v3.23.3

INSTALL_WINDOWS

@@ -1,158 +1,4 @@
-Instructions for compiling cpuminer-opt for Windows.
+Please consult the wiki for Windows compile instructions.
-
-These intructions are out of date. Please consult the wiki for
-the latest:
 
 https://github.com/JayDDee/cpuminer-opt/wiki/Compiling-from-source
 
-Windows compilation using Visual Studio is not supported. Mingw64 is
-used on a Linux system (bare metal or virtual machine) to cross-compile
-cpuminer-opt executable binaries for Windows.
-
-These instructions were written for Debian and Ubuntu compatible distributions
-but should work on other major distributions as well. However some of the
-package names or file paths may be different.
-
-It is assumed a Linux system is already available and running. And the user
-has enough Linux knowledge to find and install packages and follow these
-instructions.
-
-First it is a good idea to create new user specifically for cross compiling.
-It keeps all mingw stuff contained and isolated from the rest of the system.
-
-Step by step...
-
-1. Install necessary packages from the distribution's repositories.
-
-Refer to Linux compile instructions and install required packages.
-
-Additionally, install mingw-w64.
-
-sudo apt-get install mingw-w64 libz-mingw-w64-dev
-
-
-2. Create a local library directory for packages to be compiled in the next
-   step. Suggested location is $HOME/usr/lib/
-
-$ mkdir $HOME/usr/lib
-
-3. Download and build other packages for mingw that don't have a mingw64
-   version available in the repositories.
-
-Download the following source code packages from their respective and
-respected download locations, copy them to $HOME/usr/lib/ and uncompress them. 
-
-openssl: https://github.com/openssl/openssl/releases
-
-curl: https://github.com/curl/curl/releases
-
-gmp: https://gmplib.org/download/gmp/
-
-In most cases the latest version is ok but it's safest to download the same major and minor version as included in your distribution. The following uses versions from Ubuntu 20.04. Change version numbers as required.
-
-Run the following commands or follow the supplied instructions. Do not run "make install" unless you are using /usr/lib, which isn't recommended.
-
-Some instructions insist on running "make check". If make check fails it may still work, YMMV.
-
-You can speed up "make" by using all CPU cores available with "-j n" where n is the number of CPU threads you want to use.
-
-openssl:
-
-$ ./Configure mingw64 shared --cross-compile-prefix=x86_64-w64-mingw32-
-$ make
-
-Make may fail with an ld error, just ensure libcrypto-1_1-x64.dll is created.
-
-curl:
-
-$ ./configure --with-winssl --with-winidn --host=x86_64-w64-mingw32
-$ make
-
-gmp:
-
-$ ./configure --host=x86_64-w64-mingw32
-$ make
-
-4. Tweak the environment.
-
-This step is required everytime you login or the commands can be added to .bashrc.
-
-Define some local variables to point to local library.
-
-$ export LOCAL_LIB="$HOME/usr/lib"
-
-$ export LDFLAGS="-L$LOCAL_LIB/curl/lib/.libs -L$LOCAL_LIB/gmp/.libs -L$LOCAL_LIB/openssl"
-
-$ export CONFIGURE_ARGS="--with-curl=$LOCAL_LIB/curl --with-crypto=$LOCAL_LIB/openssl --host=x86_64-w64-mingw32"
-
-Adjust for gcc version:
-
-$ export GCC_MINGW_LIB="/usr/lib/gcc/x86_64-w64-mingw32/9.3-win32"
-
-Create a release directory and copy some dll files previously built. This can be done outside of cpuminer-opt and only needs to be done once. If the release directory is in cpuminer-opt directory it needs to be recreated every time a source package is decompressed.
-
-$ mkdir release
-$ cp /usr/x86_64-w64-mingw32/lib/zlib1.dll release/
-$ cp /usr/x86_64-w64-mingw32/lib/libwinpthread-1.dll release/
-$ cp $GCC_MINGW_LIB/libstdc++-6.dll release/
-$ cp $GCC_MINGW_LIB/libgcc_s_seh-1.dll release/
-$ cp $LOCAL_LIB/openssl/libcrypto-1_1-x64.dll release/
-$ cp $LOCAL_LIB/curl/lib/.libs/libcurl-4.dll release/
-
-The following steps need to be done every time a new source package is
-opened.
-
-5. Download cpuminer-opt
-
-Download the latest source code package of cpumuner-opt to your desired
-location. .zip or .tar.gz, your choice.
-
-https://github.com/JayDDee/cpuminer-opt/releases
-
-Decompress and change to the cpuminer-opt directory.
-
-6. compile
-
-Create a link to the locally compiled version of gmp.h
-
-$ ln -s $LOCAL_LIB/gmp-version/gmp.h ./gmp.h
-
-$ ./autogen.sh
-
-Configure the compiler for the CPU architecture of the host machine:
-
-CFLAGS="-O3 -march=native -Wall" ./configure $CONFIGURE_ARGS
-
-or cross compile for a specific CPU architecture:
-
-CFLAGS="-O3 -march=znver1 -Wall" ./configure $CONFIGURE_ARGS
-
-This will compile for AMD Ryzen.
-
-You can compile more generically for a set of specific CPU features if you know what features you want:
-
-CFLAGS="-O3 -maes -msse4.2 -Wall" ./configure $CONFIGURE_ARGS
-
-This will compile for an older CPU that does not have AVX.
-
-You can find several examples in README.txt
-
-If you have a CPU with more than 64 threads and Windows 7 or higher you can enable the CPU Groups feature by adding the following to CFLAGS:
-
-"-D_WIN32_WINNT=0x0601"
-
-Once you have run configure successfully run the compiler with n CPU threads:
-
-$ make -j n
-
-Copy cpuminer.exe to the release directory, compress and copy the release directory to a Windows system and run cpuminer.exe from the command line.
-
-Run cpuminer
-
-In a command windows change directories to the unzipped release folder. To get a list of all options:
-
-cpuminer.exe --help
-
-Command options are specific to where you mine. Refer to the pool's instructions on how to set them.
-
-

Makefile.am

@@ -36,21 +36,17 @@ cpuminer_SOURCES = \
   algo/argon2/argon2d/argon2d/argon2d_thread.c \
   algo/argon2/argon2d/argon2d/encoding.c \
   algo/blake/sph_blake.c \
-  algo/blake/blake256-hash-4way.c \
+  algo/blake/blake256-hash.c \
-  algo/blake/blake512-hash-4way.c \
+  algo/blake/blake512-hash.c \
   algo/blake/blake-gate.c \
   algo/blake/blake.c \
   algo/blake/blake-4way.c \
   algo/blake/sph_blake2b.c \
   algo/blake/sph-blake2s.c \
-  algo/blake/blake2s-hash-4way.c \
+  algo/blake/blake2s-hash.c \
   algo/blake/blake2s.c \
-  algo/blake/blake2s-gate.c \
+  algo/blake/blake2b-hash.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 \
@@ -163,8 +159,6 @@ cpuminer_SOURCES = \
   algo/sha/sph_sha2big.c \
   algo/sha/sha256-hash-4way.c \
   algo/sha/sha512-hash-4way.c \
-  algo/sha/sha256-hash-opt.c \
-  algo/sha/sha256-hash-2way-ni.c \
   algo/sha/hmac-sha256-hash.c \
   algo/sha/hmac-sha256-hash-4way.c \
   algo/sha/sha256d.c \
@@ -172,9 +166,10 @@ cpuminer_SOURCES = \
   algo/sha/sha256d-4way.c \
   algo/sha/sha256t-gate.c \
   algo/sha/sha256t-4way.c \
-  algo/sha/sha256t.c \
   algo/sha/sha256q-4way.c \
   algo/sha/sha256q.c \
+  algo/sha/sha512256d-4way.c \
+  algo/sha/sha256dt.c \
   algo/shabal/sph_shabal.c \
   algo/shabal/shabal-hash-4way.c \
   algo/shavite/sph_shavite.c \
@@ -292,10 +287,10 @@ disable_flags =
 if USE_ASM
    cpuminer_SOURCES += asm/neoscrypt_asm.S
 if ARCH_x86
-   cpuminer_SOURCES += asm/sha2-x86.S asm/scrypt-x86.S asm/aesb-x86.S
+   cpuminer_SOURCES += asm/sha2-x86.S asm/scrypt-x86.S
 endif
 if ARCH_x86_64
-   cpuminer_SOURCES += asm/sha2-x64.S asm/scrypt-x64.S asm/aesb-x64.S
+   cpuminer_SOURCES += asm/sha2-x64.S asm/scrypt-x64.S
 endif
 if ARCH_ARM
    cpuminer_SOURCES += asm/sha2-arm.S asm/scrypt-arm.S

RELEASE_NOTES

@@ -65,6 +65,53 @@ If not what makes it happen or not happen?
 Change Log
 ----------
 
+v3.23.3
+
+#400: Removed excessive thread restarts when mining solo.
+Fixed build_msys2.sh for gcc-13 by removing unsupported option "--param=evrp-mode=legacy" from CFLAGS.
+Added CPUID detection and reporting of CPUs and SW builds supporting SHA512 extension.
+Added prototype of sha-512 using SHA512 intrinsics, untested.
+Other improvements and code cleanup.
+
+v3.23.2
+
+sha256dt, sha256t & sha256d +10% with SHA, small improvement with AVX2.
+Other small improvements and code cleanup.
+
+v3.23.1
+
+#349: Fix sha256t low difficulty shares and low effective hash rate.
+Faster sha256dt:  AVX512 +7%, SHA +200%, AVX2 +5%.
+Faster blakecoin & vanilla: AVX2 +30%, AVX512 +110%.
+Other small improvements and code cleanup.
+
+v3.23.0
+
+#398: Prevent GBT fallback to Getwork on network error.
+#398: Prevent excessive logs when conditional mining is paused when mining solo.
+Fix a false start if stratum doesn't immediately send a new job after connecting.
+Tweak diagonal shuffle in Blake2b & Blake256 1-way SIMD to reduce latency.
+CPUID support for AVX10.
+Initial changes to AVX2 targeted code in preparation for AVX10.
+Code cleanup and miscellaneous small improvements.
+
+v3.22.3
+
+Data interleaving and byte swap optimizations with AVX2, AVX512 & AVX512VBMI.
+Faster Luffa with AVX2 & AVX512.
+Other small optimizations.
+Some code cleanup.
+
+v3.22.2
+
+Added sha512256d & sha256dt algos.
+Fixed intermittant invalid shares lyra2v2 AVX512.
+Removed application limits on the number of CPUs and threads, HW and OS limits still apply.
+Added a log warning if more threads are defined than active CPUs in affinity mask.
+Improved merkle tree memory management for stratum.
+Added transaction count to New Work log.
+Other small improvements.
+
 v3.22.1
 
 #393 fixed segfault in GBT, regression from v3.22.0.
@@ -187,40 +234,29 @@ v3.19.5
 
 Enhanced stratum-keepalive preemptively resets the stratum connection
 before the server to avoid lost shares.
-
 Added build-msys2.sh shell script for easier compiling on Windows, see Wiki for details.
-
 X16RT: eliminate unnecessary recalculations of the hash order.
-
 Fix a few compiler warnings.
-
 Fixed log colour error when a block is solved.
 
 v3.19.4
 
 #359: Fix verthash memory allocation for non-hugepages, broken in v3.19.3.
-
 New option stratum-keepalive prevents stratum timeouts when no shares are
 submitted for several minutes due to high difficulty.
-
 Fixed a bug displaying optimizations for some algos.
 
 v3.19.3
 
 Linux: Faster verthash (+25%), scryptn2 (+2%) when huge pages are available.
-
 Small speed up for Hamsi AVX2 & AVX512, Keccak AVX512.
 
 v3.19.2
 
 Fixed log displaying incorrect memory usage for scrypt, broken in v3.19.1.
-
 Reduce log noise when replies to submitted shares are lost due to stratum errors.
-
 Fugue prehash optimization for X16r family AVX2 & AVX512.
-
 Small speed improvement for Hamsi AVX2 & AVX512.
-
 Win: With CPU groups enabled the number of CPUs displayed in the ASCII art
 affinity map is the number of CPUs in a CPU group, was number of CPUs up to 64.
 
@@ -232,7 +268,6 @@ Changes to Windows binaries package:
  - zen build renamed to avx2-sha, supports Zen1 & Zen2,
  - avx512-sha build removed, Rocketlake CPUs can use avx512-sha-vaes,
  - see README.txt for compatibility details.
-
 Fixed a few compiler warnings that are new in GCC 11.
 Other minor fixes.
 
@@ -246,22 +281,17 @@ Changes to cpu-affinity:
   - streamlined code for more efficient initialization of miner threads,
   - precise affining of each miner thread to a specific CPU,
   - added an option to disable CPU affinity with "--cpu-affinity 0"
-
 Faster sha256t with AVX512 & AVX2.
-
 Added stratum error count to stats log, reported only when non-zero.
 
 v3.18.2
 
 Issue #342, fixed Groestl AES on Windows, broken in v3.18.0.
-
 AVX512 for sha256d.
-
 SSE42 and AVX may now be displayed as mining features at startup.
 This is hard coded for each algo, and is only implemented for scrypt
 at this time as it is the only algo with significant performance differences
 with those features.
-
 Fixed an issue where a high hashrate algo could cause excessive invalid hash
 rate log reports when starting up in benchmark mode.
 
@@ -272,9 +302,7 @@ More speed for scrypt:
  - AVX2 is now used by default on CPUS with SHA but not AVX512,
  - scrypt:1024 performance lost in v3.18.0 is restored,
  - AVX512 & AVX2 improvements to scrypt:1024.
-
 Big speedup for SwiFFTx AVX2 & SSE4.1: x22i +55%, x25x +22%.
-
 Issue #337: fixed a problem that could display negative stats values in the
 first summary report if the report was forced prematurely due to a stratum
 diff change. The stats will still be invalid but should display zeros.
@@ -287,26 +315,19 @@ Complete rewrite of Scrypt code, optimized for large N factor (scryptn2):
   - memory requirements reduced 30-60% depending on CPU architecture,
   - memory usage displayed at startup,
   - scrypt, default N=1024 (LTC), will likely perform slower.
-
 Improved stale share detection and handling for Scrypt with large N factor:
   - abort and discard partially computed hash when new work is detected,
   - quicker response to new job, less time wasted mining stale job.
-
 Improved stale share handling for all algorithms:
   - report possible stale share when new work received with a previously
     submitted share still pending,
   - when new work is detected report the submission of an already completed,
     otherwise valid, but likely stale, share,
   - fixed incorrect block height in stale share log.
-
 Small performance improvements to sha, bmw, cube & hamsi for AVX512 & AVX2.
-
 When stratum disconnects miner threads go to idle until reconnected.
-
 Colour changes to some logs.
-
 Some low level function name changes for clarity and consistency.
-
 The reference hashrate in the summary log and the benchmark total hashrate
 are now the mean hashrate for the session. 
 
@@ -419,7 +440,6 @@ Fixed neoscrypt BUG log.
 v3.14.3
 
 #265: more mutex changes to reduce blocking with high thread count.
-
 #267: fixed hodl algo potential memory alignment issue,
       add warning when thread count is not valid for mining hodl algo.
 

algo-gate-api.c

@@ -171,7 +171,7 @@ int scanhash_4way_64in_32out( struct work *work, uint32_t max_nonce,
          }
       }
       *noncev = _mm256_add_epi32( *noncev,
-                                  m256_const1_64( 0x0000000400000000 ) );
+                                  _mm256_set1_epi64x( 0x0000000400000000 ) );
       n += 4;
    } while ( likely( ( n <= last_nonce ) && !work_restart[thr_id].restart ) );
    pdata[19] = n;
@@ -227,7 +227,7 @@ int scanhash_8way_64in_32out( struct work *work, uint32_t max_nonce,
          }
       }
       *noncev = _mm512_add_epi32( *noncev,
-                                  m512_const1_64( 0x0000000800000000 ) );
+                                  _mm512_set1_epi64( 0x0000000800000000 ) );
       n += 8;
    } while ( likely( ( n < last_nonce ) && !work_restart[thr_id].restart ) );
    pdata[19] = n;
@@ -248,7 +248,7 @@ int null_hash()
    return 0;
 };
 
-void init_algo_gate( algo_gate_t* gate )
+static void init_algo_gate( algo_gate_t* gate )
 {
    gate->miner_thread_init       = (void*)&return_true;
    gate->scanhash                = (void*)&scanhash_generic;
@@ -337,9 +337,11 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
     case ALGO_QUBIT:        rc = register_qubit_algo         ( gate ); break;
     case ALGO_SCRYPT:       rc = register_scrypt_algo        ( gate ); break;
     case ALGO_SHA256D:      rc = register_sha256d_algo       ( gate ); break;
+    case ALGO_SHA256DT:     rc = register_sha256dt_algo      ( gate ); break;
     case ALGO_SHA256Q:      rc = register_sha256q_algo       ( gate ); break;
     case ALGO_SHA256T:      rc = register_sha256t_algo       ( gate ); break;
     case ALGO_SHA3D:        rc = register_sha3d_algo         ( gate ); break;
+    case ALGO_SHA512256D:   rc = register_sha512256d_algo    ( gate ); break;
     case ALGO_SHAVITE3:     rc = register_shavite_algo       ( gate ); break;
     case ALGO_SKEIN:        rc = register_skein_algo         ( gate ); break;
     case ALGO_SKEIN2:       rc = register_skein2_algo        ( gate ); break;

algo-gate-api.h

@@ -94,10 +94,14 @@ typedef  uint32_t set_t;
 #define SSE42_OPT        4
 #define AVX_OPT          8   // Sandybridge
 #define AVX2_OPT      0x10   // Haswell, Zen1
-#define SHA_OPT       0x20   // Zen1, Icelake (sha256)
+#define SHA_OPT       0x20   // Zen1, Icelake (deprecated)
-#define AVX512_OPT    0x40   // Skylake-X (AVX512[F,VL,DQ,BW])
+#define AVX512_OPT    0x40   // Skylake-X, Zen4 (AVX512[F,VL,DQ,BW])
-#define VAES_OPT      0x80   // Icelake (VAES & AVX512)
+#define VAES_OPT      0x80   // Icelake, Zen3
+#define SHA512_OPT   0x100   // Lunar Lake, Arrow Lake
 
+// AVX10 does not have explicit algo features:
+//  AVX10_512 is compatible with AVX512 + VAES
+//  AVX10_256 is compatible with AVX2 + VAES
 
