unsigned char hashbuf[128] __attribute__ ((aligned (16)));
-        sph_u64 hashctA;
+//        unsigned char hashbuf[128] __attribute__ ((aligned (16)));
+//        sph_u64 hashctA;
 //        sph_u64 hashctB;
-        size_t hashptr;
+//        size_t hashptr;

 	HEFTY1(input, 80, hash);

@@ -56,10 +56,9 @@ void bastionhash(void *output, const void *input)
 		sph_fugue512(&ctx_fugue, hash, 64);
 		sph_fugue512_close(&ctx_fugue, hash);
 	} else {
-                DECL_SKN;
-                SKN_I;
-                SKN_U;
-                SKN_C;
+   sph_skein512_init( &ctx_skein );
+   sph_skein512( &ctx_skein, hash, 64 );
+   sph_skein512_close( &ctx_skein, hash );
 	}

 	sph_whirlpool_init(&ctx_whirlpool);
@@ -95,10 +94,9 @@ void bastionhash(void *output, const void *input)
 	sph_shabal512(&ctx_shabal, hash, 64);
 	sph_shabal512_close(&ctx_shabal, hash);

-        DECL_SKN;
-        SKN_I;
-        SKN_U;
-        SKN_C;
+   sph_skein512_init( &ctx_skein );
+   sph_skein512( &ctx_skein, hash, 64 );
+   sph_skein512_close( &ctx_skein, hash );

 	if (hash[0] & 0x8)
 	{
--- a/algo/hodl/aes.c
+++ b/algo/hodl/aes.c
@@ -83,7 +83,7 @@ void ExpandAESKey256(__m128i *keys, const __m128i *KeyBuf)
    keys[14] = tmp1;
 }

-#ifdef __SSE4_2__
+#if defined(__SSE4_2__)
 //#ifdef __AVX__

 #define AESENC(i,j) \
@@ -151,7 +151,7 @@ void AES256CBC(__m128i** data, const __m128i** next, __m128i ExpandedKey[][16],
    }
 }

-#else    // NO SSE4.2
+#else    // NO AVX

 static inline __m128i AES256Core(__m128i State, const __m128i *ExpandedKey)
 {
--- a/algo/hodl/hodl-gate.c
+++ b/algo/hodl/hodl-gate.c
@@ -15,11 +15,6 @@ pthread_barrier_t hodl_barrier;
 // need to be passed.
 unsigned char *hodl_scratchbuf = NULL;

-void hodl_set_target( struct work* work, double diff )
-{
-     diff_to_target(work->target, diff / 8388608.0 );
-}
-
 void hodl_le_build_stratum_request( char* req, struct work* work,
                                    struct stratum_ctx *sctx ) 
 {
@@ -166,11 +161,10 @@ bool register_hodl_algo( algo_gate_t* gate )
 //     return false;
 //  }
  pthread_barrier_init( &hodl_barrier, NULL, opt_n_threads );
-  gate->optimizations         = AES_OPT | SSE42_OPT | AVX2_OPT;
+  gate->optimizations         = AES_OPT | AVX_OPT | AVX2_OPT;
  gate->scanhash              = (void*)&hodl_scanhash;
  gate->get_new_work          = (void*)&hodl_get_new_work;
  gate->longpoll_rpc_call     = (void*)&hodl_longpoll_rpc_call;
-  gate->set_target            = (void*)&hodl_set_target;
  gate->build_stratum_request = (void*)&hodl_le_build_stratum_request;
  gate->malloc_txs_request    = (void*)&hodl_malloc_txs_request;
  gate->build_block_header    = (void*)&hodl_build_block_header;
@@ -179,6 +173,7 @@ bool register_hodl_algo( algo_gate_t* gate )
  gate->work_cmp_size         = 76;
  hodl_scratchbuf = (unsigned char*)malloc( 1 << 30 );
  allow_getwork = false;
+  opt_target_factor = 8388608.0;
  return ( hodl_scratchbuf != NULL );
 }

--- a/algo/hodl/hodl-wolf.c
+++ b/algo/hodl/hodl-wolf.c
@@ -17,7 +17,7 @@ void GenerateGarbageCore( CacheEntry *Garbage, int ThreadID, int ThreadCount,
    const uint32_t StartChunk = ThreadID * Chunk;
    const uint32_t EndChunk   = StartChunk + Chunk;

-#ifdef __SSE4_2__
+#if defined(__SSE4_2__)
 //#ifdef __AVX__
    uint64_t* TempBufs[ SHA512_PARALLEL_N ] ;
    uint64_t* desination[ SHA512_PARALLEL_N ];
@@ -64,7 +64,7 @@ void Rev256(uint32_t *Dest, const uint32_t *Src)
 int scanhash_hodl_wolf( struct work* work, uint32_t max_nonce,
                        uint64_t *hashes_done, struct thr_info *mythr )
 {
-#ifdef __SSE4_2__
+#if defined(__SSE4_2__)
 //#ifdef __AVX__
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
@@ -140,7 +140,7 @@ int scanhash_hodl_wolf( struct work* work, uint32_t max_nonce,
    return(0);


-#else  // no SSE4.2
+#else  // no AVX

    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
@@ -148,6 +148,7 @@ int scanhash_hodl_wolf( struct work* work, uint32_t max_nonce,
    CacheEntry *Garbage = (CacheEntry*)hodl_scratchbuf;
    CacheEntry Cache;
    uint32_t CollisionCount = 0;
+    int threadNumber = mythr->id;

    swab32_array( BlockHdr, pdata, 20 );
        // Search for pattern in psuedorandom data      
@@ -205,7 +206,7 @@ int scanhash_hodl_wolf( struct work* work, uint32_t max_nonce,
    *hashes_done = CollisionCount;
    return(0);