 // return set containing all elements from sets a & b
 inline set_t set_union ( set_t a, set_t b ) { return a | b; }
@@ -264,7 +268,9 @@ void std_get_new_work( struct work *work, struct work *g_work, int thr_id,
                        uint32_t* end_nonce_ptr );
 
 void sha256d_gen_merkle_root( char *merkle_root, struct stratum_ctx *sctx );
-void SHA256_gen_merkle_root ( char *merkle_root, struct stratum_ctx *sctx );
+void sha256_gen_merkle_root ( char *merkle_root, struct stratum_ctx *sctx );
+// OpenSSL sha256 deprecated
+//void SHA256_gen_merkle_root ( char *merkle_root, struct stratum_ctx *sctx );
 
 bool std_le_work_decode( struct work *work );
 bool std_be_work_decode( struct work *work );

algo/argon2/argon2a/argon2a.c

@@ -77,7 +77,7 @@ 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->gen_merkle_root = (void*)&sha256_gen_merkle_root;
   opt_target_factor = 65536.0;
 
   return true;

algo/blake/blake-4way.c

@@ -1,5 +1,5 @@
 #include "blake-gate.h"
-#include "blake-hash-4way.h"
+#include "blake256-hash.h"
 #include <string.h>
 #include <stdint.h>
 #include <memory.h>

algo/blake/blake-hash-4way.h

@@ -1,231 +0,0 @@
-/* $Id: sph_blake.h 252 2011-06-07 17:55:14Z tp $ */
-/**
- * BLAKE interface. BLAKE is a family of functions which differ by their
- * output size; this implementation defines BLAKE for output sizes 224,
- * 256, 384 and 512 bits. This implementation conforms to the "third
- * round" specification.
- *
- * ==========================(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_blake.h
- * @author   Thomas Pornin <thomas.pornin@cryptolog.com>
- */
-
-#ifndef __BLAKE_HASH_4WAY__
-#define __BLAKE_HASH_4WAY__ 1
-
-#ifdef __cplusplus
-extern "C"{
-#endif
-
-#include <stddef.h>
-#include "algo/sha/sph_types.h"
-#include "simd-utils.h"
-
-#define SPH_SIZE_blake256   256
-
-#define SPH_SIZE_blake512   512
-
-/////////////////////////
-//
-//  Blake-256 1 way SSE2
-
-void  blake256_transform_le( uint32_t *H, const uint32_t *buf,
-                             const uint32_t T0, const uint32_t T1 );
-
-/////////////////////////
-//
-//  Blake-512 1 way SSE2
-
-void  blake512_transform_le( uint64_t *H, const uint64_t *buf,
-                             const uint64_t T0, const uint64_t T1 );
-
-//////////////////////////
-//
-//   Blake-256 4 way SSE2
-
-typedef struct {
-   unsigned char buf[64<<2];
-   uint32_t H[8<<2];
-   size_t ptr;
-   uint32_t T0, T1;
-   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
-} blake_4way_small_context __attribute__ ((aligned (64)));
-
-// Default, 14 rounds, blake, decred
-typedef blake_4way_small_context blake256_4way_context;
-void blake256_4way_init(void *ctx);
-void blake256_4way_update(void *ctx, const void *data, size_t len);
-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_update(void *cc, const void *data, size_t len);
-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_update(void *cc, const void *data, size_t len);
-void blake256r8_4way_close(void *cc, void *dst);
-
-#ifdef __AVX2__
-
-//////////////////////////
-//
-//   Blake-256 8 way AVX2
-
-typedef struct {
-   __m256i buf[16] __attribute__ ((aligned (64)));
-   __m256i H[8];
-   size_t ptr;
-   sph_u32 T0, T1;
-   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
-} blake_8way_small_context;
-
-// Default 14 rounds
-typedef blake_8way_small_context blake256_8way_context;
-void blake256_8way_init(void *cc);
-void blake256_8way_update(void *cc, const void *data, size_t len);
-void blake256_8way_close(void *cc, void *dst);
-void blake256_8way_update_le(void *cc, const void *data, size_t len);
-void blake256_8way_close_le(void *cc, void *dst);
-void blake256_8way_round0_prehash_le( void *midstate, const void *midhash,
-                                      void *data );
-void blake256_8way_final_rounds_le( void *final_hash, const void *midstate,
-                                     const void *midhash, const void *data );
-
-// 14 rounds, blake, decred
-typedef blake_8way_small_context blake256r14_8way_context;
-void blake256r14_8way_init(void *cc);
-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_update(void *cc, const void *data, size_t len);
-void blake256r8_8way_close(void *cc, void *dst);
-
-// Blake-512 4 way AVX2
-
-typedef struct {
-   __m256i buf[16];
-   __m256i H[8];
-   __m256i S[4];   
-   size_t ptr;
-   sph_u64 T0, T1;
-} blake_4way_big_context __attribute__ ((aligned (128)));
-
-typedef blake_4way_big_context blake512_4way_context;
-
-void blake512_4way_init( blake_4way_big_context *sc );
-void blake512_4way_update( void *cc, const void *data, size_t len );
-void blake512_4way_close( void *cc, void *dst );
-void blake512_4way_full( blake_4way_big_context *sc, void * dst,
-                         const void *data, size_t len );
-void blake512_4way_full_le( blake_4way_big_context *sc, void * dst,
-                            const void *data, size_t len );
-void blake512_4way_prehash_le( blake_4way_big_context *sc, __m256i *midstate,
-                               const void *data );
-void blake512_4way_final_le( blake_4way_big_context *sc, void *hash,
-                             const __m256i nonce, const __m256i *midstate );
-
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-
-////////////////////////////
-//
-//   Blake-256 16 way AVX512
-
-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);
-// Expects data in little endian order, no byte swap needed
-void blake256_16way_update_le(void *cc, const void *data, size_t len);
-void blake256_16way_close_le(void *cc, void *dst);
-void blake256_16way_round0_prehash_le( void *midstate, const void *midhash,
-                                       void *data );
-void blake256_16way_final_rounds_le( void *final_hash, const void *midstate,
-                                     const void *midhash, const void *data );
-
-
-// 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 AVX512
-
-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_full( blake_8way_big_context *sc, void * dst,
-                        const void *data, size_t len );
-void blake512_8way_full_le( blake_8way_big_context *sc, void * dst,
-                            const void *data, size_t len );
-void blake512_8way_prehash_le( blake_8way_big_context *sc, __m512i *midstate,
-                               const void *data );
-void blake512_8way_final_le( blake_8way_big_context *sc, void *hash,
-                             const __m512i nonce, const __m512i *midstate );
-
-#endif  // AVX512
-#endif  // AVX2
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif  // BLAKE_HASH_4WAY_H__

algo/blake/blake256-hash.h

@@ -0,0 +1,124 @@
+#ifndef BLAKE256_HASH__
+#define BLAKE256_HASH__ 1
+
+#include <stddef.h>
+#include "simd-utils.h"
+
+/////////////////////////
+//
+//  Blake-256 1 way SSE2
+
+void  blake256_transform_le( uint32_t *H, const uint32_t *buf,
+                             const uint32_t T0, const uint32_t T1, int rounds );
+
+//////////////////////////
+//
+//   Blake-256 4 way SSE2
+
+typedef struct {
+   unsigned char buf[64<<2];
+   uint32_t H[8<<2];
+   size_t ptr;
+   uint32_t T0, T1;
+   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
+} blake_4way_small_context __attribute__ ((aligned (64)));
+
+// Default, 14 rounds
+typedef blake_4way_small_context blake256_4way_context;
+void blake256_4way_init(void *ctx);
+void blake256_4way_update(void *ctx, const void *data, size_t len);
+void blake256_4way_close(void *ctx, void *dst);
+
+// 14 rounds
+typedef blake_4way_small_context blake256r14_4way_context;
+void blake256r14_4way_init(void *cc);
+void blake256r14_4way_update(void *cc, const void *data, size_t len);
+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_update(void *cc, const void *data, size_t len);
+void blake256r8_4way_close(void *cc, void *dst);
+
+#ifdef __AVX2__
+
+//////////////////////////
+//
+//   Blake-256 8 way AVX2
+
+typedef struct {
+   __m256i buf[16] __attribute__ ((aligned (64)));
+   __m256i H[8];
+   size_t ptr;
+   uint32_t T0, T1;
+   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
+} blake_8way_small_context;
+
+// Default 14 rounds
+typedef blake_8way_small_context blake256_8way_context;
+void blake256_8way_init(void *cc);
+void blake256_8way_update(void *cc, const void *data, size_t len);
+void blake256_8way_close(void *cc, void *dst);
+void blake256_8way_update_le(void *cc, const void *data, size_t len);
+void blake256_8way_close_le(void *cc, void *dst);
+void blake256_8way_round0_prehash_le( void *midstate, const void *midhash,
+                                      void *data );
+void blake256_8way_final_rounds_le( void *final_hash, const void *midstate,
+                    const void *midhash, const void *data, const int rounds );
+
+// 14 rounds, blake, decred
+typedef blake_8way_small_context blake256r14_8way_context;
+void blake256r14_8way_init(void *cc);
+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_update(void *cc, const void *data, size_t len);
+void blake256r8_8way_close(void *cc, void *dst);
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+////////////////////////////
+//
+//   Blake-256 16 way AVX512
+
+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);
+// Expects data in little endian order, no byte swap needed
+void blake256_16way_update_le(void *cc, const void *data, size_t len);
+void blake256_16way_close_le(void *cc, void *dst);
+void blake256_16way_round0_prehash_le( void *midstate, const void *midhash,
+                                       void *data );
+void blake256_16way_final_rounds_le( void *final_hash, const void *midstate,
+                     const void *midhash, const void *data, const int rounds );
+
+// 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);
+
+#endif  // AVX512
+#endif  // AVX2
+
+#endif  // BLAKE256_HASH_H__

algo/blake/blake2b-4way.c

@@ -1,113 +0,0 @@
-/**
- * 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_solution( work, lane_hash, mythr );
-          }
-      }
-      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_solution( work, lane_hash, mythr );
-          }
-      }
-      n += 4;
-   } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
-
-   *hashes_done = n - first_nonce + 1;
-   return 0;
-}
-
-#endif

algo/blake/blake2b-gate.c

@@ -1,20 +0,0 @@
-#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;
-};
-
-

algo/blake/blake2b-gate.h

@@ -1,34 +0,0 @@
-#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

algo/blake/blake2b-hash.c

@@ -31,7 +31,7 @@
 #include <stdint.h>
 #include <string.h>
 
-#include "blake2b-hash-4way.h"
+#include "blake2b-hash.h"
 
 #if defined(__AVX2__)
 
@@ -252,17 +252,17 @@ static void blake2b_8way_compress( blake2b_8way_ctx *ctx, int last )
    v[ 5] = ctx->h[5];
    v[ 6] = ctx->h[6];
    v[ 7] = ctx->h[7];
-   v[ 8] = m512_const1_64( 0x6A09E667F3BCC908 );
+   v[ 8] = v512_64( 0x6A09E667F3BCC908 );
-   v[ 9] = m512_const1_64( 0xBB67AE8584CAA73B );
+   v[ 9] = v512_64( 0xBB67AE8584CAA73B );
-   v[10] = m512_const1_64( 0x3C6EF372FE94F82B );
+   v[10] = v512_64( 0x3C6EF372FE94F82B );
-   v[11] = m512_const1_64( 0xA54FF53A5F1D36F1 );
+   v[11] = v512_64( 0xA54FF53A5F1D36F1 );
-   v[12] = m512_const1_64( 0x510E527FADE682D1 );
+   v[12] = v512_64( 0x510E527FADE682D1 );
-   v[13] = m512_const1_64( 0x9B05688C2B3E6C1F );
+   v[13] = v512_64( 0x9B05688C2B3E6C1F );
-   v[14] = m512_const1_64( 0x1F83D9ABFB41BD6B );
+   v[14] = v512_64( 0x1F83D9ABFB41BD6B );
-   v[15] = m512_const1_64( 0x5BE0CD19137E2179 );
+   v[15] = v512_64( 0x5BE0CD19137E2179 );
 
-   v[12] = _mm512_xor_si512( v[12], _mm512_set1_epi64( ctx->t[0] ) );
+   v[12] = _mm512_xor_si512( v[12], v512_64( ctx->t[0] ) );
-   v[13] = _mm512_xor_si512( v[13], _mm512_set1_epi64( ctx->t[1] ) );
+   v[13] = _mm512_xor_si512( v[13], v512_64( ctx->t[1] ) );
 
    if ( last )
       v[14] = mm512_not( v[14] );
@@ -310,16 +310,16 @@ int blake2b_8way_init( blake2b_8way_ctx *ctx )
 {
    size_t i;
 
-   ctx->h[0] = m512_const1_64( 0x6A09E667F3BCC908 );
+   ctx->h[0] = v512_64( 0x6A09E667F3BCC908 );
-   ctx->h[1] = m512_const1_64( 0xBB67AE8584CAA73B );
+   ctx->h[1] = v512_64( 0xBB67AE8584CAA73B );
-   ctx->h[2] = m512_const1_64( 0x3C6EF372FE94F82B );
+   ctx->h[2] = v512_64( 0x3C6EF372FE94F82B );
-   ctx->h[3] = m512_const1_64( 0xA54FF53A5F1D36F1 );
+   ctx->h[3] = v512_64( 0xA54FF53A5F1D36F1 );
-   ctx->h[4] = m512_const1_64( 0x510E527FADE682D1 );
+   ctx->h[4] = v512_64( 0x510E527FADE682D1 );
-   ctx->h[5] = m512_const1_64( 0x9B05688C2B3E6C1F );
+   ctx->h[5] = v512_64( 0x9B05688C2B3E6C1F );
-   ctx->h[6] = m512_const1_64( 0x1F83D9ABFB41BD6B );
+   ctx->h[6] = v512_64( 0x1F83D9ABFB41BD6B );
-   ctx->h[7] = m512_const1_64( 0x5BE0CD19137E2179 );
+   ctx->h[7] = v512_64( 0x5BE0CD19137E2179 );
 
-   ctx->h[0] = _mm512_xor_si512( ctx->h[0], m512_const1_64( 0x01010020 ) );
+   ctx->h[0] = _mm512_xor_si512( ctx->h[0], v512_64( 0x01010020 ) );
 
    ctx->t[0] = 0;
    ctx->t[1] = 0;
@@ -419,17 +419,17 @@ static void blake2b_4way_compress( blake2b_4way_ctx *ctx, int last )
    v[ 5] = ctx->h[5];
    v[ 6] = ctx->h[6];
    v[ 7] = ctx->h[7];
-   v[ 8] = m256_const1_64( 0x6A09E667F3BCC908 );
+   v[ 8] = v256_64( 0x6A09E667F3BCC908 );
-   v[ 9] = m256_const1_64( 0xBB67AE8584CAA73B );
+   v[ 9] = v256_64( 0xBB67AE8584CAA73B );
-   v[10] = m256_const1_64( 0x3C6EF372FE94F82B );
+   v[10] = v256_64( 0x3C6EF372FE94F82B );
-   v[11] = m256_const1_64( 0xA54FF53A5F1D36F1 );
+   v[11] = v256_64( 0xA54FF53A5F1D36F1 );
-   v[12] = m256_const1_64( 0x510E527FADE682D1 );
+   v[12] = v256_64( 0x510E527FADE682D1 );
-   v[13] = m256_const1_64( 0x9B05688C2B3E6C1F );
+   v[13] = v256_64( 0x9B05688C2B3E6C1F );
-   v[14] = m256_const1_64( 0x1F83D9ABFB41BD6B );
+   v[14] = v256_64( 0x1F83D9ABFB41BD6B );
-   v[15] = m256_const1_64( 0x5BE0CD19137E2179 );
+   v[15] = v256_64( 0x5BE0CD19137E2179 );
 
-   v[12] = _mm256_xor_si256( v[12], _mm256_set1_epi64x( ctx->t[0] ) );
+   v[12] = _mm256_xor_si256( v[12], v256_64( ctx->t[0] ) );
-   v[13] = _mm256_xor_si256( v[13], _mm256_set1_epi64x( ctx->t[1] ) );
+   v[13] = _mm256_xor_si256( v[13], v256_64( ctx->t[1] ) );
 
    if ( last )   
 		v[14] = mm256_not( v[14] );
@@ -477,16 +477,16 @@ int blake2b_4way_init( blake2b_4way_ctx *ctx )
 {
 	size_t i;
 
-   ctx->h[0] = m256_const1_64( 0x6A09E667F3BCC908 );
+   ctx->h[0] = v256_64( 0x6A09E667F3BCC908 );
-   ctx->h[1] = m256_const1_64( 0xBB67AE8584CAA73B );
+   ctx->h[1] = v256_64( 0xBB67AE8584CAA73B );
-   ctx->h[2] = m256_const1_64( 0x3C6EF372FE94F82B );
+   ctx->h[2] = v256_64( 0x3C6EF372FE94F82B );
-   ctx->h[3] = m256_const1_64( 0xA54FF53A5F1D36F1 );
+   ctx->h[3] = v256_64( 0xA54FF53A5F1D36F1 );
-   ctx->h[4] = m256_const1_64( 0x510E527FADE682D1 );
+   ctx->h[4] = v256_64( 0x510E527FADE682D1 );
-   ctx->h[5] = m256_const1_64( 0x9B05688C2B3E6C1F );
+   ctx->h[5] = v256_64( 0x9B05688C2B3E6C1F );
-   ctx->h[6] = m256_const1_64( 0x1F83D9ABFB41BD6B );
+   ctx->h[6] = v256_64( 0x1F83D9ABFB41BD6B );
-   ctx->h[7] = m256_const1_64( 0x5BE0CD19137E2179 );
+   ctx->h[7] = v256_64( 0x5BE0CD19137E2179 );
 
-   ctx->h[0] = _mm256_xor_si256( ctx->h[0], m256_const1_64( 0x01010020 ) );
+   ctx->h[0] = _mm256_xor_si256( ctx->h[0], v256_64( 0x01010020 ) );
 
 	ctx->t[0] = 0;
 	ctx->t[1] = 0;

algo/blake/blake2b.c

@@ -1,64 +1,175 @@
-/**
+#include "algo-gate-api.h"
- * Blake2-B Implementation
- * tpruvot@github 2015-2016
- */
-
-#include "blake2b-gate.h"
-
-#if !defined(BLAKE2B_8WAY) && !defined(BLAKE2B_4WAY)
-
 #include <string.h>
 #include <stdint.h>
-#include "algo/blake/sph_blake2b.h"
+#include "blake2b-hash.h"
 
-#define MIDLEN 76
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-#define A 64
+  #define BLAKE2B_8WAY
+#elif defined(__AVX2__)
+  #define BLAKE2B_4WAY
+#endif
 
-void blake2b_hash(void *output, const void *input)
+#if defined(BLAKE2B_8WAY)
+
+int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
+                           uint64_t *hashes_done, struct thr_info *mythr )
 {
-	uint8_t _ALIGN(A) hash[32];
+   uint32_t hash[8*8] __attribute__ ((aligned (128)));;
-	sph_blake2b_ctx ctx __attribute__ ((aligned (64)));
+   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];
 
-	sph_blake2b_init(&ctx, 32, NULL, 0);
+   uint32_t n = first_nonce;
-	sph_blake2b_update(&ctx, input, 80);
-	sph_blake2b_final(&ctx, hash);
 
-	memcpy(output, hash, 32);
+   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_solution( work, lane_hash, mythr );
+          }
+      }
+      n += 8;
+   } while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
 }
 
-int scanhash_blake2b( struct work *work, uint32_t max_nonce,
+#elif defined(BLAKE2B_4WAY)
-                      uint64_t *hashes_done, struct thr_info *mythr )
+
+// Function not used, code inlined.
+void blake2b_4way_hash(void *output, const void *input)
 {
-	uint32_t _ALIGN(A) vhashcpu[8];
+    blake2b_4way_ctx ctx;
-	uint32_t _ALIGN(A) endiandata[20];
+    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;  // thr_id arg is deprecated
+   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;
 
-	for (int i=0; i < 19; i++) {
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
-		be32enc(&endiandata[i], pdata[i]);
-	}
 
 	do {
-		be32enc(&endiandata[19], n);
+      *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
-		blake2b_hash(vhashcpu, endiandata);
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
-		if (vhashcpu[7] <= Htarg && fulltest(vhashcpu, ptarget))
+      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 )
       {
-			pdata[19] = n;
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
-         submit_solution( work, vhashcpu, mythr );
+          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          {
+              pdata[19] = n + lane;
+              submit_solution( work, lane_hash, mythr );
+          }
+      }
+      n += 4;
+   } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#else
+
+#include "algo/blake/sph_blake2b.h"
+
+void blake2b_hash(void *output, const void *input)
+{
+   uint8_t _ALIGN(32) hash[32];
+   sph_blake2b_ctx ctx __attribute__ ((aligned (32)));
+
+   sph_blake2b_init(&ctx, 32, NULL, 0);
+   sph_blake2b_update(&ctx, input, 80);
+   sph_blake2b_final(&ctx, hash);
+
+   memcpy(output, hash, 32);
+}
+
+int scanhash_blake2b( struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t _ALIGN(32) hash64[8];
+   uint32_t _ALIGN(32) endiandata[20];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   int thr_id = mythr->id; 
+   const uint32_t first_nonce = pdata[19];
+   uint32_t n = first_nonce;
+
+   mm128_bswap32_80( endiandata, pdata );
+
+   do {
+      endiandata[19] = n;
+      blake2b_hash( hash64, endiandata );
+      if ( unlikely( valid_hash( hash64, ptarget ) ) && !opt_benchmark )
+      {
+         pdata[19] = bswap_32( n );
+         submit_solution( work, hash64, mythr );
       }
       n++;
-	} while (n < max_nonce && !work_restart[thr_id].restart);
+   } while (n < max_nonce && !work_restart[thr_id].restart);
-	*hashes_done = n - first_nonce + 1;
+   *hashes_done = n - first_nonce + 1;
-	pdata[19] = n;
+   pdata[19] = n;
 
-	return 0;
+   return 0;
 }
 
 #endif
+
+bool register_blake2b_algo( algo_gate_t* gate )
+{
+#if defined(BLAKE2B_8WAY)
+  gate->scanhash  = (void*)&scanhash_blake2b_8way;
+#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;
+};
+

algo/blake/blake2s-4way.c

@@ -1,170 +0,0 @@
-#include "blake2s-gate.h"
-#include "blake2s-hash-4way.h"
-#include <string.h>
-#include <stdint.h>
-
-#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_solution( work, lane_hash, mythr );
-         }
-      }
-      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 )
-{
-   blake2s_8way_state ctx;
-   memcpy( &ctx, &blake2s_8w_ctx, sizeof ctx );
-   blake2s_8way_update( &ctx, input + (64<<3), 16 );
-   blake2s_8way_final( &ctx, output, BLAKE2S_OUTBYTES );
-}
-
-int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
-                      uint64_t *hashes_done, struct thr_info *mythr )
-{
-   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;
-   const uint32_t Htarg = ptarget[7];
-   const uint32_t first_nonce = pdata[19];
-   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
-   uint32_t n = first_nonce;
-   int thr_id = mythr->id; 
-
-   mm256_bswap32_intrlv80_8x32( vdata, pdata );
-   blake2s_8way_init( &blake2s_8w_ctx, BLAKE2S_OUTBYTES );
-   blake2s_8way_update( &blake2s_8w_ctx, vdata, 64 );
-
-   do {
-      *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 lane = 0; lane < 8; lane++ )
-      if ( unlikely( hash7[lane] <= Htarg ) )
-      {
-         extr_lane_8x32( lane_hash, hash, lane, 256 );
-         if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
-         {
-              pdata[19] = n + lane;
-              submit_solution( work, lane_hash, mythr );
-         }
-      }
-      n += 8;
-   } while ( (n < max_nonce) && !work_restart[thr_id].restart );
-
-   *hashes_done = n - first_nonce + 1;
-   return 0;
-}
-
-#elif defined(BLAKE2S_4WAY)
-
-static __thread blake2s_4way_state blake2s_4w_ctx;
-
-void blake2s_4way_hash( void *output, const void *input )
-{
-   blake2s_4way_state ctx;
-   memcpy( &ctx, &blake2s_4w_ctx, sizeof ctx );
-   blake2s_4way_update( &ctx, input + (64<<2), 16 );
-   blake2s_4way_final( &ctx, output, BLAKE2S_OUTBYTES );
-}
-
-int scanhash_blake2s_4way( struct work *work, uint32_t max_nonce,
-                      uint64_t *hashes_done, struct thr_info *mythr )
-{
-   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;
-   const uint32_t Htarg = ptarget[7];
-   const uint32_t first_nonce = pdata[19];
-   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
-   uint32_t n = first_nonce;
-   int thr_id = mythr->id; 
-
-   mm128_bswap32_intrlv80_4x32( vdata, pdata );
-   blake2s_4way_init( &blake2s_4w_ctx, BLAKE2S_OUTBYTES );
-   blake2s_4way_update( &blake2s_4w_ctx, vdata, 64 );
-
-   do {
-      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
-      pdata[19] = n;
-
-      blake2s_4way_hash( hash, vdata );
-
-      for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
-      {
-         extr_lane_4x32( lane_hash, hash, lane, 256 );
-         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
-         {
-              pdata[19] = n + lane;
-              submit_solution( work, lane_hash, mythr );
-              }
-      }
-      n += 4;
-   } while ( (n < max_nonce) && !work_restart[thr_id].restart );
-
-   *hashes_done = n - first_nonce + 1;
-   return 0;
-}
-
-#endif

algo/blake/blake2s-gate.c

@@ -1,23 +0,0 @@
-#include "blake2s-gate.h"
-
-bool register_blake2s_algo( algo_gate_t* gate )
-{
-#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)
-  gate->scanhash  = (void*)&scanhash_blake2s_4way;
-  gate->hash      = (void*)&blake2s_4way_hash;
-#else
-  gate->scanhash  = (void*)&scanhash_blake2s;
-  gate->hash      = (void*)&blake2s_hash;
-#endif
-  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
-  return true;
-};
-
-

algo/blake/blake2s-gate.h

@@ -1,46 +0,0 @@
-#ifndef __BLAKE2S_GATE_H__
-#define __BLAKE2S_GATE_H__ 1
-
-#include <stdint.h>
-#include "algo-gate-api.h"
-
-#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_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)
-
-void blake2s_8way_hash( void *state, const void *input );
-int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done, struct thr_info *mythr );
-
-#elif defined (BLAKE2S_4WAY)
-
-void blake2s_4way_hash( void *state, const void *input );
-int scanhash_blake2s_4way( struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done, struct thr_info *mythr );
-#else
-
-void blake2s_hash( void *state, const void *input );
-int scanhash_blake2s( struct work *work, uint32_t max_nonce,
-                      uint64_t *hashes_done, struct thr_info *mythr );
-
-#endif
-
-#endif

algo/blake/blake2s-hash.c

@@ -11,7 +11,7 @@
  * this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
  */
 
-#include "blake2s-hash-4way.h"
+#include "blake2s-hash.h"
 
 #include <stdint.h>
 #include <string.h>
@@ -62,23 +62,23 @@ int blake2s_4way_init( blake2s_4way_state *S, const uint8_t outlen )
 
    memset( S, 0, sizeof( blake2s_4way_state ) );
 
-   S->h[0] = m128_const1_64( 0x6A09E6676A09E667ULL );
+   S->h[0] = v128_64( 0x6A09E6676A09E667ULL );
-   S->h[1] = m128_const1_64( 0xBB67AE85BB67AE85ULL );
+   S->h[1] = v128_64( 0xBB67AE85BB67AE85ULL );
-   S->h[2] = m128_const1_64( 0x3C6EF3723C6EF372ULL );
+   S->h[2] = v128_64( 0x3C6EF3723C6EF372ULL );
-   S->h[3] = m128_const1_64( 0xA54FF53AA54FF53AULL );
+   S->h[3] = v128_64( 0xA54FF53AA54FF53AULL );
-   S->h[4] = m128_const1_64( 0x510E527F510E527FULL );
+   S->h[4] = v128_64( 0x510E527F510E527FULL );
-   S->h[5] = m128_const1_64( 0x9B05688C9B05688CULL );
+   S->h[5] = v128_64( 0x9B05688C9B05688CULL );
-   S->h[6] = m128_const1_64( 0x1F83D9AB1F83D9ABULL );
+   S->h[6] = v128_64( 0x1F83D9AB1F83D9ABULL );
-   S->h[7] = m128_const1_64( 0x5BE0CD195BE0CD19ULL );
+   S->h[7] = v128_64( 0x5BE0CD195BE0CD19ULL );
    
 //   for( int i = 0; i < 8; ++i )
-//      S->h[i] = _mm_set1_epi32( blake2s_IV[i] );
+//      S->h[i] = v128_32( blake2s_IV[i] );
 
    uint32_t *p = ( uint32_t * )( P );
 
    /* IV XOR ParamBlock */
    for ( size_t i = 0; i < 8; ++i )
-      S->h[i] = _mm_xor_si128( S->h[i], _mm_set1_epi32( p[i] ) );
+      S->h[i] = _mm_xor_si128( S->h[i], v128_32( p[i] ) );
    return 0;
 }
 
@@ -90,18 +90,18 @@ int blake2s_4way_compress( blake2s_4way_state *S, const __m128i* block )
    memcpy_128( m, block, 16 );
    memcpy_128( v, S->h, 8 );
 
-   v[ 8] = m128_const1_64( 0x6A09E6676A09E667ULL );
+   v[ 8] = v128_64( 0x6A09E6676A09E667ULL );
-   v[ 9] = m128_const1_64( 0xBB67AE85BB67AE85ULL );
+   v[ 9] = v128_64( 0xBB67AE85BB67AE85ULL );
-   v[10] = m128_const1_64( 0x3C6EF3723C6EF372ULL );
+   v[10] = v128_64( 0x3C6EF3723C6EF372ULL );
-   v[11] = m128_const1_64( 0xA54FF53AA54FF53AULL );
+   v[11] = v128_64( 0xA54FF53AA54FF53AULL );
-   v[12] = _mm_xor_si128( _mm_set1_epi32( S->t[0] ),
+   v[12] = _mm_xor_si128( v128_32( S->t[0] ),
-                          m128_const1_64( 0x510E527F510E527FULL ) );
+                          v128_64( 0x510E527F510E527FULL ) );
-   v[13] = _mm_xor_si128( _mm_set1_epi32( S->t[1] ),
+   v[13] = _mm_xor_si128( v128_32( S->t[1] ),
-                          m128_const1_64( 0x9B05688C9B05688CULL ) );
+                          v128_64( 0x9B05688C9B05688CULL ) );
-   v[14] = _mm_xor_si128( _mm_set1_epi32( S->f[0] ),
+   v[14] = _mm_xor_si128( v128_32( S->f[0] ),
-                          m128_const1_64( 0x1F83D9AB1F83D9ABULL ) );
+                          v128_64( 0x1F83D9AB1F83D9ABULL ) );
-   v[15] = _mm_xor_si128( _mm_set1_epi32( S->f[1] ),
+   v[15] = _mm_xor_si128( v128_32( S->f[1] ),
-                          m128_const1_64( 0x5BE0CD195BE0CD19ULL ) );
+                          v128_64( 0x5BE0CD195BE0CD19ULL ) );
 
 #define G4W( sigma0, sigma1, a, b, c, d ) \
 do { \
@@ -269,35 +269,35 @@ 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[ 8] = v256_64( 0x6A09E6676A09E667ULL );
-   v[ 9] = m256_const1_64( 0xBB67AE85BB67AE85ULL );
+   v[ 9] = v256_64( 0xBB67AE85BB67AE85ULL );
-   v[10] = m256_const1_64( 0x3C6EF3723C6EF372ULL );
+   v[10] = v256_64( 0x3C6EF3723C6EF372ULL );
-   v[11] = m256_const1_64( 0xA54FF53AA54FF53AULL );
+   v[11] = v256_64( 0xA54FF53AA54FF53AULL );
-   v[12] = _mm256_xor_si256( _mm256_set1_epi32( S->t[0] ),
+   v[12] = _mm256_xor_si256( v256_32( S->t[0] ),
-                          m256_const1_64( 0x510E527F510E527FULL ) );
+                          v256_64( 0x510E527F510E527FULL ) );
 
-   v[13] = _mm256_xor_si256( _mm256_set1_epi32( S->t[1] ),
+   v[13] = _mm256_xor_si256( v256_32( S->t[1] ),
-                          m256_const1_64( 0x9B05688C9B05688CULL ) );
+                          v256_64( 0x9B05688C9B05688CULL ) );
 
-   v[14] = _mm256_xor_si256( _mm256_set1_epi32( S->f[0] ),
+   v[14] = _mm256_xor_si256( v256_32( S->f[0] ),
-                          m256_const1_64( 0x1F83D9AB1F83D9ABULL ) );
+                          v256_64( 0x1F83D9AB1F83D9ABULL ) );
 
-   v[15] = _mm256_xor_si256( _mm256_set1_epi32( S->f[1] ),
+   v[15] = _mm256_xor_si256( v256_32( S->f[1] ),
-                          m256_const1_64( 0x5BE0CD195BE0CD19ULL ) );
+                          v256_64( 0x5BE0CD195BE0CD19ULL ) );
 
 /*
-   v[ 8] = _mm256_set1_epi32( blake2s_IV[0] );
+   v[ 8] = v256_32( blake2s_IV[0] );
-   v[ 9] = _mm256_set1_epi32( blake2s_IV[1] );
+   v[ 9] = v256_32( blake2s_IV[1] );
-   v[10] = _mm256_set1_epi32( blake2s_IV[2] );
+   v[10] = v256_32( blake2s_IV[2] );
-   v[11] = _mm256_set1_epi32( blake2s_IV[3] );
+   v[11] = v256_32( blake2s_IV[3] );
-   v[12] = _mm256_xor_si256( _mm256_set1_epi32( S->t[0] ),
+   v[12] = _mm256_xor_si256( v256_32( S->t[0] ),
-                             _mm256_set1_epi32( blake2s_IV[4] ) );
+                             v256_32( blake2s_IV[4] ) );
-   v[13] = _mm256_xor_si256( _mm256_set1_epi32( S->t[1] ),
+   v[13] = _mm256_xor_si256( v256_32( S->t[1] ),
-                             _mm256_set1_epi32( blake2s_IV[5] ) );
+                             v256_32( blake2s_IV[5] ) );
-   v[14] = _mm256_xor_si256( _mm256_set1_epi32( S->f[0] ),
+   v[14] = _mm256_xor_si256( v256_32( S->f[0] ),
-                             _mm256_set1_epi32( blake2s_IV[6] ) );
+                             v256_32( blake2s_IV[6] ) );
-   v[15] = _mm256_xor_si256( _mm256_set1_epi32( S->f[1] ),
+   v[15] = _mm256_xor_si256( v256_32( S->f[1] ),
-                             _mm256_set1_epi32( blake2s_IV[7] ) );
+                             v256_32( blake2s_IV[7] ) );
 
 
 #define G8W(r,i,a,b,c,d) \
@@ -391,24 +391,24 @@ 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 ) );
-   S->h[0] = m256_const1_64( 0x6A09E6676A09E667ULL );
+   S->h[0] = v256_64( 0x6A09E6676A09E667ULL );
-   S->h[1] = m256_const1_64( 0xBB67AE85BB67AE85ULL );
+   S->h[1] = v256_64( 0xBB67AE85BB67AE85ULL );
-   S->h[2] = m256_const1_64( 0x3C6EF3723C6EF372ULL );
+   S->h[2] = v256_64( 0x3C6EF3723C6EF372ULL );
-   S->h[3] = m256_const1_64( 0xA54FF53AA54FF53AULL );
+   S->h[3] = v256_64( 0xA54FF53AA54FF53AULL );
-   S->h[4] = m256_const1_64( 0x510E527F510E527FULL );
+   S->h[4] = v256_64( 0x510E527F510E527FULL );
-   S->h[5] = m256_const1_64( 0x9B05688C9B05688CULL );
+   S->h[5] = v256_64( 0x9B05688C9B05688CULL );
-   S->h[6] = m256_const1_64( 0x1F83D9AB1F83D9ABULL );
+   S->h[6] = v256_64( 0x1F83D9AB1F83D9ABULL );
-   S->h[7] = m256_const1_64( 0x5BE0CD195BE0CD19ULL );
+   S->h[7] = v256_64( 0x5BE0CD195BE0CD19ULL );
 
 
 //   for( int i = 0; i < 8; ++i )
-//      S->h[i] = _mm256_set1_epi32( blake2s_IV[i] );
+//      S->h[i] = v256_32( blake2s_IV[i] );
 
    uint32_t *p = ( uint32_t * )( P );
 