-#endif  // SSE4.2 else
+#endif  // AVX else

 }

--- a/algo/hodl/sha512-avx.h
+++ b/algo/hodl/sha512-avx.h
@@ -23,6 +23,7 @@ typedef struct
   __m256i h[8];
   __m256i w[80];
 #elif defined(__SSE4_2__)
+//#elif defined(__AVX__)
   __m128i h[8];
   __m128i w[80];
 #else
@@ -32,7 +33,8 @@ typedef struct

 #ifdef __AVX2__
 #define SHA512_PARALLEL_N 8
-#elif defined(__SSE$_2__)
+#elif defined(__SSE4_2__)
+//#elif defined(__AVX__)
 #define SHA512_PARALLEL_N 4
 #else
 #define SHA512_PARALLEL_N 1   // dummy value
--- a/algo/hodl/sha512_avx.c
+++ b/algo/hodl/sha512_avx.c
@@ -1,6 +1,6 @@
 #ifndef __AVX2__

-#ifdef __SSE4_2__
+#if defined(__SSE4_2__)
 //#ifdef __AVX__

 //Dependencies
--- a/algo/hodl/wolf-aes.h
+++ b/algo/hodl/wolf-aes.h
@@ -6,7 +6,7 @@

 void ExpandAESKey256(__m128i *keys, const __m128i *KeyBuf);

-#ifdef __SSE4_2__
+#if defined(__SSE4_2__)
 //#ifdef __AVX__

 #define AES_PARALLEL_N 8
--- a/algo/jh/jh-hash-4way.c
+++ b/algo/jh/jh-hash-4way.c
@@ -92,9 +92,41 @@ extern "C"{

 #endif

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define Sb_8W(x0, x1, x2, x3, c) \
+do { \
+   __m512i cc = _mm512_set1_epi64( c ); \
+    x3 = mm512_not( x3 ); \
+    x0 = _mm512_xor_si512( x0, _mm512_andnot_si512( x2, cc ) ); \
+    tmp = _mm512_xor_si512( cc, _mm512_and_si512( x0, x1 ) ); \
+    x0 = _mm512_xor_si512( x0, _mm512_and_si512( x2, x3 ) ); \
+    x3 = _mm512_xor_si512( x3, _mm512_andnot_si512( x1, x2 ) ); \
+    x1 = _mm512_xor_si512( x1, _mm512_and_si512( x0, x2 ) ); \
+    x2 = _mm512_xor_si512( x2, _mm512_andnot_si512( x3, x0 ) ); \
+    x0 = _mm512_xor_si512( x0, _mm512_or_si512( x1, x3 ) ); \
+    x3 = _mm512_xor_si512( x3, _mm512_and_si512( x1, x2 ) ); \
+    x1 = _mm512_xor_si512( x1, _mm512_and_si512( tmp, x0 ) ); \
+    x2 = _mm512_xor_si512( x2, tmp ); \
+} while (0)
+
+#define Lb_8W(x0, x1, x2, x3, x4, x5, x6, x7) \
+do { \
+    x4 = _mm512_xor_si512( x4, x1 ); \
+    x5 = _mm512_xor_si512( x5, x2 ); \
+    x6 = _mm512_xor_si512( x6, _mm512_xor_si512( x3, x0 ) ); \
+    x7 = _mm512_xor_si512( x7, x0 ); \
+    x0 = _mm512_xor_si512( x0, x5 ); \
+    x1 = _mm512_xor_si512( x1, x6 ); \
+    x2 = _mm512_xor_si512( x2, _mm512_xor_si512( x7, x4 ) ); \
+    x3 = _mm512_xor_si512( x3, x4 ); \
+} while (0)
+
+#endif
+
 #define Sb(x0, x1, x2, x3, c) \
 do { \
-   __m256i cc = _mm256_set_epi64x( c, c, c, c ); \
+   __m256i cc = _mm256_set1_epi64x( c ); \
    x3 = mm256_not( x3 ); \
    x0 = _mm256_xor_si256( x0, _mm256_andnot_si256( x2, cc ) ); \
    tmp = _mm256_xor_si256( cc, _mm256_and_si256( x0, x1 ) ); \
@@ -226,6 +258,48 @@ static const sph_u64 C[] = {
 			x4 ## l, x5 ## l, x6 ## l, x7 ## l); \
 	} while (0)

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define S_8W(x0, x1, x2, x3, cb, r)   do { \
+      Sb_8W(x0 ## h, x1 ## h, x2 ## h, x3 ## h, cb ## hi(r)); \
+      Sb_8W(x0 ## l, x1 ## l, x2 ## l, x3 ## l, cb ## lo(r)); \
+   } while (0)
+
+#define L_8W(x0, x1, x2, x3, x4, x5, x6, x7)   do { \
+      Lb_8W(x0 ## h, x1 ## h, x2 ## h, x3 ## h, \
+         x4 ## h, x5 ## h, x6 ## h, x7 ## h); \
+      Lb_8W(x0 ## l, x1 ## l, x2 ## l, x3 ## l, \
+         x4 ## l, x5 ## l, x6 ## l, x7 ## l); \
+   } while (0)
+
+#define Wz_8W(x, c, n) \
+do { \
+   __m512i t = _mm512_slli_epi64( _mm512_and_si512(x ## h, (c)), (n) ); \
+   x ## h = _mm512_or_si512( _mm512_and_si512( \
+                                _mm512_srli_epi64(x ## h, (n)), (c)), t ); \
+   t = _mm512_slli_epi64( _mm512_and_si512(x ## l, (c)), (n) ); \
+   x ## l = _mm512_or_si512( _mm512_and_si512((x ## l >> (n)), (c)), t ); \
+} while (0)
+
+#define W80(x)   Wz_8W(x, m512_const1_64( 0x5555555555555555 ),  1 )
+#define W81(x)   Wz_8W(x, m512_const1_64( 0x3333333333333333 ),  2 )
+#define W82(x)   Wz_8W(x, m512_const1_64( 0x0F0F0F0F0F0F0F0F ),  4 )
+#define W83(x)   Wz_8W(x, m512_const1_64( 0x00FF00FF00FF00FF ),  8 ) 
+#define W84(x)   Wz_8W(x, m512_const1_64( 0x0000FFFF0000FFFF ), 16 )
+#define W85(x)   Wz_8W(x, m512_const1_64( 0x00000000FFFFFFFF ), 32 )
+#define W86(x) \
+do { \
+   __m512i t = x ## h; \
+   x ## h = x ## l; \
+   x ## l = t; \
+} while (0)
+
+#define DECL_STATE_8W \
+   __m512i h0h, h1h, h2h, h3h, h4h, h5h, h6h, h7h; \
+   __m512i h0l, h1l, h2l, h3l, h4l, h5l, h6l, h7l; \
+   __m512i tmp;
+
+#endif

 #define Wz(x, c, n) \
 do { \
@@ -236,28 +310,12 @@ do { \
   x ## l = _mm256_or_si256( _mm256_and_si256((x ## l >> (n)), (c)), t ); \
 } while (0)

-
-/*
-#define Wz(x, c, n)   do { \
-		sph_u64 t = (x ## h & (c)) << (n); \
-		x ## h = ((x ## h >> (n)) & (c)) | t; \
-		t = (x ## l & (c)) << (n); \
-		x ## l = ((x ## l >> (n)) & (c)) | t; \
-	} while (0)
-*/
-
-#define W0(x)   Wz(x, _mm256_set_epi64x( 0x5555555555555555, \
-       0x5555555555555555, 0x5555555555555555, 0x5555555555555555 ), 1 )
-#define W1(x)   Wz(x, _mm256_set_epi64x( 0x3333333333333333, \
-       0x3333333333333333, 0x3333333333333333, 0x3333333333333333 ), 2 )
-#define W2(x)   Wz(x, _mm256_set_epi64x( 0x0F0F0F0F0F0F0F0F, \
-       0x0F0F0F0F0F0F0F0F, 0x0F0F0F0F0F0F0F0F, 0x0F0F0F0F0F0F0F0F ), 4 )
-#define W3(x)   Wz(x, _mm256_set_epi64x( 0x00FF00FF00FF00FF, \
-       0x00FF00FF00FF00FF, 0x00FF00FF00FF00FF, 0x00FF00FF00FF00FF ), 8 ) 
-#define W4(x)   Wz(x, _mm256_set_epi64x( 0x0000FFFF0000FFFF, \
-       0x0000FFFF0000FFFF, 0x0000FFFF0000FFFF, 0x0000FFFF0000FFFF ), 16 )
-#define W5(x)   Wz(x, _mm256_set_epi64x( 0x00000000FFFFFFFF, \
-       0x00000000FFFFFFFF, 0x00000000FFFFFFFF, 0x00000000FFFFFFFF ), 32 )
+#define W0(x)   Wz(x, m256_const1_64( 0x5555555555555555 ),  1 )
+#define W1(x)   Wz(x, m256_const1_64( 0x3333333333333333 ),  2 )
+#define W2(x)   Wz(x, m256_const1_64( 0x0F0F0F0F0F0F0F0F ),  4 )
+#define W3(x)   Wz(x, m256_const1_64( 0x00FF00FF00FF00FF ),  8 ) 
+#define W4(x)   Wz(x, m256_const1_64( 0x0000FFFF0000FFFF ), 16 )
+#define W5(x)   Wz(x, m256_const1_64( 0x00000000FFFFFFFF ), 32 )
 #define W6(x) \
 do { \
   __m256i t = x ## h; \
@@ -265,25 +323,12 @@ do { \
   x ## l = t; \
 } while (0)

-/*
-#define W0(x)   Wz(x, SPH_C64(0x5555555555555555),  1)
-#define W1(x)   Wz(x, SPH_C64(0x3333333333333333),  2)
-#define W2(x)   Wz(x, SPH_C64(0x0F0F0F0F0F0F0F0F),  4)
-#define W3(x)   Wz(x, SPH_C64(0x00FF00FF00FF00FF),  8)
-#define W4(x)   Wz(x, SPH_C64(0x0000FFFF0000FFFF), 16)
-#define W5(x)   Wz(x, SPH_C64(0x00000000FFFFFFFF), 32)
-#define W6(x)   do { \
-		sph_u64 t = x ## h; \
-		x ## h = x ## l; \
-		x ## l = t; \
-	} while (0)
-*/
-
 #define DECL_STATE \
 	__m256i h0h, h1h, h2h, h3h, h4h, h5h, h6h, h7h; \
 	__m256i h0l, h1l, h2l, h3l, h4l, h5l, h6l, h7l; \
 	__m256i tmp;

+
 #define READ_STATE(state)   do { \
 		h0h = (state)->H[ 0]; \
 		h0l = (state)->H[ 1]; \
@@ -322,6 +367,38 @@ do { \
 		(state)->H[15] = h7l; \
 	} while (0)