    /* IV XOR ParamBlock */
    for ( size_t i = 0; i < 8; ++i )
-      S->h[i] = _mm256_xor_si256( S->h[i], _mm256_set1_epi32( p[i] ) );
+      S->h[i] = _mm256_xor_si256( S->h[i], v256_32( p[i] ) );
    return 0;
 }
 
@@ -510,21 +510,21 @@ int blake2s_16way_compress( blake2s_16way_state *S, const __m512i *block )
    memcpy_512( m, block, 16 );
    memcpy_512( v, S->h, 8 );
 
-   v[ 8] = m512_const1_64( 0x6A09E6676A09E667ULL );
+   v[ 8] = v512_64( 0x6A09E6676A09E667ULL );
-   v[ 9] = m512_const1_64( 0xBB67AE85BB67AE85ULL );
+   v[ 9] = v512_64( 0xBB67AE85BB67AE85ULL );
-   v[10] = m512_const1_64( 0x3C6EF3723C6EF372ULL );
+   v[10] = v512_64( 0x3C6EF3723C6EF372ULL );
-   v[11] = m512_const1_64( 0xA54FF53AA54FF53AULL );
+   v[11] = v512_64( 0xA54FF53AA54FF53AULL );
-   v[12] = _mm512_xor_si512( _mm512_set1_epi32( S->t[0] ),
+   v[12] = _mm512_xor_si512( v512_32( S->t[0] ),
-                          m512_const1_64( 0x510E527F510E527FULL ) );
+                          v512_64( 0x510E527F510E527FULL ) );
 
-   v[13] = _mm512_xor_si512( _mm512_set1_epi32( S->t[1] ),
+   v[13] = _mm512_xor_si512( v512_32( S->t[1] ),
-                          m512_const1_64( 0x9B05688C9B05688CULL ) );
+                          v512_64( 0x9B05688C9B05688CULL ) );
 
-   v[14] = _mm512_xor_si512( _mm512_set1_epi32( S->f[0] ),
+   v[14] = _mm512_xor_si512( v512_32( S->f[0] ),
-                          m512_const1_64( 0x1F83D9AB1F83D9ABULL ) );
+                          v512_64( 0x1F83D9AB1F83D9ABULL ) );
 
-   v[15] = _mm512_xor_si512( _mm512_set1_epi32( S->f[1] ),
+   v[15] = _mm512_xor_si512( v512_32( S->f[1] ),
-                          m512_const1_64( 0x5BE0CD195BE0CD19ULL ) );
+                          v512_64( 0x5BE0CD195BE0CD19ULL ) );
 
 
 #define G16W( sigma0, sigma1, a, b, c, d) \
@@ -589,20 +589,20 @@ int blake2s_16way_init( blake2s_16way_state *S, const uint8_t outlen )
    memset( P->personal, 0, sizeof( P->personal ) );
 
    memset( S, 0, sizeof( blake2s_16way_state ) );
-   S->h[0] = m512_const1_64( 0x6A09E6676A09E667ULL );
+   S->h[0] = v512_64( 0x6A09E6676A09E667ULL );
-   S->h[1] = m512_const1_64( 0xBB67AE85BB67AE85ULL );
+   S->h[1] = v512_64( 0xBB67AE85BB67AE85ULL );
-   S->h[2] = m512_const1_64( 0x3C6EF3723C6EF372ULL );
+   S->h[2] = v512_64( 0x3C6EF3723C6EF372ULL );
-   S->h[3] = m512_const1_64( 0xA54FF53AA54FF53AULL );
+   S->h[3] = v512_64( 0xA54FF53AA54FF53AULL );
-   S->h[4] = m512_const1_64( 0x510E527F510E527FULL );
+   S->h[4] = v512_64( 0x510E527F510E527FULL );
-   S->h[5] = m512_const1_64( 0x9B05688C9B05688CULL );
+   S->h[5] = v512_64( 0x9B05688C9B05688CULL );
-   S->h[6] = m512_const1_64( 0x1F83D9AB1F83D9ABULL );
+   S->h[6] = v512_64( 0x1F83D9AB1F83D9ABULL );
-   S->h[7] = m512_const1_64( 0x5BE0CD195BE0CD19ULL );
+   S->h[7] = v512_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] ) );
+      S->h[i] = _mm512_xor_si512( S->h[i], v512_32( p[i] ) );
    return 0;
 }
 

algo/blake/blake2s.c

@@ -1,75 +1,252 @@
-#include "blake2s-gate.h"
+#include "algo-gate-api.h"
-
+#include "blake2s-hash.h"
-#if  !defined(BLAKE2S_16WAY) && !defined(BLAKE2S_8WAY) && !defined(BLAKE2S)
-
 #include <string.h>
 #include <stdint.h>
 
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define BLAKE2S_16WAY
+#elif defined(__AVX2__)
+  #define BLAKE2S_8WAY
+#elif defined(__SSE2__)
+  #define BLAKE2S_4WAY
+#endif
+
+#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_solution( work, lane_hash, mythr );
+         }
+      }
+      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 )
+{
+   blake2s_8way_state ctx;
+   memcpy( &ctx, &blake2s_8w_ctx, sizeof ctx );
+   blake2s_8way_update( &ctx, input + (64<<3), 16 );
+   blake2s_8way_final( &ctx, output, BLAKE2S_OUTBYTES );
+}
+
+int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
+{
+   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;
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
+   uint32_t n = first_nonce;
+   int thr_id = mythr->id; 
+
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
+   blake2s_8way_init( &blake2s_8w_ctx, BLAKE2S_OUTBYTES );
+   blake2s_8way_update( &blake2s_8w_ctx, vdata, 64 );
+
+   do {
+      *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 lane = 0; lane < 8; lane++ )
+      if ( unlikely( hash7[lane] <= Htarg ) )
+      {
+         extr_lane_8x32( lane_hash, hash, lane, 256 );
+         if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
+         {
+              pdata[19] = n + lane;
+              submit_solution( work, lane_hash, mythr );
+         }
+      }
+      n += 8;
+   } while ( (n < max_nonce) && !work_restart[thr_id].restart );
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#elif defined(BLAKE2S_4WAY)
+
+static __thread blake2s_4way_state blake2s_4w_ctx;
+
+void blake2s_4way_hash( void *output, const void *input )
+{
+   blake2s_4way_state ctx;
+   memcpy( &ctx, &blake2s_4w_ctx, sizeof ctx );
+   blake2s_4way_update( &ctx, input + (64<<2), 16 );
+   blake2s_4way_final( &ctx, output, BLAKE2S_OUTBYTES );
+}
+
+int scanhash_blake2s_4way( struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
+{
+   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;
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
+   uint32_t n = first_nonce;
+   int thr_id = mythr->id; 
+
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
+   blake2s_4way_init( &blake2s_4w_ctx, BLAKE2S_OUTBYTES );
+   blake2s_4way_update( &blake2s_4w_ctx, vdata, 64 );
+
+   do {
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
+      pdata[19] = n;
+
+      blake2s_4way_hash( hash, vdata );
+
+      for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
+      {
+         extr_lane_4x32( lane_hash, hash, lane, 256 );
+         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+         {
+              pdata[19] = n + lane;
+              submit_solution( work, lane_hash, mythr );
+              }
+      }
+      n += 4;
+   } while ( (n < max_nonce) && !work_restart[thr_id].restart );
+
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#else
+
 #include "sph-blake2s.h"
 
 static __thread blake2s_state blake2s_ctx;
-//static __thread blake2s_state s_ctx;
-#define MIDLEN 76
 
 void blake2s_hash( void *output, const void *input )
 {
-   unsigned char _ALIGN(64) hash[BLAKE2S_OUTBYTES];
+   unsigned char _ALIGN(32) hash[BLAKE2S_OUTBYTES];
-   blake2s_state ctx __attribute__ ((aligned (64)));
+   blake2s_state ctx __attribute__ ((aligned (32)));
-  
+
    memcpy( &ctx, &blake2s_ctx, sizeof ctx );
    blake2s_update( &ctx, input+64, 16 );
- 
+   blake2s_final( &ctx, hash, BLAKE2S_OUTBYTES );
-//	blake2s_init(&ctx, BLAKE2S_OUTBYTES);
-//	blake2s_update(&ctx, input, 80);
-	blake2s_final( &ctx, hash, BLAKE2S_OUTBYTES );
 
-	memcpy(output, hash, 32);
+   memcpy(output, hash, 32);
 }
-/*
+
-static void blake2s_hash_end(uint32_t *output, const uint32_t *input)
+int scanhash_blake2s( struct work *work,uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
 {
-	s_ctx.buflen = MIDLEN;
+   uint32_t *pdata = work->data;
-	memcpy(&s_ctx, &s_midstate, 32 + 16 + MIDLEN);
+   const uint32_t *ptarget = work->target;
-	blake2s_update(&s_ctx, (uint8_t*) &input[MIDLEN/4], 80 - MIDLEN);
+   uint32_t _ALIGN(32) hash32[8];
-	blake2s_final(&s_ctx, (uint8_t*) output, BLAKE2S_OUTBYTES);
+   uint32_t _ALIGN(32) endiandata[20];
+   const int thr_id = mythr->id;
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+   uint32_t n = first_nonce;
+
+   mm128_bswap32_80( endiandata, pdata );
+
+   // midstate
+   blake2s_init( &blake2s_ctx, BLAKE2S_OUTBYTES );
+   blake2s_update( &blake2s_ctx, (uint8_t*) endiandata, 64 );
+
+   do
+   {
+      endiandata[19] = n;
+      blake2s_hash( hash32, endiandata );
+      if ( unlikely( valid_hash( hash32, ptarget ) ) && !opt_benchmark )
+      {
+         pdata[19] = bswap_32( n );
+         submit_solution( work, hash32, mythr );
+      }
+      n++;
+   } while (n < max_nonce && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce + 1;
+   pdata[19] = n;
+
+   return 0;
 }
-*/
-int scanhash_blake2s( 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) hash64[8];
-	uint32_t _ALIGN(64) endiandata[20];
-   int thr_id = mythr->id;  // thr_id arg is deprecated
-
-	const uint32_t Htarg = ptarget[7];
-	const uint32_t first_nonce = pdata[19];
-
-	uint32_t n = first_nonce;
-
-        swab32_array( endiandata, pdata, 20 );
-
-	// midstate
-	blake2s_init( &blake2s_ctx, BLAKE2S_OUTBYTES );
-	blake2s_update( &blake2s_ctx, (uint8_t*) endiandata, 64 );
-
-	do {
-		be32enc(&endiandata[19], n);
-		blake2s_hash( hash64, endiandata );
-		if (hash64[7] <= Htarg && fulltest(hash64, ptarget)) {
-			*hashes_done = n - first_nonce + 1;
-			pdata[19] = n;
-			return true;
-		}
-		n++;
-
-	} while (n < max_nonce && !work_restart[thr_id].restart);
-
-	*hashes_done = n - first_nonce + 1;
-	pdata[19] = n;
-
-	return 0;
-}
 #endif
+
+bool register_blake2s_algo( algo_gate_t* gate )
+{
+#if defined(BLAKE2S_16WAY)
+  gate->scanhash  = (void*)&scanhash_blake2s_16way;
+  gate->hash      = (void*)&blake2s_16way_hash;
+#elif defined(BLAKE2S_8WAY)
+  gate->scanhash  = (void*)&scanhash_blake2s_8way;
+  gate->hash      = (void*)&blake2s_8way_hash;
+#elif defined(BLAKE2S_4WAY)
+  gate->scanhash  = (void*)&scanhash_blake2s_4way;
+  gate->hash      = (void*)&blake2s_4way_hash;
+#else
+  gate->scanhash  = (void*)&scanhash_blake2s;
+  gate->hash      = (void*)&blake2s_hash;
+#endif
+  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
+  return true;
+};
+

algo/blake/blake512-hash.h

@@ -0,0 +1,83 @@
+#ifndef BLAKE512_HASH__
+#define BLAKE512_HASH__ 1
+
+#include <stddef.h>
+#include "simd-utils.h"
+
+/////////////////////////
+//
+//  Blake-512 1 way SSE2 & AVX2
+
+typedef struct {
+   unsigned char buf[128];    /* first field, for alignment */
+   uint64_t H[8];
+   uint64_t T0, T1;
+   size_t ptr;
+} blake512_context __attribute__ ((aligned (32)));
+
+void  blake512_transform( uint64_t *H, const uint64_t *buf,
+                          const uint64_t T0, const uint64_t T1 );
+void blake512_init( blake512_context *sc );
+void blake512_update( blake512_context *sc, const void *data, size_t len );
+void blake512_close( blake512_context *sc, void *dst );
+void blake512_full( blake512_context *sc, void *dst, const void *data,
+                    size_t len );
+
+#ifdef __AVX2__
+
+// Blake-512 4 way AVX2
+
+typedef struct {
+   __m256i buf[16];
+   __m256i H[8];
+   __m256i S[4];   
+   size_t ptr;
+   uint64_t T0, T1;
+} blake_4way_big_context __attribute__ ((aligned (64)));
+
+typedef blake_4way_big_context blake512_4way_context;
+
+void blake512_4way_init( blake_4way_big_context *sc );
+void blake512_4way_update( void *cc, const void *data, size_t len );
+void blake512_4way_close( void *cc, void *dst );
+void blake512_4way_full( blake_4way_big_context *sc, void * dst,
+                         const void *data, size_t len );
+void blake512_4way_full_le( blake_4way_big_context *sc, void * dst,
+                            const void *data, size_t len );
+void blake512_4way_prehash_le( blake_4way_big_context *sc, __m256i *midstate,
+                               const void *data );
+void blake512_4way_final_le( blake_4way_big_context *sc, void *hash,
+                             const __m256i nonce, const __m256i *midstate );
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+////////////////////////////
+//
+//// Blake-512 8 way AVX512
+
+typedef struct {
+   __m512i buf[16];
+   __m512i H[8];
+   __m512i S[4];
+   size_t ptr;
+   uint64_t 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_full( blake_8way_big_context *sc, void * dst,
+                        const void *data, size_t len );
+void blake512_8way_full_le( blake_8way_big_context *sc, void * dst,
+                            const void *data, size_t len );
+void blake512_8way_prehash_le( blake_8way_big_context *sc, __m512i *midstate,
+                               const void *data );
+void blake512_8way_final_le( blake_8way_big_context *sc, void *hash,
+                             const __m512i nonce, const __m512i *midstate );
+
+#endif  // AVX512
+#endif  // AVX2
+
+#endif  // BLAKE512_HASH_H__

algo/blake/blakecoin-4way.c

@@ -1,10 +1,152 @@
 #include "blakecoin-gate.h"
-#include "blake-hash-4way.h"
+#include "blake256-hash.h"
 #include <string.h>
 #include <stdint.h>
 #include <memory.h>
 
-#if defined (BLAKECOIN_4WAY)
+#define rounds 8
+
+#if defined (BLAKECOIN_16WAY)
+
+int scanhash_blakecoin_16way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t hash32[8*16] __attribute__ ((aligned (64)));
+   uint32_t midstate_vars[16*16] __attribute__ ((aligned (64)));
+   __m512i block0_hash[8] __attribute__ ((aligned (64)));
+   __m512i block_buf[16] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash32_d7 =  (uint32_t*)&( ((__m512i*)hash32)[7] );
+   uint32_t *pdata = work->data;
+   const uint32_t *ptarget = work->target;
+   const uint32_t targ32_d7 = ptarget[7];
+   uint32_t phash[8] __attribute__ ((aligned (64))) =
+   {
+      0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
+      0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19
+   };
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = (const uint32_t) n;
+   const uint32_t last_nonce = max_nonce - 16;
+   const int thr_id = mythr->id;
+   const bool bench = opt_benchmark;
+   const __m512i sixteen = v512_32( 16 );
+
+   // Prehash first block
+   blake256_transform_le( phash, pdata, 512, 0, rounds );
+
+   block0_hash[0] = v512_32( phash[0] );
+   block0_hash[1] = v512_32( phash[1] );
+   block0_hash[2] = v512_32( phash[2] );
+   block0_hash[3] = v512_32( phash[3] );
+   block0_hash[4] = v512_32( phash[4] );
+   block0_hash[5] = v512_32( phash[5] );
+   block0_hash[6] = v512_32( phash[6] );
+   block0_hash[7] = v512_32( phash[7] );
+
+   // Build vectored second block, interleave last 16 bytes of data using
+   // unique nonces.
+   block_buf[0] = v512_32( pdata[16] );
+   block_buf[1] = v512_32( pdata[17] );
+   block_buf[2] = v512_32( pdata[18] );
+   block_buf[3] =
+             _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 );
+
+   // Partialy prehash second block without touching nonces in block_buf[3].
+   blake256_16way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+
+   do {
+      blake256_16way_final_rounds_le( hash32, midstate_vars, block0_hash,
+                                      block_buf, rounds );
+      for ( int lane = 0; lane < 16; lane++ )
+      if ( unlikely( hash32_d7[ lane ] <= targ32_d7 ) )
+      {
+         extr_lane_16x32( lane_hash, hash32, lane, 256 );
+         if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
+         {
+            pdata[19] = n + lane;
+            submit_solution( work, lane_hash, mythr );
+         }
+      }
+      block_buf[3] = _mm512_add_epi32( block_buf[3], sixteen );
+      n += 16;
+   } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart) );
+   pdata[19] = n;
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+
+#elif defined (BLAKECOIN_8WAY)
+
+int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t hash32[8*8] __attribute__ ((aligned (64)));
+   uint32_t midstate_vars[16*8] __attribute__ ((aligned (32)));
+   __m256i block0_hash[8] __attribute__ ((aligned (32)));
+   __m256i block_buf[16] __attribute__ ((aligned (32)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash32_d7 =  (uint32_t*)&( ((__m256i*)hash32)[7] );
+   uint32_t *pdata = work->data;
+   const uint32_t *ptarget = work->target;
+   const uint32_t targ32_d7 = ptarget[7];
+   uint32_t phash[8] __attribute__ ((aligned (32))) =
+   {
+      0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
+      0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19
+   };
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = (const uint32_t) n;
+   const uint32_t last_nonce = max_nonce - 8;
+   const int thr_id = mythr->id;
+   const bool bench = opt_benchmark;
+   const __m256i eight = v256_32( 8 );
+
+   // Prehash first block
+   blake256_transform_le( phash, pdata, 512, 0, rounds );
+
+   block0_hash[0] = v256_32( phash[0] );
+   block0_hash[1] = v256_32( phash[1] );
+   block0_hash[2] = v256_32( phash[2] );
+   block0_hash[3] = v256_32( phash[3] );
+   block0_hash[4] = v256_32( phash[4] );
+   block0_hash[5] = v256_32( phash[5] );
+   block0_hash[6] = v256_32( phash[6] );
+   block0_hash[7] = v256_32( phash[7] );
+
+   // Build vectored second block, interleave last 16 bytes of data using
+   // unique nonces.
+   block_buf[0] = v256_32( pdata[16] );
+   block_buf[1] = v256_32( pdata[17] );
+   block_buf[2] = v256_32( pdata[18] );
+   block_buf[3] = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
+
+   // Partialy prehash second block without touching nonces in block_buf[3].
+   blake256_8way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+
+   do {
+      blake256_8way_final_rounds_le( hash32, midstate_vars, block0_hash,
+                                     block_buf, rounds );
+      for ( int lane = 0; lane < 8; lane++ )
+      if ( unlikely( hash32_d7[ lane ] <= targ32_d7 ) )
+      {
+         extr_lane_8x32( lane_hash, hash32, lane, 256 );
+         if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
+         {
+            pdata[19] = n + lane;
+            submit_solution( work, lane_hash, mythr );
+         }
+      }
+      block_buf[3] = _mm256_add_epi32( block_buf[3], eight );
+      n += 8;
+   } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart) );
+   pdata[19] = n;
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+   
+#elif defined (BLAKECOIN_4WAY)
 
 blake256r8_4way_context blakecoin_4w_ctx;
 
@@ -61,61 +203,3 @@ int scanhash_blakecoin_4way( struct work *work, uint32_t max_nonce,
 
 #endif
 
-#if defined(BLAKECOIN_8WAY)
-
-blake256r8_8way_context blakecoin_8w_ctx;
-
-void blakecoin_8way_hash( void *state, const void *input )
-{
-     uint32_t vhash[8*8] __attribute__ ((aligned (64)));
-     blake256r8_8way_context ctx;
-
-     memcpy( &ctx, &blakecoin_8w_ctx, sizeof ctx );
-     blake256r8_8way_update( &ctx, input + (64<<3), 16 );
-     blake256r8_8way_close( &ctx, vhash );
-
-     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,
-                         uint64_t *hashes_done, struct thr_info *mythr )
-{
-   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
-   uint32_t hash[8*8] __attribute__ ((aligned (32)));
-   uint32_t *pdata = work->data;
-   uint32_t *ptarget = work->target;
-   const uint32_t first_nonce = pdata[19];
-   uint32_t HTarget = ptarget[7];
-   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;
-
-   mm256_bswap32_intrlv80_8x32( vdata, pdata );
-   blake256r8_8way_init( &blakecoin_8w_ctx );
-   blake256r8_8way_update( &blakecoin_8w_ctx, vdata, 64 );
-
-   do {
-      *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 );
-
-      for ( int i = 0; i < 8; i++ )
-      if (  (hash+(i<<3))[7] <= HTarget && fulltest( hash+(i<<3), ptarget )
-          && !opt_benchmark )
-      {
-          pdata[19] = n+i;
-          submit_solution( work, hash+(i<<3), mythr );
-      }
-      n += 8;
-   } while ( (n < max_nonce) && !work_restart[thr_id].restart );
-
-   *hashes_done = n - first_nonce + 1;
-   return 0;
-}
-
-#endif
-

algo/blake/blakecoin-gate.c

@@ -4,10 +4,10 @@
 // vanilla uses default gen merkle root, otherwise identical to blakecoin
 bool register_vanilla_algo( algo_gate_t* gate )
 {
-#if defined(BLAKECOIN_8WAY)
+#if defined(BLAKECOIN_16WAY)
+  gate->scanhash  = (void*)&scanhash_blakecoin_16way;
+#elif defined(BLAKECOIN_8WAY)
   gate->scanhash  = (void*)&scanhash_blakecoin_8way;
-  gate->hash      = (void*)&blakecoin_8way_hash;
-
 #elif defined(BLAKECOIN_4WAY)
   gate->scanhash  = (void*)&scanhash_blakecoin_4way;
   gate->hash      = (void*)&blakecoin_4way_hash;
@@ -15,14 +15,14 @@ bool register_vanilla_algo( algo_gate_t* gate )
   gate->scanhash = (void*)&scanhash_blakecoin;
   gate->hash     = (void*)&blakecoinhash;
 #endif
-  gate->optimizations = SSE42_OPT | AVX2_OPT;
+  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
   return true;
 }
 
 bool register_blakecoin_algo( algo_gate_t* gate )
 {
   register_vanilla_algo( gate );
-  gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
+  gate->gen_merkle_root = (void*)&sha256_gen_merkle_root;
   return true;
 }
 

algo/blake/blakecoin-gate.h

@@ -1,30 +1,36 @@
-#ifndef __BLAKECOIN_GATE_H__
+#ifndef BLAKECOIN_GATE_H__
-#define __BLAKECOIN_GATE_H__ 1
+#define BLAKECOIN_GATE_H__ 1
 
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__SSE4_2__)
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+  #define BLAKECOIN_16WAY
+#elif defined(__AVX2__)
+  #define BLAKECOIN_8WAY
+#elif defined(__SSE2__)  // always true
   #define BLAKECOIN_4WAY
 #endif
-#if defined(__AVX2__)
-  #define BLAKECOIN_8WAY
-#endif
 
-#if defined (BLAKECOIN_8WAY)
+#if defined (BLAKECOIN_16WAY)
-void blakecoin_8way_hash(void *state, const void *input);
+int scanhash_blakecoin_16way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+#elif defined (BLAKECOIN_8WAY)
+//void blakecoin_8way_hash(void *state, const void *input);
 int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
                          uint64_t *hashes_done, struct thr_info *mythr );
-#endif
 
-#if defined (BLAKECOIN_4WAY)
+#elif defined (BLAKECOIN_4WAY)
 void blakecoin_4way_hash(void *state, const void *input);
 int scanhash_blakecoin_4way( struct work *work, uint32_t max_nonce,
                          uint64_t *hashes_done, struct thr_info *mythr );
-#endif
+#else  // never used
 
 void blakecoinhash( void *state, const void *input );
 int scanhash_blakecoin( struct work *work, uint32_t max_nonce,
                       uint64_t *hashes_done, struct thr_info *mythr );
 
 #endif
+
+#endif

algo/blake/blakecoin.c

@@ -1,6 +1,6 @@
 #include "blakecoin-gate.h"
 
-#if !defined(BLAKECOIN_8WAY) && !defined(BLAKECOIN_4WAY)
+#if !defined(BLAKECOIN_16WAY) && !defined(BLAKECOIN_8WAY) && !defined(BLAKECOIN_4WAY)
 
 #define BLAKE32_ROUNDS 8
 #include "sph_blake.h"
@@ -12,7 +12,6 @@ void blakecoin_close(void *cc, void *dst);
 #include <string.h>
 #include <stdint.h>
 #include <memory.h>
-#include <openssl/sha.h>
 
 // context management is staged for efficiency.
 // 1. global initial ctx cached on startup
@@ -35,8 +34,8 @@ void blakecoinhash( void *state, const void *input )
 	uint8_t hash[64] __attribute__ ((aligned (32)));
 	uint8_t *ending = (uint8_t*) input + 64;
 
-        // copy cached midstate
+   // copy cached midstate
-        memcpy( &ctx, &blake_mid_ctx, sizeof ctx );
+   memcpy( &ctx, &blake_mid_ctx, sizeof ctx );
 	blakecoin( &ctx, ending, 16 );
 	blakecoin_close( &ctx, hash );
 	memcpy( state, hash, 32 );
@@ -45,8 +44,8 @@ void blakecoinhash( void *state, const void *input )
 int scanhash_blakecoin( struct work *work, uint32_t max_nonce,
                         uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t *pdata = work->data;
+   uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
+   uint32_t *ptarget = work->target;
 	const uint32_t first_nonce = pdata[19];
 	uint32_t HTarget = ptarget[7];
    int thr_id = mythr->id;  // thr_id arg is deprecated
@@ -60,10 +59,10 @@ int scanhash_blakecoin( struct work *work, uint32_t max_nonce,
 		HTarget = 0x7f;
 
 	// we need big endian data...
-        for (int kk=0; kk < 19; kk++) 
+   for (int kk=0; kk < 19; kk++) 
-                be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]);
+      be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]);
 
-        blake_midstate_init( endiandata );
+   blake_midstate_init( endiandata );
 
 #ifdef DEBUG_ALGO
 	applog(LOG_DEBUG,"[%d] Target=%08x %08x", thr_id, ptarget[6], ptarget[7]);

algo/blake/pentablake-4way.c

@@ -6,9 +6,7 @@
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
-
+#include "blake512-hash.h"
-#include "blake-hash-4way.h"
-#include "sph_blake.h"
 
 extern void pentablakehash_4way( void *output, const void *input )
 {

algo/blake/sph-blake2s.c

@@ -14,8 +14,9 @@
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
-
+#include "simd-utils.h"
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
+#include "compat.h"
 #include "sph-blake2s.h"
 
 static const uint32_t blake2s_IV[8] =
@@ -208,8 +209,8 @@ int blake2s_init_key( blake2s_state *S, const uint8_t outlen, const void *key, c
 
 int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES] )
 {
-	uint32_t m[16];
+	uint32_t _ALIGN(32) m[16];
-	uint32_t v[16];
+	uint32_t _ALIGN(32) v[16];
 
 	for( size_t i = 0; i < 16; ++i )
 		m[i] = load32( block + i * sizeof( m[i] ) );
@@ -225,6 +226,58 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES]
 	v[13] = S->t[1] ^ blake2s_IV[5];
 	v[14] = S->f[0] ^ blake2s_IV[6];
 	v[15] = S->f[1] ^ blake2s_IV[7];
+
+#if defined(__SSE2__)
+
+   __m128i *V = (__m128i*)v;
+
+#define BLAKE2S_ROUND( r ) \
+   V[0] = _mm_add_epi32( V[0], _mm_add_epi32( V[1], _mm_set_epi32( \
+                  m[blake2s_sigma[r][ 6]], m[blake2s_sigma[r][ 4]], \
+                  m[blake2s_sigma[r][ 2]], m[blake2s_sigma[r][ 0]] ) ) ); \
+   V[3] = mm128_swap32_16( _mm_xor_si128( V[3], V[0] ) ); \
+   V[2] = _mm_add_epi32( V[2], V[3] ); \
+   V[1] = mm128_ror_32( _mm_xor_si128( V[1], V[2] ), 12 ); \
+   V[0] = _mm_add_epi32( V[0], _mm_add_epi32( V[1], _mm_set_epi32( \
+                   m[blake2s_sigma[r][ 7]], m[blake2s_sigma[r][ 5]], \
+                   m[blake2s_sigma[r][ 3]], m[blake2s_sigma[r][ 1]] ) ) ); \
+   V[3] = mm128_shuflr32_8( _mm_xor_si128( V[3], V[0] ) ); \
+   V[2] = _mm_add_epi32( V[2], V[3] ); \
+   V[1] = mm128_ror_32( _mm_xor_si128( V[1], V[2] ), 7 ); \
+   V[0] = mm128_shufll_32( V[0] ); \
+   V[3] = mm128_swap_64( V[3] ); \
+   V[2] = mm128_shuflr_32( V[2] ); \
+   V[0] = _mm_add_epi32( V[0], _mm_add_epi32( V[1], _mm_set_epi32( \
+                    m[blake2s_sigma[r][12]], m[blake2s_sigma[r][10]], \
+                    m[blake2s_sigma[r][ 8]], m[blake2s_sigma[r][14]] ) ) ); \
+   V[3] = mm128_swap32_16( _mm_xor_si128( V[3], V[0] ) ); \
+   V[2] = _mm_add_epi32( V[2], V[3] ); \
+   V[1] = mm128_ror_32( _mm_xor_si128( V[1], V[2] ), 12 ); \
+   V[0] = _mm_add_epi32( V[0], _mm_add_epi32( V[1], _mm_set_epi32( \
+                    m[blake2s_sigma[r][13]], m[blake2s_sigma[r][11]], \
+                    m[blake2s_sigma[r][ 9]], m[blake2s_sigma[r][15]] ) ) ); \
+   V[3] = mm128_shuflr32_8( _mm_xor_si128( V[3], V[0] ) ); \
+   V[2] = _mm_add_epi32( V[2], V[3] ); \
+   V[1] = mm128_ror_32( _mm_xor_si128( V[1], V[2] ), 7 ); \
+   V[0] = mm128_shuflr_32( V[0] ); \
+   V[3] = mm128_swap_64( V[3] ); \
+   V[2] = mm128_shufll_32( V[2] )
+
+   BLAKE2S_ROUND(0);
+   BLAKE2S_ROUND(1);
+   BLAKE2S_ROUND(2);
+   BLAKE2S_ROUND(3);
+   BLAKE2S_ROUND(4);
+   BLAKE2S_ROUND(5);
+   BLAKE2S_ROUND(6);
+   BLAKE2S_ROUND(7);
+   BLAKE2S_ROUND(8);
+   BLAKE2S_ROUND(9);
+   
+#undef BLAKE2S_ROUND
+
+#else
+
 #define G(r,i,a,b,c,d) \
 	do { \
 		a = a + b + m[blake2s_sigma[r][2*i+0]]; \
@@ -236,6 +289,7 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES]
 		c = c + d; \
 		b = SPH_ROTR32(b ^ c, 7); \
 	} while(0)
+
 #define ROUND(r)  \
 	do { \
 		G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
@@ -247,7 +301,8 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES]
 		G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
 		G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
 	} while(0)
-	ROUND( 0 );
+
+   ROUND( 0 );
 	ROUND( 1 );
 	ROUND( 2 );
 	ROUND( 3 );
@@ -258,6 +313,8 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES]
 	ROUND( 8 );
 	ROUND( 9 );
 
+#endif
+
 	for( size_t i = 0; i < 8; ++i )
 		S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
 

algo/blake/sph_blake.h

@@ -42,7 +42,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for BLAKE-224.

algo/blake/sph_blake2b.c

@@ -31,7 +31,7 @@
 #include <stdint.h>
 #include <string.h>
 #include "simd-utils.h"
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 #include "sph_blake2b.h"
 
 // Little-endian byte access.
@@ -64,6 +64,22 @@
   V[1] = mm256_ror_64( _mm256_xor_si256( V[1], V[2] ), 63 ); \
 }
 
+// Pivot about V[1] instead of V[0] reduces latency.
+#define BLAKE2B_ROUND( R ) \
+{ \
+  __m256i *V = (__m256i*)v; \
+  const uint8_t *sigmaR = sigma[R]; \
+  BLAKE2B_G(  0,  1,  2,  3,  4,  5,  6,  7 ); \
+  V[0] = mm256_shufll_64( V[0] ); \
+  V[3] = mm256_swap_128( V[3] ); \
+  V[2] = mm256_shuflr_64( V[2] ); \
+  BLAKE2B_G( 14, 15,  8,  9, 10, 11, 12, 13 ); \
+  V[0] = mm256_shuflr_64( V[0] ); \
+  V[3] = mm256_swap_128( V[3] ); \
+  V[2] = mm256_shufll_64( V[2] ); \
+}
+
+/*
 #define BLAKE2B_ROUND( R ) \
 { \
   __m256i *V = (__m256i*)v; \
@@ -77,6 +93,7 @@
   V[2] = mm256_swap_128( V[2] ); \
   V[1] = mm256_shufll_64( V[1] ); \
 }
+*/
 
 #elif defined(__SSE2__)
 // always true

algo/bmw/bmw-hash-4way.h

@@ -41,8 +41,6 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-
-#include "algo/sha/sph_types.h"
 #include "simd-utils.h"
 
 #define SPH_SIZE_bmw256   256
@@ -57,7 +55,7 @@ typedef struct {
    __m128i buf[64];
    __m128i H[16];
    size_t ptr;
-   sph_u32 bit_count;  // assume bit_count fits in 32 bits
+   uint32_t bit_count;  // assume bit_count fits in 32 bits
 } bmw_4way_small_context;
 
 typedef bmw_4way_small_context bmw256_4way_context;
@@ -144,7 +142,7 @@ typedef struct {
    __m256i buf[16];
    __m256i H[16];
    size_t ptr;
-   sph_u64 bit_count;
+   uint64_t bit_count;
 } bmw_4way_big_context __attribute__((aligned(128)));
 
 typedef bmw_4way_big_context bmw512_4way_context;

algo/bmw/bmw256-hash-4way.c

@@ -109,7 +109,7 @@ static const uint32_t IV256[] = {
              _mm_sub_epi32( _mm_add_epi32( rol_off_32( M, j, 0 ), \
                                            rol_off_32( M, j, 3 ) ), \
                             rol_off_32( M, j, 10 ) ), \
-       _mm_set1_epi32( ( (j)+16 ) * SPH_C32(0x05555555UL) ) ), \
+       _mm_set1_epi32( ( (j)+16 ) * 0x05555555UL ) ), \
    H[ ( (j)+7 ) & 0xF ] )
 