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define INPUT_BUF1_8W \
+   __m512i m0h = buf[0]; \
+   __m512i m0l = buf[1]; \
+   __m512i m1h = buf[2]; \
+   __m512i m1l = buf[3]; \
+   __m512i m2h = buf[4]; \
+   __m512i m2l = buf[5]; \
+   __m512i m3h = buf[6]; \
+   __m512i m3l = buf[7]; \
+   h0h = _mm512_xor_si512( h0h, m0h ); \
+   h0l = _mm512_xor_si512( h0l, m0l ); \
+   h1h = _mm512_xor_si512( h1h, m1h ); \
+   h1l = _mm512_xor_si512( h1l, m1l ); \
+   h2h = _mm512_xor_si512( h2h, m2h ); \
+   h2l = _mm512_xor_si512( h2l, m2l ); \
+   h3h = _mm512_xor_si512( h3h, m3h ); \
+   h3l = _mm512_xor_si512( h3l, m3l ); \
+
+#define INPUT_BUF2_8W \
+   h4h = _mm512_xor_si512( h4h, m0h ); \
+   h4l = _mm512_xor_si512( h4l, m0l ); \
+   h5h = _mm512_xor_si512( h5h, m1h ); \
+   h5l = _mm512_xor_si512( h5l, m1l ); \
+   h6h = _mm512_xor_si512( h6h, m2h ); \
+   h6l = _mm512_xor_si512( h6l, m2l ); \
+   h7h = _mm512_xor_si512( h7h, m3h ); \
+   h7l = _mm512_xor_si512( h7l, m3l ); \
+
+#endif
+
 #define INPUT_BUF1 \
 	__m256i m0h = buf[0]; \
 	__m256i m0l = buf[1]; \
@@ -331,14 +408,14 @@ do { \
 	__m256i m2l = buf[5]; \
 	__m256i m3h = buf[6]; \
 	__m256i m3l = buf[7]; \
-        h0h = _mm256_xor_si256( h0h, m0h ); \
-        h0l = _mm256_xor_si256( h0l, m0l ); \
-        h1h = _mm256_xor_si256( h1h, m1h ); \
-        h1l = _mm256_xor_si256( h1l, m1l ); \
-        h2h = _mm256_xor_si256( h2h, m2h ); \
-        h2l = _mm256_xor_si256( h2l, m2l ); \
-        h3h = _mm256_xor_si256( h3h, m3h ); \
-        h3l = _mm256_xor_si256( h3l, m3l ); \
+   h0h = _mm256_xor_si256( h0h, m0h ); \
+   h0l = _mm256_xor_si256( h0l, m0l ); \
+   h1h = _mm256_xor_si256( h1h, m1h ); \
+   h1l = _mm256_xor_si256( h1l, m1l ); \
+   h2h = _mm256_xor_si256( h2h, m2h ); \
+   h2l = _mm256_xor_si256( h2l, m2l ); \
+   h3h = _mm256_xor_si256( h3h, m3h ); \
+   h3l = _mm256_xor_si256( h3l, m3l ); \

 #define INPUT_BUF2 \
   h4h = _mm256_xor_si256( h4h, m0h ); \
@@ -350,6 +427,7 @@ do { \
   h7h = _mm256_xor_si256( h7h, m3h ); \
   h7l = _mm256_xor_si256( h7l, m3l ); \

+
 static const sph_u64 IV256[] = {
 	C64e(0xeb98a3412c20d3eb), C64e(0x92cdbe7b9cb245c1),
 	C64e(0x1c93519160d4c7fa), C64e(0x260082d67e508a03),
@@ -376,6 +454,22 @@ static const sph_u64 IV512[] = {
 #else


+#endif
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define SL_8W(ro)   SLu_8W(r + ro, ro)
+
+#define SLu_8W(r, ro)   do { \
+      S_8W(h0, h2, h4, h6, Ceven_, r); \
+      S_8W(h1, h3, h5, h7, Codd_, r); \
+      L_8W(h0, h2, h4, h6, h1, h3, h5, h7); \
+      W8 ## ro(h1); \
+      W8 ## ro(h3); \
+      W8 ## ro(h5); \
+      W8 ## ro(h7); \
+   } while (0)
+
 #endif

 #define SL(ro)   SLu(r + ro, ro)
@@ -399,6 +493,23 @@ static const sph_u64 IV512[] = {
 * loop.
 */

+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define E8_8W   do { \
+      unsigned r; \
+      for (r = 0; r < 42; r += 7) { \
+         SL_8W(0); \
+         SL_8W(1); \
+         SL_8W(2); \
+         SL_8W(3); \
+         SL_8W(4); \
+         SL_8W(5); \
+         SL_8W(6); \
+      } \
+   } while (0)
+
+#endif
+
 #define E8   do { \
 		unsigned r; \
 		for (r = 0; r < 42; r += 7) { \
@@ -425,51 +536,100 @@ static const sph_u64 IV512[] = {
 * On a "true 64-bit" architecture, we can unroll at will.
 */

-#define E8   do { \
-		SLu( 0, 0); \
-		SLu( 1, 1); \
-		SLu( 2, 2); \
-		SLu( 3, 3); \
-		SLu( 4, 4); \
-		SLu( 5, 5); \
-		SLu( 6, 6); \
-		SLu( 7, 0); \
-		SLu( 8, 1); \
-		SLu( 9, 2); \
-		SLu(10, 3); \
-		SLu(11, 4); \
-		SLu(12, 5); \
-		SLu(13, 6); \
-		SLu(14, 0); \
-		SLu(15, 1); \
-		SLu(16, 2); \
-		SLu(17, 3); \
-		SLu(18, 4); \
-		SLu(19, 5); \
-		SLu(20, 6); \
-		SLu(21, 0); \
-		SLu(22, 1); \
-		SLu(23, 2); \
-		SLu(24, 3); \
-		SLu(25, 4); \
-		SLu(26, 5); \
-		SLu(27, 6); \
-		SLu(28, 0); \
-		SLu(29, 1); \
-		SLu(30, 2); \
-		SLu(31, 3); \
-		SLu(32, 4); \
-		SLu(33, 5); \
-		SLu(34, 6); \
-		SLu(35, 0); \
-		SLu(36, 1); \
-		SLu(37, 2); \
-		SLu(38, 3); \
-		SLu(39, 4); \
-		SLu(40, 5); \
-		SLu(41, 6); \
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+#define E8_8W   do { \
+		SLu_8W( 0, 0); \
+		SLu_8W( 1, 1); \
+		SLu_8W( 2, 2); \
+		SLu_8W( 3, 3); \
+		SLu_8W( 4, 4); \
+		SLu_8W( 5, 5); \
+		SLu_8W( 6, 6); \
+		SLu_8W( 7, 0); \
+		SLu_8W( 8, 1); \
+		SLu_8W( 9, 2); \
+		SLu_8W(10, 3); \
+		SLu_8W(11, 4); \
+		SLu_8W(12, 5); \
+		SLu_8W(13, 6); \
+		SLu_8W(14, 0); \
+		SLu_8W(15, 1); \
+		SLu_8W(16, 2); \
+		SLu_8W(17, 3); \
+		SLu_8W(18, 4); \
+		SLu_8W(19, 5); \
+		SLu_8W(20, 6); \
+		SLu_8W(21, 0); \
+		SLu_8W(22, 1); \
+		SLu_8W(23, 2); \
+		SLu_8W(24, 3); \
+		SLu_8W(25, 4); \
+		SLu_8W(26, 5); \
+		SLu_8W(27, 6); \
+		SLu_8W(28, 0); \
+		SLu_8W(29, 1); \
+		SLu_8W(30, 2); \
+		SLu_8W(31, 3); \
+		SLu_8W(32, 4); \
+		SLu_8W(33, 5); \
+		SLu_8W(34, 6); \
+		SLu_8W(35, 0); \
+		SLu_8W(36, 1); \
+		SLu_8W(37, 2); \
+		SLu_8W(38, 3); \
+		SLu_8W(39, 4); \
+		SLu_8W(40, 5); \
+		SLu_8W(41, 6); \
 	} while (0)

+#endif  // AVX512
+
+#define E8   do { \
+      SLu( 0, 0); \
+      SLu( 1, 1); \
+      SLu( 2, 2); \
+      SLu( 3, 3); \
+      SLu( 4, 4); \
+      SLu( 5, 5); \
+      SLu( 6, 6); \
+      SLu( 7, 0); \
+      SLu( 8, 1); \
+      SLu( 9, 2); \
+      SLu(10, 3); \
+      SLu(11, 4); \
+      SLu(12, 5); \
+      SLu(13, 6); \
+      SLu(14, 0); \
+      SLu(15, 1); \
+      SLu(16, 2); \
+      SLu(17, 3); \
+      SLu(18, 4); \
+      SLu(19, 5); \
+      SLu(20, 6); \
+      SLu(21, 0); \
+      SLu(22, 1); \
+      SLu(23, 2); \
+      SLu(24, 3); \
+      SLu(25, 4); \
+      SLu(26, 5); \
+      SLu(27, 6); \
+      SLu(28, 0); \
+      SLu(29, 1); \
+      SLu(30, 2); \
+      SLu(31, 3); \
+      SLu(32, 4); \
+      SLu(33, 5); \
+      SLu(34, 6); \
+      SLu(35, 0); \
+      SLu(36, 1); \
+      SLu(37, 2); \
+      SLu(38, 3); \
+      SLu(39, 4); \
+      SLu(40, 5); \
+      SLu(41, 6); \
+   } while (0)
+
 #else