 
@@ -451,22 +451,22 @@ static const __m128i final_s[16] =
 */
 void bmw256_4way_init( bmw256_4way_context *ctx )
 {
-   ctx->H[ 0] = m128_const1_64( 0x4041424340414243 );
+   ctx->H[ 0] = _mm_set1_epi64x( 0x4041424340414243 );
-   ctx->H[ 1] = m128_const1_64( 0x4445464744454647 );
+   ctx->H[ 1] = _mm_set1_epi64x( 0x4445464744454647 );
-   ctx->H[ 2] = m128_const1_64( 0x48494A4B48494A4B );
+   ctx->H[ 2] = _mm_set1_epi64x( 0x48494A4B48494A4B );
-   ctx->H[ 3] = m128_const1_64( 0x4C4D4E4F4C4D4E4F );
+   ctx->H[ 3] = _mm_set1_epi64x( 0x4C4D4E4F4C4D4E4F );
-   ctx->H[ 4] = m128_const1_64( 0x5051525350515253 );
+   ctx->H[ 4] = _mm_set1_epi64x( 0x5051525350515253 );
-   ctx->H[ 5] = m128_const1_64( 0x5455565754555657 );
+   ctx->H[ 5] = _mm_set1_epi64x( 0x5455565754555657 );
-   ctx->H[ 6] = m128_const1_64( 0x58595A5B58595A5B );
+   ctx->H[ 6] = _mm_set1_epi64x( 0x58595A5B58595A5B );
-   ctx->H[ 7] = m128_const1_64( 0x5C5D5E5F5C5D5E5F );
+   ctx->H[ 7] = _mm_set1_epi64x( 0x5C5D5E5F5C5D5E5F );
-   ctx->H[ 8] = m128_const1_64( 0x6061626360616263 );
+   ctx->H[ 8] = _mm_set1_epi64x( 0x6061626360616263 );
-   ctx->H[ 9] = m128_const1_64( 0x6465666764656667 );
+   ctx->H[ 9] = _mm_set1_epi64x( 0x6465666764656667 );
-   ctx->H[10] = m128_const1_64( 0x68696A6B68696A6B );
+   ctx->H[10] = _mm_set1_epi64x( 0x68696A6B68696A6B );
-   ctx->H[11] = m128_const1_64( 0x6C6D6E6F6C6D6E6F );
+   ctx->H[11] = _mm_set1_epi64x( 0x6C6D6E6F6C6D6E6F );
-   ctx->H[12] = m128_const1_64( 0x7071727370717273 );
+   ctx->H[12] = _mm_set1_epi64x( 0x7071727370717273 );
-   ctx->H[13] = m128_const1_64( 0x7475767774757677 );
+   ctx->H[13] = _mm_set1_epi64x( 0x7475767774757677 );
-   ctx->H[14] = m128_const1_64( 0x78797A7B78797A7B );
+   ctx->H[14] = _mm_set1_epi64x( 0x78797A7B78797A7B );
-   ctx->H[15] = m128_const1_64( 0x7C7D7E7F7C7D7E7F );
+   ctx->H[15] = _mm_set1_epi64x( 0x7C7D7E7F7C7D7E7F );
 
 
 //   for ( int i = 0; i < 16; i++ )
@@ -485,7 +485,7 @@ bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
    size_t ptr;
    const int buf_size = 64;  // bytes of one lane, compatible with len
 
-   sc->bit_count += (sph_u32)len << 3;
+   sc->bit_count += (uint32_t)len << 3;
    buf = sc->buf;
    ptr = sc->ptr;
    h1 = sc->H;
@@ -529,7 +529,7 @@ bmw32_4way_close(bmw_4way_small_context *sc, unsigned ub, unsigned n,
 
    buf = sc->buf;
    ptr = sc->ptr;
-   buf[ ptr>>2 ] = m128_const1_64( 0x0000008000000080 );
+   buf[ ptr>>2 ] = _mm_set1_epi64x( 0x0000008000000080 );
    ptr += 4;
    h = sc->H;
 
@@ -959,22 +959,22 @@ static const __m256i final_s8[16] =
 
 void bmw256_8way_init( bmw256_8way_context *ctx )
 {
-   ctx->H[ 0] = m256_const1_64( 0x4041424340414243 );
+   ctx->H[ 0] = _mm256_set1_epi64x( 0x4041424340414243 );
-   ctx->H[ 1] = m256_const1_64( 0x4445464744454647 );
+   ctx->H[ 1] = _mm256_set1_epi64x( 0x4445464744454647 );
-   ctx->H[ 2] = m256_const1_64( 0x48494A4B48494A4B );
+   ctx->H[ 2] = _mm256_set1_epi64x( 0x48494A4B48494A4B );
-   ctx->H[ 3] = m256_const1_64( 0x4C4D4E4F4C4D4E4F );
+   ctx->H[ 3] = _mm256_set1_epi64x( 0x4C4D4E4F4C4D4E4F );
-   ctx->H[ 4] = m256_const1_64( 0x5051525350515253 );
+   ctx->H[ 4] = _mm256_set1_epi64x( 0x5051525350515253 );
-   ctx->H[ 5] = m256_const1_64( 0x5455565754555657 );
+   ctx->H[ 5] = _mm256_set1_epi64x( 0x5455565754555657 );
-   ctx->H[ 6] = m256_const1_64( 0x58595A5B58595A5B );
+   ctx->H[ 6] = _mm256_set1_epi64x( 0x58595A5B58595A5B );
-   ctx->H[ 7] = m256_const1_64( 0x5C5D5E5F5C5D5E5F );
+   ctx->H[ 7] = _mm256_set1_epi64x( 0x5C5D5E5F5C5D5E5F );
-   ctx->H[ 8] = m256_const1_64( 0x6061626360616263 );
+   ctx->H[ 8] = _mm256_set1_epi64x( 0x6061626360616263 );
-   ctx->H[ 9] = m256_const1_64( 0x6465666764656667 );
+   ctx->H[ 9] = _mm256_set1_epi64x( 0x6465666764656667 );
-   ctx->H[10] = m256_const1_64( 0x68696A6B68696A6B );
+   ctx->H[10] = _mm256_set1_epi64x( 0x68696A6B68696A6B );
-   ctx->H[11] = m256_const1_64( 0x6C6D6E6F6C6D6E6F );
+   ctx->H[11] = _mm256_set1_epi64x( 0x6C6D6E6F6C6D6E6F );
-   ctx->H[12] = m256_const1_64( 0x7071727370717273 );
+   ctx->H[12] = _mm256_set1_epi64x( 0x7071727370717273 );
-   ctx->H[13] = m256_const1_64( 0x7475767774757677 );
+   ctx->H[13] = _mm256_set1_epi64x( 0x7475767774757677 );
-   ctx->H[14] = m256_const1_64( 0x78797A7B78797A7B );
+   ctx->H[14] = _mm256_set1_epi64x( 0x78797A7B78797A7B );
-   ctx->H[15] = m256_const1_64( 0x7C7D7E7F7C7D7E7F );
+   ctx->H[15] = _mm256_set1_epi64x( 0x7C7D7E7F7C7D7E7F );
    ctx->ptr       = 0;
    ctx->bit_count = 0;
 }
@@ -1030,7 +1030,7 @@ void bmw256_8way_close( bmw256_8way_context *ctx, void *dst )
 
    buf = ctx->buf;
    ptr = ctx->ptr;
-   buf[ ptr>>2 ] = m256_const1_64( 0x0000008000000080 );
+   buf[ ptr>>2 ] = _mm256_set1_epi64x( 0x0000008000000080 );
    ptr += 4;
    h = ctx->H;
 
@@ -1460,22 +1460,22 @@ static const __m512i final_s16[16] =
 
 void bmw256_16way_init( bmw256_16way_context *ctx )
 {
-   ctx->H[ 0] = m512_const1_64( 0x4041424340414243 );
+   ctx->H[ 0] = _mm512_set1_epi64( 0x4041424340414243 );
-   ctx->H[ 1] = m512_const1_64( 0x4445464744454647 );
+   ctx->H[ 1] = _mm512_set1_epi64( 0x4445464744454647 );
-   ctx->H[ 2] = m512_const1_64( 0x48494A4B48494A4B );
+   ctx->H[ 2] = _mm512_set1_epi64( 0x48494A4B48494A4B );
-   ctx->H[ 3] = m512_const1_64( 0x4C4D4E4F4C4D4E4F );
+   ctx->H[ 3] = _mm512_set1_epi64( 0x4C4D4E4F4C4D4E4F );
-   ctx->H[ 4] = m512_const1_64( 0x5051525350515253 );
+   ctx->H[ 4] = _mm512_set1_epi64( 0x5051525350515253 );
-   ctx->H[ 5] = m512_const1_64( 0x5455565754555657 );
+   ctx->H[ 5] = _mm512_set1_epi64( 0x5455565754555657 );
-   ctx->H[ 6] = m512_const1_64( 0x58595A5B58595A5B );
+   ctx->H[ 6] = _mm512_set1_epi64( 0x58595A5B58595A5B );
-   ctx->H[ 7] = m512_const1_64( 0x5C5D5E5F5C5D5E5F );
+   ctx->H[ 7] = _mm512_set1_epi64( 0x5C5D5E5F5C5D5E5F );
-   ctx->H[ 8] = m512_const1_64( 0x6061626360616263 );
+   ctx->H[ 8] = _mm512_set1_epi64( 0x6061626360616263 );
-   ctx->H[ 9] = m512_const1_64( 0x6465666764656667 );
+   ctx->H[ 9] = _mm512_set1_epi64( 0x6465666764656667 );
-   ctx->H[10] = m512_const1_64( 0x68696A6B68696A6B );
+   ctx->H[10] = _mm512_set1_epi64( 0x68696A6B68696A6B );
-   ctx->H[11] = m512_const1_64( 0x6C6D6E6F6C6D6E6F );
+   ctx->H[11] = _mm512_set1_epi64( 0x6C6D6E6F6C6D6E6F );
-   ctx->H[12] = m512_const1_64( 0x7071727370717273 );
+   ctx->H[12] = _mm512_set1_epi64( 0x7071727370717273 );
-   ctx->H[13] = m512_const1_64( 0x7475767774757677 );
+   ctx->H[13] = _mm512_set1_epi64( 0x7475767774757677 );
-   ctx->H[14] = m512_const1_64( 0x78797A7B78797A7B );
+   ctx->H[14] = _mm512_set1_epi64( 0x78797A7B78797A7B );
-   ctx->H[15] = m512_const1_64( 0x7C7D7E7F7C7D7E7F );
+   ctx->H[15] = _mm512_set1_epi64( 0x7C7D7E7F7C7D7E7F );
    ctx->ptr       = 0;
    ctx->bit_count = 0;
 }
@@ -1531,7 +1531,7 @@ void bmw256_16way_close( bmw256_16way_context *ctx, void *dst )
 
    buf = ctx->buf;
    ptr = ctx->ptr;
-   buf[ ptr>>2 ] = m512_const1_64( 0x0000008000000080 );
+   buf[ ptr>>2 ] = _mm512_set1_epi64( 0x0000008000000080 );
    ptr += 4;
    h = ctx->H;
 

algo/bmw/bmw512-hash-4way.c

@@ -45,15 +45,15 @@ extern "C"{
 
 #define LPAR   (
 
-static const sph_u64 IV512[] = {
+static const uint64_t IV512[] = {
-        SPH_C64(0x8081828384858687), SPH_C64(0x88898A8B8C8D8E8F),
+        0x8081828384858687, 0x88898A8B8C8D8E8F,
-        SPH_C64(0x9091929394959697), SPH_C64(0x98999A9B9C9D9E9F),
+        0x9091929394959697, 0x98999A9B9C9D9E9F,
-        SPH_C64(0xA0A1A2A3A4A5A6A7), SPH_C64(0xA8A9AAABACADAEAF),
+        0xA0A1A2A3A4A5A6A7, 0xA8A9AAABACADAEAF,
-        SPH_C64(0xB0B1B2B3B4B5B6B7), SPH_C64(0xB8B9BABBBCBDBEBF),
+        0xB0B1B2B3B4B5B6B7, 0xB8B9BABBBCBDBEBF,
-        SPH_C64(0xC0C1C2C3C4C5C6C7), SPH_C64(0xC8C9CACBCCCDCECF),
+        0xC0C1C2C3C4C5C6C7, 0xC8C9CACBCCCDCECF,
-        SPH_C64(0xD0D1D2D3D4D5D6D7), SPH_C64(0xD8D9DADBDCDDDEDF),
+        0xD0D1D2D3D4D5D6D7, 0xD8D9DADBDCDDDEDF,
-        SPH_C64(0xE0E1E2E3E4E5E6E7), SPH_C64(0xE8E9EAEBECEDEEEF),
+        0xE0E1E2E3E4E5E6E7, 0xE8E9EAEBECEDEEEF,
-        SPH_C64(0xF0F1F2F3F4F5F6F7), SPH_C64(0xF8F9FAFBFCFDFEFF)
+        0xF0F1F2F3F4F5F6F7, 0xF8F9FAFBFCFDFEFF
 };
 
 #if defined(__SSE2__)
@@ -894,24 +894,24 @@ static const __m256i final_b[16] =
 };
 
 static void
-bmw64_4way_init( bmw_4way_big_context *sc, const sph_u64 *iv )
+bmw64_4way_init( bmw_4way_big_context *sc, const uint64_t *iv )
 {
-   sc->H[ 0] = m256_const1_64( 0x8081828384858687 );
+   sc->H[ 0] = _mm256_set1_epi64x( 0x8081828384858687 );
-   sc->H[ 1] = m256_const1_64( 0x88898A8B8C8D8E8F );
+   sc->H[ 1] = _mm256_set1_epi64x( 0x88898A8B8C8D8E8F );
-   sc->H[ 2] = m256_const1_64( 0x9091929394959697 );
+   sc->H[ 2] = _mm256_set1_epi64x( 0x9091929394959697 );
-   sc->H[ 3] = m256_const1_64( 0x98999A9B9C9D9E9F );
+   sc->H[ 3] = _mm256_set1_epi64x( 0x98999A9B9C9D9E9F );
-   sc->H[ 4] = m256_const1_64( 0xA0A1A2A3A4A5A6A7 );
+   sc->H[ 4] = _mm256_set1_epi64x( 0xA0A1A2A3A4A5A6A7 );
-   sc->H[ 5] = m256_const1_64( 0xA8A9AAABACADAEAF );
+   sc->H[ 5] = _mm256_set1_epi64x( 0xA8A9AAABACADAEAF );
-   sc->H[ 6] = m256_const1_64( 0xB0B1B2B3B4B5B6B7 );
+   sc->H[ 6] = _mm256_set1_epi64x( 0xB0B1B2B3B4B5B6B7 );
-   sc->H[ 7] = m256_const1_64( 0xB8B9BABBBCBDBEBF );
+   sc->H[ 7] = _mm256_set1_epi64x( 0xB8B9BABBBCBDBEBF );
-   sc->H[ 8] = m256_const1_64( 0xC0C1C2C3C4C5C6C7 );
+   sc->H[ 8] = _mm256_set1_epi64x( 0xC0C1C2C3C4C5C6C7 );
-   sc->H[ 9] = m256_const1_64( 0xC8C9CACBCCCDCECF );
+   sc->H[ 9] = _mm256_set1_epi64x( 0xC8C9CACBCCCDCECF );
-   sc->H[10] = m256_const1_64( 0xD0D1D2D3D4D5D6D7 );
+   sc->H[10] = _mm256_set1_epi64x( 0xD0D1D2D3D4D5D6D7 );
-   sc->H[11] = m256_const1_64( 0xD8D9DADBDCDDDEDF );
+   sc->H[11] = _mm256_set1_epi64x( 0xD8D9DADBDCDDDEDF );
-   sc->H[12] = m256_const1_64( 0xE0E1E2E3E4E5E6E7 );
+   sc->H[12] = _mm256_set1_epi64x( 0xE0E1E2E3E4E5E6E7 );
-   sc->H[13] = m256_const1_64( 0xE8E9EAEBECEDEEEF );
+   sc->H[13] = _mm256_set1_epi64x( 0xE8E9EAEBECEDEEEF );
-   sc->H[14] = m256_const1_64( 0xF0F1F2F3F4F5F6F7 );
+   sc->H[14] = _mm256_set1_epi64x( 0xF0F1F2F3F4F5F6F7 );
-   sc->H[15] = m256_const1_64( 0xF8F9FAFBFCFDFEFF );
+   sc->H[15] = _mm256_set1_epi64x( 0xF8F9FAFBFCFDFEFF );
    sc->ptr = 0;
    sc->bit_count = 0;
 }
@@ -926,7 +926,7 @@ bmw64_4way( bmw_4way_big_context *sc, const void *data, size_t len )
    size_t ptr;
    const int buf_size = 128;  // bytes of one lane, compatible with len
 
-   sc->bit_count += (sph_u64)len << 3;
+   sc->bit_count += (uint64_t)len << 3;
    buf = sc->buf;
    ptr = sc->ptr;
    h1 = sc->H;
@@ -967,7 +967,7 @@ bmw64_4way_close(bmw_4way_big_context *sc, unsigned ub, unsigned n,
 
    buf = sc->buf;
    ptr = sc->ptr;
-   buf[ ptr>>3 ] = m256_const1_64( 0x80 );
+   buf[ ptr>>3 ] = _mm256_set1_epi64x( 0x80 );
    ptr += 8;
    h = sc->H;
 
@@ -1377,24 +1377,24 @@ static const __m512i final_b8[16] =
 
 
 void bmw512_8way_init( bmw512_8way_context *ctx )
-//bmw64_4way_init( bmw_4way_big_context *sc, const sph_u64 *iv )
+//bmw64_4way_init( bmw_4way_big_context *sc, const uint64_t *iv )
 {
-   ctx->H[ 0] = m512_const1_64( 0x8081828384858687 );
+   ctx->H[ 0] = _mm512_set1_epi64( 0x8081828384858687 );
-   ctx->H[ 1] = m512_const1_64( 0x88898A8B8C8D8E8F );
+   ctx->H[ 1] = _mm512_set1_epi64( 0x88898A8B8C8D8E8F );
-   ctx->H[ 2] = m512_const1_64( 0x9091929394959697 );
+   ctx->H[ 2] = _mm512_set1_epi64( 0x9091929394959697 );
-   ctx->H[ 3] = m512_const1_64( 0x98999A9B9C9D9E9F );
+   ctx->H[ 3] = _mm512_set1_epi64( 0x98999A9B9C9D9E9F );
-   ctx->H[ 4] = m512_const1_64( 0xA0A1A2A3A4A5A6A7 );
+   ctx->H[ 4] = _mm512_set1_epi64( 0xA0A1A2A3A4A5A6A7 );
-   ctx->H[ 5] = m512_const1_64( 0xA8A9AAABACADAEAF );
+   ctx->H[ 5] = _mm512_set1_epi64( 0xA8A9AAABACADAEAF );
-   ctx->H[ 6] = m512_const1_64( 0xB0B1B2B3B4B5B6B7 );
+   ctx->H[ 6] = _mm512_set1_epi64( 0xB0B1B2B3B4B5B6B7 );
-   ctx->H[ 7] = m512_const1_64( 0xB8B9BABBBCBDBEBF );
+   ctx->H[ 7] = _mm512_set1_epi64( 0xB8B9BABBBCBDBEBF );
-   ctx->H[ 8] = m512_const1_64( 0xC0C1C2C3C4C5C6C7 );
+   ctx->H[ 8] = _mm512_set1_epi64( 0xC0C1C2C3C4C5C6C7 );
-   ctx->H[ 9] = m512_const1_64( 0xC8C9CACBCCCDCECF );
+   ctx->H[ 9] = _mm512_set1_epi64( 0xC8C9CACBCCCDCECF );
-   ctx->H[10] = m512_const1_64( 0xD0D1D2D3D4D5D6D7 );
+   ctx->H[10] = _mm512_set1_epi64( 0xD0D1D2D3D4D5D6D7 );
-   ctx->H[11] = m512_const1_64( 0xD8D9DADBDCDDDEDF );
+   ctx->H[11] = _mm512_set1_epi64( 0xD8D9DADBDCDDDEDF );
-   ctx->H[12] = m512_const1_64( 0xE0E1E2E3E4E5E6E7 );
+   ctx->H[12] = _mm512_set1_epi64( 0xE0E1E2E3E4E5E6E7 );
-   ctx->H[13] = m512_const1_64( 0xE8E9EAEBECEDEEEF );
+   ctx->H[13] = _mm512_set1_epi64( 0xE8E9EAEBECEDEEEF );
-   ctx->H[14] = m512_const1_64( 0xF0F1F2F3F4F5F6F7 );
+   ctx->H[14] = _mm512_set1_epi64( 0xF0F1F2F3F4F5F6F7 );
-   ctx->H[15] = m512_const1_64( 0xF8F9FAFBFCFDFEFF );
+   ctx->H[15] = _mm512_set1_epi64( 0xF8F9FAFBFCFDFEFF );
    ctx->ptr = 0;
    ctx->bit_count = 0;
 }
@@ -1448,7 +1448,7 @@ void bmw512_8way_close( bmw512_8way_context *ctx, void *dst )
 
    buf = ctx->buf;
    ptr = ctx->ptr;
-   buf[ ptr>>3 ] = m512_const1_64( 0x80 );
+   buf[ ptr>>3 ] = _mm512_set1_epi64( 0x80 );
    ptr += 8;
    h = ctx->H;
 
@@ -1483,22 +1483,22 @@ void bmw512_8way_full( bmw512_8way_context *ctx, void *out, const void *data,
 
 // Init
 
-   H[ 0] = m512_const1_64( 0x8081828384858687 );
+   H[ 0] = _mm512_set1_epi64( 0x8081828384858687 );
-   H[ 1] = m512_const1_64( 0x88898A8B8C8D8E8F );
+   H[ 1] = _mm512_set1_epi64( 0x88898A8B8C8D8E8F );
-   H[ 2] = m512_const1_64( 0x9091929394959697 );
+   H[ 2] = _mm512_set1_epi64( 0x9091929394959697 );
-   H[ 3] = m512_const1_64( 0x98999A9B9C9D9E9F );
+   H[ 3] = _mm512_set1_epi64( 0x98999A9B9C9D9E9F );
-   H[ 4] = m512_const1_64( 0xA0A1A2A3A4A5A6A7 );
+   H[ 4] = _mm512_set1_epi64( 0xA0A1A2A3A4A5A6A7 );
-   H[ 5] = m512_const1_64( 0xA8A9AAABACADAEAF );
+   H[ 5] = _mm512_set1_epi64( 0xA8A9AAABACADAEAF );
-   H[ 6] = m512_const1_64( 0xB0B1B2B3B4B5B6B7 );
+   H[ 6] = _mm512_set1_epi64( 0xB0B1B2B3B4B5B6B7 );
-   H[ 7] = m512_const1_64( 0xB8B9BABBBCBDBEBF );
+   H[ 7] = _mm512_set1_epi64( 0xB8B9BABBBCBDBEBF );
-   H[ 8] = m512_const1_64( 0xC0C1C2C3C4C5C6C7 );
+   H[ 8] = _mm512_set1_epi64( 0xC0C1C2C3C4C5C6C7 );
-   H[ 9] = m512_const1_64( 0xC8C9CACBCCCDCECF );
+   H[ 9] = _mm512_set1_epi64( 0xC8C9CACBCCCDCECF );
-   H[10] = m512_const1_64( 0xD0D1D2D3D4D5D6D7 );
+   H[10] = _mm512_set1_epi64( 0xD0D1D2D3D4D5D6D7 );
-   H[11] = m512_const1_64( 0xD8D9DADBDCDDDEDF );
+   H[11] = _mm512_set1_epi64( 0xD8D9DADBDCDDDEDF );
-   H[12] = m512_const1_64( 0xE0E1E2E3E4E5E6E7 );
+   H[12] = _mm512_set1_epi64( 0xE0E1E2E3E4E5E6E7 );
-   H[13] = m512_const1_64( 0xE8E9EAEBECEDEEEF );
+   H[13] = _mm512_set1_epi64( 0xE8E9EAEBECEDEEEF );
-   H[14] = m512_const1_64( 0xF0F1F2F3F4F5F6F7 );
+   H[14] = _mm512_set1_epi64( 0xF0F1F2F3F4F5F6F7 );
-   H[15] = m512_const1_64( 0xF8F9FAFBFCFDFEFF );
+   H[15] = _mm512_set1_epi64( 0xF8F9FAFBFCFDFEFF );
 
 // Update
 
@@ -1530,7 +1530,7 @@ void bmw512_8way_full( bmw512_8way_context *ctx, void *out, const void *data,
    __m512i h1[16], h2[16];
    size_t u, v;
 
-   buf[ ptr>>3 ] = m512_const1_64( 0x80 );
+   buf[ ptr>>3 ] = _mm512_set1_epi64( 0x80 );
    ptr += 8;
 
    if (  ptr > (buf_size - 8) )

algo/bmw/sph_bmw.h

@@ -41,7 +41,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for BMW-224.

algo/cubehash/cube-hash-2way.c

@@ -221,14 +221,14 @@ int cube_4way_init( cube_4way_context *sp, int hashbitlen, int rounds,
     sp->rounds    = rounds;
     sp->pos       = 0;
 
-    h[ 0] = m512_const1_128( iv[0] );
+    h[ 0] = mm512_bcast_m128( iv[0] );
-    h[ 1] = m512_const1_128( iv[1] );
+    h[ 1] = mm512_bcast_m128( iv[1] );
-    h[ 2] = m512_const1_128( iv[2] );
+    h[ 2] = mm512_bcast_m128( iv[2] );
-    h[ 3] = m512_const1_128( iv[3] );
+    h[ 3] = mm512_bcast_m128( iv[3] );
-    h[ 4] = m512_const1_128( iv[4] );
+    h[ 4] = mm512_bcast_m128( iv[4] );
-    h[ 5] = m512_const1_128( iv[5] );
+    h[ 5] = mm512_bcast_m128( iv[5] );
-    h[ 6] = m512_const1_128( iv[6] );
+    h[ 6] = mm512_bcast_m128( iv[6] );
-    h[ 7] = m512_const1_128( iv[7] );
+    h[ 7] = mm512_bcast_m128( iv[7] );
 
     return 0;
 }
@@ -259,11 +259,11 @@ int cube_4way_close( cube_4way_context *sp, void *output )
 
     // 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 ) );
+                         mm512_bcast128lo_64( 0x0000000000000080 ) );
     transform_4way( sp );
 
     sp->h[7] = _mm512_xor_si512( sp->h[7],
-                                 m512_const2_64( 0x0000000100000000, 0 ) );
+                         mm512_bcast128hi_64( 0x0000000100000000 ) );
 
     for ( i = 0; i < 10; ++i ) 
        transform_4way( sp );
@@ -283,14 +283,14 @@ int cube_4way_full( cube_4way_context *sp, void *output,  int hashbitlen,
     sp->rounds    = 16;
     sp->pos       = 0;
 
-    h[ 0] = m512_const1_128( iv[0] );
+    h[ 0] = mm512_bcast_m128( iv[0] );
-    h[ 1] = m512_const1_128( iv[1] );
+    h[ 1] = mm512_bcast_m128( iv[1] );
-    h[ 2] = m512_const1_128( iv[2] );
+    h[ 2] = mm512_bcast_m128( iv[2] );
-    h[ 3] = m512_const1_128( iv[3] );
+    h[ 3] = mm512_bcast_m128( iv[3] );
-    h[ 4] = m512_const1_128( iv[4] );
+    h[ 4] = mm512_bcast_m128( iv[4] );
-    h[ 5] = m512_const1_128( iv[5] );
+    h[ 5] = mm512_bcast_m128( iv[5] );
-    h[ 6] = m512_const1_128( iv[6] );
+    h[ 6] = mm512_bcast_m128( iv[6] );
-    h[ 7] = m512_const1_128( iv[7] );
+    h[ 7] = mm512_bcast_m128( iv[7] );
 
     const int len = size >> 4;
     const __m512i *in = (__m512i*)data;
@@ -310,11 +310,11 @@ int cube_4way_full( cube_4way_context *sp, void *output,  int hashbitlen,
 
     // 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 ) );
+                         mm512_bcast128lo_64( 0x0000000000000080 ) );
     transform_4way( sp );
 
     sp->h[7] = _mm512_xor_si512( sp->h[7],
-                                    m512_const2_64( 0x0000000100000000, 0 ) );
+                         mm512_bcast128hi_64( 0x0000000100000000 ) );
 
     for ( i = 0; i < 10; ++i )
        transform_4way( sp );
@@ -336,14 +336,14 @@ int cube_4way_2buf_full( cube_4way_2buf_context *sp,
     sp->rounds    = 16;
     sp->pos       = 0;
 
-    h1[0] = h0[0] = m512_const1_128( iv[0] );
+    h1[0] = h0[0] = mm512_bcast_m128( iv[0] );
-    h1[1] = h0[1] = m512_const1_128( iv[1] );
+    h1[1] = h0[1] = mm512_bcast_m128( iv[1] );
-    h1[2] = h0[2] = m512_const1_128( iv[2] );
+    h1[2] = h0[2] = mm512_bcast_m128( iv[2] );
-    h1[3] = h0[3] = m512_const1_128( iv[3] );
+    h1[3] = h0[3] = mm512_bcast_m128( iv[3] );
-    h1[4] = h0[4] = m512_const1_128( iv[4] );
+    h1[4] = h0[4] = mm512_bcast_m128( iv[4] );
-    h1[5] = h0[5] = m512_const1_128( iv[5] );
+    h1[5] = h0[5] = mm512_bcast_m128( iv[5] );
-    h1[6] = h0[6] = m512_const1_128( iv[6] );
+    h1[6] = h0[6] = mm512_bcast_m128( iv[6] );
-    h1[7] = h0[7] = m512_const1_128( iv[7] );
+    h1[7] = h0[7] = mm512_bcast_m128( iv[7] );
 
     const int len = size >> 4;
     const __m512i *in0 = (__m512i*)data0;
@@ -365,13 +365,13 @@ int cube_4way_2buf_full( cube_4way_2buf_context *sp,
     }
 
     // pos is zero for 64 byte data, 1 for 80 byte data.
-    __m512i tmp = m512_const2_64( 0, 0x0000000000000080 );
+    __m512i tmp = mm512_bcast128lo_64( 0x0000000000000080 );
     sp->h0[ sp->pos ] = _mm512_xor_si512( sp->h0[ sp->pos ], tmp );
     sp->h1[ sp->pos ] = _mm512_xor_si512( sp->h1[ sp->pos ], tmp );
 
     transform_4way_2buf( sp );
 
-    tmp = m512_const2_64( 0x0000000100000000, 0 );
+    tmp = mm512_bcast128hi_64( 0x0000000100000000 );
     sp->h0[7] = _mm512_xor_si512( sp->h0[7], tmp );
     sp->h1[7] = _mm512_xor_si512( sp->h1[7], tmp );
 
@@ -384,7 +384,6 @@ int cube_4way_2buf_full( cube_4way_2buf_context *sp,
     return 0;
 }
 
-
 int cube_4way_update_close( cube_4way_context *sp, void *output,
                                const void *data, size_t size )
 {
@@ -406,11 +405,11 @@ int cube_4way_update_close( cube_4way_context *sp, void *output,
 
     // 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 ) );
+                          mm512_bcast128lo_64( 0x0000000000000080 ) );
     transform_4way( sp );
 
     sp->h[7] = _mm512_xor_si512( sp->h[7],
-                                    m512_const2_64( 0x0000000100000000, 0 ) );
+                          mm512_bcast128hi_64( 0x0000000100000000 ) );
 
     for ( i = 0; i < 10; ++i )
        transform_4way( sp );
@@ -424,21 +423,6 @@ int cube_4way_update_close( cube_4way_context *sp, void *output,
 
 // 2 way 128 
 
-// This isn't expected to be used with AVX512 so HW rotate intruction
-// is assumed not avaiable.
-// Use double buffering to optimize serial bit rotations. Full double
-// buffering isn't practical because it needs twice as many registers
-// with AVX2 having only half as many as AVX512.
-#define ROL2( out0, out1, in0, in1, c ) \
-{ \
- __m256i t0 = _mm256_slli_epi32( in0, c ); \
- __m256i t1 = _mm256_slli_epi32( in1, c ); \
- out0 = _mm256_srli_epi32( in0, 32-(c) ); \
- out1 = _mm256_srli_epi32( in1, 32-(c) ); \
- out0 = _mm256_or_si256( out0, t0 ); \
- out1 = _mm256_or_si256( out1, t1 ); \
-}
-
 static void transform_2way( cube_2way_context *sp )
 {
     int r;
@@ -461,8 +445,10 @@ 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 );
-        ROL2( y0, y1, x2, x3, 7 );
+        y0 = mm256_rol_32( x2, 7 );
-        ROL2( x2, x3, x0, x1, 7 );
+        y1 = mm256_rol_32( x3, 7 );
+        x2 = mm256_rol_32( x0, 7 );
+        x3 = mm256_rol_32( x1, 7 );
         x0 = _mm256_xor_si256( y0, x4 );
         x1 = _mm256_xor_si256( y1, x5 );
         x2 = _mm256_xor_si256( x2, x6 );
@@ -475,8 +461,10 @@ 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 );
-        ROL2( y0, x1, x1, x0, 11 );
+        y0 = mm256_rol_32( x1, 11 );
-        ROL2( y1, x3, x3, x2, 11 );
+        x1 = mm256_rol_32( x0, 11 );
+        y1 = mm256_rol_32( x3, 11 );
+        x3 = mm256_rol_32( x2, 11 );
         x0 = _mm256_xor_si256( y0, x4 );
         x1 = _mm256_xor_si256( x1, x5 );
         x2 = _mm256_xor_si256( y1, x6 );
@@ -508,14 +496,14 @@ int cube_2way_init( cube_2way_context *sp, int hashbitlen, int rounds,
     sp->rounds    = rounds;
     sp->pos       = 0;
 
-    h[ 0] = m256_const1_128( iv[0] );
+    h[ 0] = mm256_bcast_m128( iv[0] );
-    h[ 1] = m256_const1_128( iv[1] );
+    h[ 1] = mm256_bcast_m128( iv[1] );
-    h[ 2] = m256_const1_128( iv[2] );
+    h[ 2] = mm256_bcast_m128( iv[2] );
-    h[ 3] = m256_const1_128( iv[3] );
+    h[ 3] = mm256_bcast_m128( iv[3] );
-    h[ 4] = m256_const1_128( iv[4] );
+    h[ 4] = mm256_bcast_m128( iv[4] );
-    h[ 5] = m256_const1_128( iv[5] );
+    h[ 5] = mm256_bcast_m128( iv[5] );
-    h[ 6] = m256_const1_128( iv[6] );
+    h[ 6] = mm256_bcast_m128( iv[6] );
-    h[ 7] = m256_const1_128( iv[7] );
+    h[ 7] = mm256_bcast_m128( iv[7] );
     
     return 0;
 }
@@ -546,13 +534,14 @@ 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 ],
-                                   m256_const2_64( 0, 0x0000000000000080 ) );
+                                   mm256_bcast128lo_64( 0x0000000000000080 ) );
     transform_2way( sp );
 
     sp->h[7] = _mm256_xor_si256( sp->h[7],
-                                   m256_const2_64( 0x0000000100000000, 0 ) );
+                                   mm256_bcast128hi_64( 0x0000000100000000 ) );
 
-    for ( i = 0; i < 10; ++i )           transform_2way( sp );
+    for ( i = 0; i < 10; ++i )  
+       transform_2way( sp );
 
     memcpy( hash, sp->h, sp->hashlen<<5 );
     return 0;
@@ -579,13 +568,14 @@ 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 ],
-                                    m256_const2_64( 0, 0x0000000000000080 ) );
+                                    mm256_bcast128lo_64( 0x0000000000000080 ) );
     transform_2way( sp );
 
     sp->h[7] = _mm256_xor_si256( sp->h[7],
-                                    m256_const2_64( 0x0000000100000000, 0 ) );
+                                    mm256_bcast128hi_64( 0x0000000100000000 ) );
 
-    for ( i = 0; i < 10; ++i )    transform_2way( sp );
+    for ( i = 0; i < 10; ++i )
+       transform_2way( sp );
 
     memcpy( hash, sp->h, sp->hashlen<<5 );
     return 0;
@@ -602,14 +592,14 @@ int cube_2way_full( cube_2way_context *sp, void *output, int hashbitlen,
     sp->rounds    = 16;
     sp->pos       = 0;
 
-    h[ 0] = m256_const1_128( iv[0] );
+    h[ 0] = mm256_bcast_m128( iv[0] );
-    h[ 1] = m256_const1_128( iv[1] );
+    h[ 1] = mm256_bcast_m128( iv[1] );
-    h[ 2] = m256_const1_128( iv[2] );
+    h[ 2] = mm256_bcast_m128( iv[2] );
-    h[ 3] = m256_const1_128( iv[3] );
+    h[ 3] = mm256_bcast_m128( iv[3] );
-    h[ 4] = m256_const1_128( iv[4] );
+    h[ 4] = mm256_bcast_m128( iv[4] );
-    h[ 5] = m256_const1_128( iv[5] );
+    h[ 5] = mm256_bcast_m128( iv[5] );
-    h[ 6] = m256_const1_128( iv[6] );
+    h[ 6] = mm256_bcast_m128( iv[6] );
-    h[ 7] = m256_const1_128( iv[7] );
+    h[ 7] = mm256_bcast_m128( iv[7] );
 
     const int len = size >> 4;
     const __m256i *in = (__m256i*)data;
@@ -629,13 +619,14 @@ int cube_2way_full( cube_2way_context *sp, void *output, int hashbitlen,
 
     // pos is zero for 64 byte data, 1 for 80 byte data.
     sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ],
-                                    m256_const2_64( 0, 0x0000000000000080 ) );
+                                    mm256_bcast128lo_64( 0x0000000000000080 ) );
     transform_2way( sp );
 
     sp->h[7] = _mm256_xor_si256( sp->h[7],
-                                    m256_const2_64( 0x0000000100000000, 0 ) );
+                                    mm256_bcast128hi_64( 0x0000000100000000 ) );
 