@@ -477,13 +637,200 @@ static const sph_u64 IV512[] = {

 #endif

-static void
-jh_4way_init( jh_4way_context *sc, const void *iv )
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+void jh256_8way_init( jh_8way_context *sc )
 {
-    uint64_t *v = (uint64_t*)iv;
-    
-    for ( int i = 0; i < 16; i++ )
-        sc->H[i] = _mm256_set_epi64x( v[i], v[i], v[i], v[i] );
+    // bswapped IV256
+    sc->H[ 0] = m512_const1_64( 0xebd3202c41a398eb );
+    sc->H[ 1] = m512_const1_64( 0xc145b29c7bbecd92 );
+    sc->H[ 2] = m512_const1_64( 0xfac7d4609151931c );
+    sc->H[ 3] = m512_const1_64( 0x038a507ed6820026 );
+    sc->H[ 4] = m512_const1_64( 0x45b92677269e23a4 );
+    sc->H[ 5] = m512_const1_64( 0x77941ad4481afbe0 );
+    sc->H[ 6] = m512_const1_64( 0x7a176b0226abb5cd );
+    sc->H[ 7] = m512_const1_64( 0xa82fff0f4224f056 );
+    sc->H[ 8] = m512_const1_64( 0x754d2e7f8996a371 );
+    sc->H[ 9] = m512_const1_64( 0x62e27df70849141d );
+    sc->H[10] = m512_const1_64( 0x948f2476f7957627 );
+    sc->H[11] = m512_const1_64( 0x6c29804757b6d587 );
+    sc->H[12] = m512_const1_64( 0x6c0d8eac2d275e5c );
+    sc->H[13] = m512_const1_64( 0x0f7a0557c6508451 );
+    sc->H[14] = m512_const1_64( 0xea12247067d3e47b );
+    sc->H[15] = m512_const1_64( 0x69d71cd313abe389 );
+    sc->ptr = 0;
+    sc->block_count = 0;
+}
+
+void jh512_8way_init( jh_8way_context *sc )
+{
+    // bswapped IV512
+    sc->H[ 0] = m512_const1_64( 0x17aa003e964bd16f );
+    sc->H[ 1] = m512_const1_64( 0x43d5157a052e6a63 );
+    sc->H[ 2] = m512_const1_64( 0x0bef970c8d5e228a );
+    sc->H[ 3] = m512_const1_64( 0x61c3b3f2591234e9 );
+    sc->H[ 4] = m512_const1_64( 0x1e806f53c1a01d89 );
+    sc->H[ 5] = m512_const1_64( 0x806d2bea6b05a92a );
+    sc->H[ 6] = m512_const1_64( 0xa6ba7520dbcc8e58 );
+    sc->H[ 7] = m512_const1_64( 0xf73bf8ba763a0fa9 );
+    sc->H[ 8] = m512_const1_64( 0x694ae34105e66901 );
+    sc->H[ 9] = m512_const1_64( 0x5ae66f2e8e8ab546 );
+    sc->H[10] = m512_const1_64( 0x243c84c1d0a74710 );
+    sc->H[11] = m512_const1_64( 0x99c15a2db1716e3b );
+    sc->H[12] = m512_const1_64( 0x56f8b19decf657cf );
+    sc->H[13] = m512_const1_64( 0x56b116577c8806a7 );
+    sc->H[14] = m512_const1_64( 0xfb1785e6dffcc2e3 );
+    sc->H[15] = m512_const1_64( 0x4bdd8ccc78465a54 );
+    sc->ptr = 0;
+    sc->block_count = 0;
+}
+
+static void
+jh_8way_core( jh_8way_context *sc, const void *data, size_t len )
+{
+    __m512i *buf;
+    __m512i *vdata = (__m512i*)data;
+   const int buf_size = 64;   // 64 * _m512i
+   size_t ptr;
+   DECL_STATE_8W
+
+   buf = sc->buf;
+   ptr = sc->ptr;
+
+   if ( len < (buf_size - ptr) )
+   {
+       memcpy_512( buf + (ptr>>3), vdata, len>>3 );
+       ptr += len;
+       sc->ptr = ptr;
+       return;
+   }
+
+   READ_STATE(sc);
+   while ( len > 0 )
+   {
+       size_t clen;
+       clen = buf_size - ptr;
+       if ( clen > len )
+          clen = len;
+
+       memcpy_512( buf + (ptr>>3), vdata, clen>>3 );
+       ptr += clen;
+       vdata += (clen>>3);
+       len -= clen;
+       if ( ptr == buf_size )
+       {
+          INPUT_BUF1_8W;
+          E8_8W;
+          INPUT_BUF2_8W;
+          sc->block_count ++;
+          ptr = 0;
+       }
+   }
+   WRITE_STATE(sc);
+   sc->ptr = ptr;
+}
+
+static void
+jh_8way_close( jh_8way_context *sc, unsigned ub, unsigned n, void *dst,
+               size_t out_size_w32, const void *iv )
+{
+   __m512i buf[16*4];
+   __m512i *dst512 = (__m512i*)dst;
+   size_t numz, u;
+   sph_u64 l0, l1, l0e, l1e;
+
+   buf[0] = m512_const1_64( 0x80ULL );
+
+   if ( sc->ptr == 0 )
+       numz = 48;
+   else
+       numz = 112 - sc->ptr;
+
+   memset_zero_512( buf+1, (numz>>3) - 1 );
+
+   l0 = SPH_T64(sc->block_count << 9) + (sc->ptr << 3);
+   l1 = SPH_T64(sc->block_count >> 55);
+   sph_enc64be( &l0e, l0 );
+   sph_enc64be( &l1e, l1 );
+   *(buf + (numz>>3)    ) = _mm512_set1_epi64( l1e );
+   *(buf + (numz>>3) + 1) = _mm512_set1_epi64( l0e );
+
+   jh_8way_core( sc, buf, numz + 16 );
+
+   for ( u=0; u < 8; u++ )
+       buf[u] = sc->H[u+8];
+
+    memcpy_512( dst512, buf, 8 );
+}
+
+void
+jh256_8way_update(void *cc, const void *data, size_t len)
+{
+   jh_8way_core(cc, data, len);
+}
+
+void
+jh256_8way_close(void *cc, void *dst)
+{
+   jh_8way_close(cc, 0, 0, dst, 8, IV256);
+}
+
+void
+jh512_8way_update(void *cc, const void *data, size_t len)
+{
+   jh_8way_core(cc, data, len);
+}
+
+void
+jh512_8way_close(void *cc, void *dst)
+{
+   jh_8way_close(cc, 0, 0, dst, 16, IV512);
+}
+
+#endif
+
+void jh256_4way_init( jh_4way_context *sc )
+{
+    // bswapped IV256
+    sc->H[ 0] = m256_const1_64( 0xebd3202c41a398eb );
+    sc->H[ 1] = m256_const1_64( 0xc145b29c7bbecd92 );
+    sc->H[ 2] = m256_const1_64( 0xfac7d4609151931c );
+    sc->H[ 3] = m256_const1_64( 0x038a507ed6820026 );
+    sc->H[ 4] = m256_const1_64( 0x45b92677269e23a4 );
+    sc->H[ 5] = m256_const1_64( 0x77941ad4481afbe0 );
+    sc->H[ 6] = m256_const1_64( 0x7a176b0226abb5cd );
+    sc->H[ 7] = m256_const1_64( 0xa82fff0f4224f056 );
+    sc->H[ 8] = m256_const1_64( 0x754d2e7f8996a371 );
+    sc->H[ 9] = m256_const1_64( 0x62e27df70849141d );
+    sc->H[10] = m256_const1_64( 0x948f2476f7957627 );
+    sc->H[11] = m256_const1_64( 0x6c29804757b6d587 );
+    sc->H[12] = m256_const1_64( 0x6c0d8eac2d275e5c );
+    sc->H[13] = m256_const1_64( 0x0f7a0557c6508451 );
+    sc->H[14] = m256_const1_64( 0xea12247067d3e47b );
+    sc->H[15] = m256_const1_64( 0x69d71cd313abe389 );
+    sc->ptr = 0;
+    sc->block_count = 0;
+}
+
+void jh512_4way_init( jh_4way_context *sc )
+{
+    // bswapped IV512
+    sc->H[ 0] = m256_const1_64( 0x17aa003e964bd16f );
+    sc->H[ 1] = m256_const1_64( 0x43d5157a052e6a63 );
+    sc->H[ 2] = m256_const1_64( 0x0bef970c8d5e228a );
+    sc->H[ 3] = m256_const1_64( 0x61c3b3f2591234e9 );
+    sc->H[ 4] = m256_const1_64( 0x1e806f53c1a01d89 );
+    sc->H[ 5] = m256_const1_64( 0x806d2bea6b05a92a );
+    sc->H[ 6] = m256_const1_64( 0xa6ba7520dbcc8e58 );
+    sc->H[ 7] = m256_const1_64( 0xf73bf8ba763a0fa9 );
+    sc->H[ 8] = m256_const1_64( 0x694ae34105e66901 );
+    sc->H[ 9] = m256_const1_64( 0x5ae66f2e8e8ab546 );
+    sc->H[10] = m256_const1_64( 0x243c84c1d0a74710 );
+    sc->H[11] = m256_const1_64( 0x99c15a2db1716e3b );
+    sc->H[12] = m256_const1_64( 0x56f8b19decf657cf );
+    sc->H[13] = m256_const1_64( 0x56b116577c8806a7 );
+    sc->H[14] = m256_const1_64( 0xfb1785e6dffcc2e3 );
+    sc->H[15] = m256_const1_64( 0x4bdd8ccc78465a54 );
    sc->ptr = 0;
    sc->block_count = 0;
 }
@@ -542,7 +889,7 @@ jh_4way_close( jh_4way_context *sc, unsigned ub, unsigned n, void *dst,
   size_t numz, u;
   sph_u64 l0, l1, l0e, l1e;

-   buf[0] = _mm256_set_epi64x( 0x80, 0x80, 0x80, 0x80 );
+   buf[0] = m256_const1_64( 0x80ULL );

   if ( sc->ptr == 0 )
       numz = 48;
@@ -555,8 +902,8 @@ jh_4way_close( jh_4way_context *sc, unsigned ub, unsigned n, void *dst,
   l1 = SPH_T64(sc->block_count >> 55);
   sph_enc64be( &l0e, l0 );
   sph_enc64be( &l1e, l1 );
-   *(buf + (numz>>3)    ) = _mm256_set_epi64x( l1e, l1e, l1e, l1e );
-   *(buf + (numz>>3) + 1) = _mm256_set_epi64x( l0e, l0e, l0e, l0e ); 
+   *(buf + (numz>>3)    ) = _mm256_set1_epi64x( l1e );
+   *(buf + (numz>>3) + 1) = _mm256_set1_epi64x( l0e ); 

   jh_4way_core( sc, buf, numz + 16 );

@@ -567,13 +914,7 @@ jh_4way_close( jh_4way_context *sc, unsigned ub, unsigned n, void *dst,
 }

 void
-jh256_4way_init(void *cc)
-{
-	jh_4way_init(cc, IV256);
-}
-
-void
-jh256_4way(void *cc, const void *data, size_t len)
+jh256_4way_update(void *cc, const void *data, size_t len)
 {
 	jh_4way_core(cc, data, len);
 }
@@ -585,13 +926,7 @@ jh256_4way_close(void *cc, void *dst)
 }

 void
-jh512_4way_init(void *cc)
-{
-	jh_4way_init(cc, IV512);
-}
-
-void
-jh512_4way(void *cc, const void *data, size_t len)
+jh512_4way_update(void *cc, const void *data, size_t len)
 {
 	jh_4way_core(cc, data, len);
 }
@@ -602,6 +937,7 @@ jh512_4way_close(void *cc, void *dst)
 	jh_4way_close(cc, 0, 0, dst, 16, IV512);
 }

+
 #ifdef __cplusplus
 }
 #endif
--- a/algo/jh/jh-hash-4way.h
+++ b/algo/jh/jh-hash-4way.h
@@ -60,34 +60,55 @@ extern "C"{
 * can be cloned by copying the context (e.g. with a simple
 * <code>memcpy()</code>).
 */
+
+ 
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
 typedef struct {
-    __m256i buf[8] __attribute__ ((aligned (64)));
+    __m512i buf[8];
+    __m512i H[16];
+    size_t ptr;
+    uint64_t block_count;
+} jh_8way_context __attribute__ ((aligned (128)));
+
+typedef jh_8way_context jh256_8way_context;
+
+typedef jh_8way_context jh512_8way_context;
+
+void jh256_8way_init( jh_8way_context *sc);
+
+void jh256_8way_update(void *cc, const void *data, size_t len);
+
+void jh256_8way_close(void *cc, void *dst);
+
+void jh512_8way_init( jh_8way_context *sc );
+
+void jh512_8way_update(void *cc, const void *data, size_t len);
+
+void jh512_8way_close(void *cc, void *dst);
+
+#endif
+
+typedef struct {
+    __m256i buf[8];
    __m256i H[16];
    size_t ptr;
    uint64_t block_count;
-/*
-	unsigned char buf[64]; 
-	size_t ptr;
-	union {
-		sph_u64 wide[16];
-	} H;
-	sph_u64 block_count;
-*/
-} jh_4way_context;
+} jh_4way_context __attribute__ ((aligned (128)));

 typedef jh_4way_context jh256_4way_context;

 typedef jh_4way_context jh512_4way_context;

-void jh256_4way_init(void *cc);
+void jh256_4way_init( jh_4way_context *sc);

-void jh256_4way(void *cc, const void *data, size_t len);
+void jh256_4way_update(void *cc, const void *data, size_t len);

 void jh256_4way_close(void *cc, void *dst);

-void jh512_4way_init(void *cc);
+void jh512_4way_init( jh_4way_context *sc );

-void jh512_4way(void *cc, const void *data, size_t len);
+void jh512_4way_update(void *cc, const void *data, size_t len);

 void jh512_4way_close(void *cc, void *dst);

@@ -95,6 +116,6 @@ void jh512_4way_close(void *cc, void *dst);
 }
 #endif

-#endif
+#endif // AVX2

 #endif
--- a/algo/jh/jha-4way.c
+++ b/algo/jh/jha-4way.c
@@ -3,7 +3,6 @@
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
-//#include "avxdefs.h"

 #if defined(JHA_4WAY)

@@ -13,9 +12,6 @@
 #include "algo/keccak/keccak-hash-4way.h"
 #include "algo/groestl/aes_ni/hash-groestl.h"

-//static __thread keccak512_4way_context jha_kec_mid
-//                                   __attribute__ ((aligned (64)));
-
 void jha_hash_4way( void *out, const void *input )
 {
    uint64_t hash0[8] __attribute__ ((aligned (64)));
@@ -37,7 +33,7 @@ void jha_hash_4way( void *out, const void *input )
    keccak512_4way_context ctx_keccak;

    keccak512_4way_init( &ctx_keccak );
-    keccak512_4way( &ctx_keccak, input, 80 );
+    keccak512_4way_update( &ctx_keccak, input, 80 );
    keccak512_4way_close( &ctx_keccak, vhash );

    // Heavy & Light Pair Loop
@@ -46,7 +42,7 @@ void jha_hash_4way( void *out, const void *input )
       vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256(
               vh[0], _mm256_set1_epi64x( 1 ) ), m256_zero );

-       mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+       dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
       init_groestl( &ctx_groestl, 64 );
       update_and_final_groestl( &ctx_groestl, (char*)hash0,
                                               (char*)hash0, 512 );
@@ -59,10 +55,10 @@ void jha_hash_4way( void *out, const void *input )
       init_groestl( &ctx_groestl, 64 );
       update_and_final_groestl( &ctx_groestl, (char*)hash3,
                                               (char*)hash3, 512 );
-       mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+       intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );

       skein512_4way_init( &ctx_skein );
-       skein512_4way( &ctx_skein, vhash, 64 );
+       skein512_4way_update( &ctx_skein, vhash, 64 );
       skein512_4way_close( &ctx_skein, vhashB );

       for ( int i = 0; i < 8; i++ )
@@ -73,14 +69,12 @@ void jha_hash_4way( void *out, const void *input )
       blake512_4way_close( &ctx_blake, vhashA );

       jh512_4way_init( &ctx_jh );
-       jh512_4way( &ctx_jh, vhash, 64 );
+       jh512_4way_update( &ctx_jh, vhash, 64 );
       jh512_4way_close( &ctx_jh, vhashB );

       for ( int i = 0; i < 8; i++ )
          casti_m256i( out, i ) = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
    }
-
-//    mm256_dintrlv_4x64( out, out+32, out+64, out+96, vhash, 256 );
 }

 int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
@@ -115,12 +109,7 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
 		0
 	};

-//   for ( int i=0; i < 19; i++ )
-//      be32enc( &endiandata[i], pdata[i] );
-
-//   uint64_t *edata = (uint64_t*)endiandata;
-//   mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
-   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );

   for ( int m = 0; m < 6; m++ )
   {
@@ -130,26 +119,17 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
         do {
              *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
-//              be32enc( noncep,   n   );
-//              be32enc( noncep+2, n+1 );
-//              be32enc( noncep+4, n+2 );
-//              be32enc( noncep+6, n+3 );

              jha_hash_4way( hash, vdata );
              pdata[19] = n;

-//              for ( int i = 0; i < 4; i++ )
-//              if ( ( !( (hash+(i<<3))[7] & mask ) == 0 )
-//                  && fulltest( hash+(i<<3), ptarget ) )
              for ( int i = 0; i < 4; i++ ) if ( !( (hash7[i] & mask ) == 0 ) )
              {
-                 mm256_extr_lane_4x64( lane_hash, hash, i, 256 );
+                 extr_lane_4x64( lane_hash, hash, i, 256 );
                 if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
                 {
                    pdata[19] = n+i;
                    submit_lane_solution( work, lane_hash, mythr, i );
-//                 nonces[ num_found++ ] = n+i;
-//                 work_set_target_ratio( work, hash+(i<<3) );
                 }
              }
              n += 4;
--- a/algo/jh/jha-gate.c
+++ b/algo/jh/jha-gate.c
@@ -12,7 +12,7 @@ bool register_jha_algo( algo_gate_t* gate )
  gate->hash             = (void*)&jha_hash;
 #endif
  gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
-  gate->set_target       = (void*)&scrypt_set_target;
+  opt_target_factor = 65536.0;
  return true;
 };

--- a/algo/jh/sse2/jh.c
+++ b/algo/jh/sse2/jh.c
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Jay D Dee	82c2605d77	v3.11.1	2020-01-06 13:55:17 -05:00
Jay D Dee	3572cb53c4	v3.11.0	2020-01-02 23:54:08 -05:00
Jay D Dee	241bc26767	v3.10.6	2019-12-25 01:26:26 -05:00
Jay D Dee	c65b0ff7a6	v3.10.5	2019-12-21 13:19:29 -05:00
Jay D Dee	a17ff6f189	v3.10.2	2019-12-09 15:59:02 -05:00
Jay D Dee	73430b13b1	v3.10.1	2019-12-05 19:09:23 -05:00
Jay D Dee	40039386a0	v3.10.0	2019-12-03 12:26:11 -05:00
Jay D Dee	91ec6f1771	v3.9.11	2019-11-26 09:22:03 -05:00
Jay D Dee	a52c5eccf7	v3.9.10	2019-11-22 20:29:18 -05:00
Jay D Dee	86b889e1b0	v3.9.9.1	2019-10-24 14:11:26 -04:00
Jay D Dee	72330eb5a7	v3.9.9	2019-10-10 19:58:34 -04:00
Jay D Dee	789c8b70bc	v3.9.8.1	2019-10-01 14:17:36 -04:00
Jay D Dee	01550d94a2	v3.9.8	2019-09-26 22:37:26 -04:00
Jay D Dee	a042fb7612	v3.9.7	2019-08-03 10:39:54 -04:00
Jay D Dee	9d49e0be7a	v3.9.6.2	2019-07-30 10:16:43 -04:00
Jay D Dee	a51f59086b	v3.9.6.1	2019-07-18 19:46:57 -04:00
Jay D Dee	6f49ba09b7	v3.9.6	2019-07-17 17:54:38 -04:00
Jay D Dee	e2d5762ef2	v3.9.5.4	2019-07-15 17:00:26 -04:00
Jay D Dee	e625ed5420	v3.9.5.3	2019-07-12 10:42:38 -04:00
Jay D Dee	9abc19a30a	v3.9.5.2	2019-07-04 12:12:11 -04:00