-    for ( i = 0; i < 10; ++i )    transform_2way( sp );
+    for ( i = 0; i < 10; ++i )
+       transform_2way( sp );
 
     memcpy( hash, sp->h, sp->hashlen<<5 );
     return 0;

algo/cubehash/cubehash_sse2.c

@@ -9,7 +9,6 @@
 #include <immintrin.h>
 #endif
 #include "cubehash_sse2.h"
-#include "algo/sha/sha3-defs.h"
 #include <stdbool.h>
 #include <unistd.h>
 #include <memory.h>
@@ -32,7 +31,7 @@ static void transform( cubehashParam *sp )
     { 
         x1 = _mm512_add_epi32( x0, x1 );
         x0 = mm512_swap_256( x0 );
-        x0 = mm512_rol_32(  x0, 7 );
+        x0 = mm512_rol_32( x0, 7 );
         x0 = _mm512_xor_si512( x0, x1 );
         x1 = mm512_swap128_64( x1 );
         x1 = _mm512_add_epi32( x0, x1 );
@@ -58,19 +57,18 @@ static void transform( cubehashParam *sp )
     { 
         x2 = _mm256_add_epi32( x0, x2 );
         x3 = _mm256_add_epi32( x1, x3 );
-        y0 = x0;
+        y0 = mm256_rol_32( x1, 7 );
-        x0 = mm256_rol_32( x1, 7 );
+        y1 = mm256_rol_32( x0, 7 );
-        x1 = mm256_rol_32( y0, 7 );
+        x0 = _mm256_xor_si256( y0, x2 );
-        x0 = _mm256_xor_si256( x0, x2 );
+        x1 = _mm256_xor_si256( y1, x3 );
-        x1 = _mm256_xor_si256( x1, x3 );
         x2 = mm256_swap128_64( x2 );
         x3 = mm256_swap128_64( x3 );
         x2 = _mm256_add_epi32( x0, x2 );
         x3 = _mm256_add_epi32( x1, x3 );
-        y0 = mm256_swap_128( x0 );
+        x0 = mm256_swap_128( x0 );
-        y1 = mm256_swap_128( x1 );
+        x1 = mm256_swap_128( x1 );
-        x0 = mm256_rol_32( y0, 11 );
+        x0 = mm256_rol_32( x0, 11 );
-        x1 = mm256_rol_32( y1, 11 );
+        x1 = mm256_rol_32( x1, 11 );
         x0 = _mm256_xor_si256( x0, x2 );
         x1 = _mm256_xor_si256( x1, x3 );
         x2 = mm256_swap64_32( x2 );
@@ -94,47 +92,48 @@ static void transform( cubehashParam *sp )
     x6 = _mm_load_si128( (__m128i*)sp->x + 6 );
     x7 = _mm_load_si128( (__m128i*)sp->x + 7 );
 
-    for (r = 0; r < rounds; ++r) {
+    for ( r = 0; r < rounds; ++r )
-	x4 = _mm_add_epi32(x0, x4);
+    {
-	x5 = _mm_add_epi32(x1, x5);
+       x4 = _mm_add_epi32( x0, x4 );
-	x6 = _mm_add_epi32(x2, x6);
+       x5 = _mm_add_epi32( x1, x5 );
-	x7 = _mm_add_epi32(x3, x7);
+       x6 = _mm_add_epi32( x2, x6 );
-	y0 = x2;
+       x7 = _mm_add_epi32( x3, x7 );
-	y1 = x3;
+       y0 = x2;
-	y2 = x0;
+       y1 = x3;
-	y3 = x1;
+       y2 = x0;
-	x0 = _mm_xor_si128(_mm_slli_epi32(y0, 7), _mm_srli_epi32(y0, 25));
+       y3 = x1;
-	x1 = _mm_xor_si128(_mm_slli_epi32(y1, 7), _mm_srli_epi32(y1, 25));
+       x0 = mm128_rol_32( y0, 7 );
-	x2 = _mm_xor_si128(_mm_slli_epi32(y2, 7), _mm_srli_epi32(y2, 25));
+       x1 = mm128_rol_32( y1, 7 );
-	x3 = _mm_xor_si128(_mm_slli_epi32(y3, 7), _mm_srli_epi32(y3, 25));
+       x2 = mm128_rol_32( y2, 7 );
-	x0 = _mm_xor_si128(x0, x4);
+       x3 = mm128_rol_32( y3, 7 );
-	x1 = _mm_xor_si128(x1, x5);
+       x0 = _mm_xor_si128( x0, x4 );
-	x2 = _mm_xor_si128(x2, x6);
+       x1 = _mm_xor_si128( x1, x5 );
-	x3 = _mm_xor_si128(x3, x7);
+       x2 = _mm_xor_si128( x2, x6 );
-	x4 = _mm_shuffle_epi32(x4, 0x4e);
+       x3 = _mm_xor_si128( x3, x7 );
-	x5 = _mm_shuffle_epi32(x5, 0x4e);
+       x4 = _mm_shuffle_epi32( x4, 0x4e );
-	x6 = _mm_shuffle_epi32(x6, 0x4e);
+       x5 = _mm_shuffle_epi32( x5, 0x4e );
-	x7 = _mm_shuffle_epi32(x7, 0x4e);
+       x6 = _mm_shuffle_epi32( x6, 0x4e );
-	x4 = _mm_add_epi32(x0, x4);
+       x7 = _mm_shuffle_epi32( x7, 0x4e );
-	x5 = _mm_add_epi32(x1, x5);
+       x4 = _mm_add_epi32( x0, x4 );
-	x6 = _mm_add_epi32(x2, x6);
+       x5 = _mm_add_epi32( x1, x5 );
-	x7 = _mm_add_epi32(x3, x7);
+       x6 = _mm_add_epi32( x2, x6 );
-	y0 = x1;
+       x7 = _mm_add_epi32( x3, x7 );
-	y1 = x0;
+       y0 = x1;
-	y2 = x3;
+       y1 = x0;
-	y3 = x2;
+       y2 = x3;
-	x0 = _mm_xor_si128(_mm_slli_epi32(y0, 11), _mm_srli_epi32(y0, 21));
+       y3 = x2;
-	x1 = _mm_xor_si128(_mm_slli_epi32(y1, 11), _mm_srli_epi32(y1, 21));
+       x0 = mm128_rol_32( y0, 11 );
-	x2 = _mm_xor_si128(_mm_slli_epi32(y2, 11), _mm_srli_epi32(y2, 21));
+       x1 = mm128_rol_32( y1, 11 );
-	x3 = _mm_xor_si128(_mm_slli_epi32(y3, 11), _mm_srli_epi32(y3, 21));
+       x2 = mm128_rol_32( y2, 11 );
-	x0 = _mm_xor_si128(x0, x4);
+       x3 = mm128_rol_32( y3, 11 );
-	x1 = _mm_xor_si128(x1, x5);
+	    x0 = _mm_xor_si128( x0, x4 );
-	x2 = _mm_xor_si128(x2, x6);
+	    x1 = _mm_xor_si128( x1, x5 );
-	x3 = _mm_xor_si128(x3, x7);
+	    x2 = _mm_xor_si128( x2, x6 );
-	x4 = _mm_shuffle_epi32(x4, 0xb1);
+	    x3 = _mm_xor_si128( x3, x7 );
-	x5 = _mm_shuffle_epi32(x5, 0xb1);
+	    x4 = _mm_shuffle_epi32( x4, 0xb1 );
-	x6 = _mm_shuffle_epi32(x6, 0xb1);
+	    x5 = _mm_shuffle_epi32( x5, 0xb1 );
-	x7 = _mm_shuffle_epi32(x7, 0xb1);
+	    x6 = _mm_shuffle_epi32( x6, 0xb1 );
+	    x7 = _mm_shuffle_epi32( x7, 0xb1 );
     }
 
     _mm_store_si128( (__m128i*)sp->x,     x0 );
@@ -180,25 +179,25 @@ int cubehashInit(cubehashParam *sp, int hashbitlen, int rounds, int blockbytes)
     if ( hashbitlen == 512 )
     {
 
-       x[0] = m128_const_64( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
+       x[0] = _mm_set_epi64x( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
-       x[1] = m128_const_64( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
+       x[1] = _mm_set_epi64x( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
-       x[2] = m128_const_64( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
+       x[2] = _mm_set_epi64x( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
-       x[3] = m128_const_64( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
+       x[3] = _mm_set_epi64x( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
-       x[4] = m128_const_64( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
+       x[4] = _mm_set_epi64x( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
-       x[5] = m128_const_64( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
+       x[5] = _mm_set_epi64x( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
-       x[6] = m128_const_64( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
+       x[6] = _mm_set_epi64x( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
-       x[7] = m128_const_64( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
+       x[7] = _mm_set_epi64x( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
     }
     else
     {
-       x[0] = m128_const_64( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
+       x[0] = _mm_set_epi64x( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
-       x[1] = m128_const_64( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
+       x[1] = _mm_set_epi64x( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
-       x[2] = m128_const_64( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
+       x[2] = _mm_set_epi64x( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
-       x[3] = m128_const_64( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
+       x[3] = _mm_set_epi64x( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
-       x[4] = m128_const_64( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
+       x[4] = _mm_set_epi64x( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
-       x[5] = m128_const_64( 0x93CB628565C892FD, 0x5FA2560309392549 );
+       x[5] = _mm_set_epi64x( 0x93CB628565C892FD, 0x5FA2560309392549 );
-       x[6] = m128_const_64( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
+       x[6] = _mm_set_epi64x( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
-       x[7] = m128_const_64( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
+       x[7] = _mm_set_epi64x( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
     }   
 
     return SUCCESS;
@@ -234,10 +233,10 @@ int cubehashDigest( cubehashParam *sp, byte *digest )
 
     // pos is zero for 64 byte data, 1 for 80 byte data.
     sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ],
-                                      m128_const_64( 0, 0x80 ) );
+                                      _mm_set_epi64x( 0, 0x80 ) );
     transform( sp );
 
-    sp->x[7] = _mm_xor_si128( sp->x[7], m128_const_64( 0x100000000, 0 ) );
+    sp->x[7] = _mm_xor_si128( sp->x[7], _mm_set_epi64x( 0x100000000, 0 ) );
     transform( sp );
     transform( sp );
     transform( sp );
@@ -279,10 +278,10 @@ int cubehashUpdateDigest( cubehashParam *sp, byte *digest,
 
     // pos is zero for 64 byte data, 1 for 80 byte data.
     sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ],
-                                      m128_const_64( 0, 0x80 ) );
+                                      _mm_set_epi64x( 0, 0x80 ) );
     transform( sp );
 
-    sp->x[7] = _mm_xor_si128( sp->x[7], m128_const_64( 0x100000000, 0 ) );
+    sp->x[7] = _mm_xor_si128( sp->x[7], _mm_set_epi64x( 0x100000000, 0 ) );
 
     transform( sp );
     transform( sp );
@@ -313,25 +312,25 @@ int cubehash_full( cubehashParam *sp, byte *digest, int hashbitlen,
     if ( hashbitlen == 512 )
     {
 
-       x[0] = m128_const_64( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
+       x[0] = _mm_set_epi64x( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
-       x[1] = m128_const_64( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
+       x[1] = _mm_set_epi64x( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
-       x[2] = m128_const_64( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
+       x[2] = _mm_set_epi64x( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
-       x[3] = m128_const_64( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
+       x[3] = _mm_set_epi64x( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
-       x[4] = m128_const_64( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
+       x[4] = _mm_set_epi64x( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
-       x[5] = m128_const_64( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
+       x[5] = _mm_set_epi64x( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
-       x[6] = m128_const_64( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
+       x[6] = _mm_set_epi64x( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
-       x[7] = m128_const_64( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
+       x[7] = _mm_set_epi64x( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
     }
     else
     {
-       x[0] = m128_const_64( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
+       x[0] = _mm_set_epi64x( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
-       x[1] = m128_const_64( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
+       x[1] = _mm_set_epi64x( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
-       x[2] = m128_const_64( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
+       x[2] = _mm_set_epi64x( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
-       x[3] = m128_const_64( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
+       x[3] = _mm_set_epi64x( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
-       x[4] = m128_const_64( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
+       x[4] = _mm_set_epi64x( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
-       x[5] = m128_const_64( 0x93CB628565C892FD, 0x5FA2560309392549 );
+       x[5] = _mm_set_epi64x( 0x93CB628565C892FD, 0x5FA2560309392549 );
-       x[6] = m128_const_64( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
+       x[6] = _mm_set_epi64x( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
-       x[7] = m128_const_64( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
+       x[7] = _mm_set_epi64x( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
     }
 
 
@@ -358,10 +357,10 @@ int cubehash_full( cubehashParam *sp, byte *digest, int hashbitlen,
 
     // pos is zero for 64 byte data, 1 for 80 byte data.
     sp->x[ sp->pos ] = _mm_xor_si128( sp->x[ sp->pos ],
-                                      m128_const_64( 0, 0x80 ) );
+                                      _mm_set_epi64x( 0, 0x80 ) );
     transform( sp );
 
-    sp->x[7] = _mm_xor_si128( sp->x[7], m128_const_64( 0x100000000, 0 ) );
+    sp->x[7] = _mm_xor_si128( sp->x[7], _mm_set_epi64x( 0x100000000, 0 ) );
 
     transform( sp );
     transform( sp );

algo/cubehash/cubehash_sse2.h

@@ -3,7 +3,7 @@
 
 #include "compat.h"
 #include <stdint.h>
-#include "algo/sha/sha3-defs.h"
+#include "compat/sha3-defs.h"
 
 #define	OPTIMIZE_SSE2
 

algo/cubehash/sph_cubehash.h

@@ -42,7 +42,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for CubeHash-224.

algo/echo/aes_ni/hash.c

@@ -566,16 +566,16 @@ HashReturn echo_full( hashState_echo *state, BitSequence *hashval,
          state->uHashSize = 256;
          state->uBlockLength = 192;
          state->uRounds = 8;
-         state->hashsize = m128_const_64( 0, 0x100 );
+         state->hashsize = _mm_set_epi64x( 0, 0x100 );
-         state->const1536 = m128_const_64( 0, 0x600 );
+         state->const1536 = _mm_set_epi64x( 0, 0x600 );
          break;
 
       case 512:
          state->uHashSize = 512;
          state->uBlockLength = 128;
          state->uRounds = 10;
-         state->hashsize = m128_const_64( 0, 0x200 );
+         state->hashsize = _mm_set_epi64x( 0, 0x200 );
-         state->const1536 = m128_const_64( 0, 0x400 );
+         state->const1536 = _mm_set_epi64x( 0, 0x400 );
          break;
 
       default:

algo/echo/aes_ni/hash_api.h

@@ -22,7 +22,7 @@
 #endif
 
 
-#include "algo/sha/sha3_common.h"
+#include "compat/sha3_common.h"
 
 #include <emmintrin.h>
 

algo/echo/echo-hash-4way.c

@@ -162,9 +162,9 @@ void echo_4way_compress( echo_4way_context *ctx, const __m512i *pmsg,
   unsigned int r, b, i, j;
   __m512i t1, t2, s2, k1;
   __m512i _state[4][4], _state2[4][4], _statebackup[4][4]; 
-  __m512i one = m512_one_128;
+  const __m512i one = mm512_bcast128lo_64( 1 ); 
-  __m512i mul2mask = m512_const2_64( 0, 0x00001b00 );
+  const __m512i mul2mask = mm512_bcast128lo_64( 0x00001b00 );
-  __m512i lsbmask  = m512_const1_32( 0x01010101 ); 
+  const __m512i lsbmask  = _mm512_set1_epi32( 0x01010101 ); 
 
   _state[ 0 ][ 0 ] = ctx->state[ 0 ][ 0 ];
   _state[ 0 ][ 1 ] = ctx->state[ 0 ][ 1 ];
@@ -264,16 +264,16 @@ int echo_4way_init( echo_4way_context *ctx, int nHashSize )
 		ctx->uHashSize = 256;
 		ctx->uBlockLength = 192;
 		ctx->uRounds = 8;
-		ctx->hashsize = m512_const2_64( 0, 0x100 );
+      ctx->hashsize = mm512_bcast128lo_64( 0x100 );
-		ctx->const1536 = m512_const2_64( 0, 0x600 );
+      ctx->const1536 = mm512_bcast128lo_64( 0x600 );
 		break;
 
 	case 512:
 		ctx->uHashSize = 512;
 		ctx->uBlockLength = 128;
 		ctx->uRounds = 10;
-		ctx->hashsize = m512_const2_64( 0, 0x200 );
+      ctx->hashsize = mm512_bcast128lo_64( 0x200 );
-		ctx->const1536 = m512_const2_64( 0, 0x400);
+      ctx->const1536 = mm512_bcast128lo_64( 0x400);
 		break;
 
 	default:
@@ -305,7 +305,7 @@ int echo_4way_update_close( echo_4way_context *state, void *hashval,
    {
       echo_4way_compress( state, data, 1 );
       state->processed_bits = 1024;
-      remainingbits = m512_const2_64( 0, -1024 );
+      remainingbits = mm512_bcast128lo_64( -1024 );
       vlen = 0;
    }
    else
@@ -313,13 +313,15 @@ int echo_4way_update_close( echo_4way_context *state, void *hashval,
       vlen = databitlen / 128;  // * 4 lanes / 128 bits per lane
       memcpy_512( state->buffer, data, vlen );
       state->processed_bits += (unsigned int)( databitlen );
-      remainingbits = m512_const2_64( 0, (uint64_t)databitlen );
+      remainingbits = mm512_bcast128lo_64( (uint64_t)databitlen );
    }
 
-   state->buffer[ vlen ] = m512_const2_64( 0, 0x80 );
+   state->buffer[ vlen ] = mm512_bcast128lo_64( 0x80 );
    memset_zero_512( state->buffer + vlen + 1, vblen - vlen - 2 );
-   state->buffer[ vblen-2 ] = m512_const2_64( (uint64_t)state->uHashSize << 48, 0 );
+   state->buffer[ vblen-2 ] =
-   state->buffer[ vblen-1 ] = m512_const2_64( 0, state->processed_bits);
+           mm512_bcast128hi_64( (uint64_t)state->uHashSize << 48 );
+   state->buffer[ vblen-1 ] =
+           mm512_bcast128lo_64( state->processed_bits );
 
    state->k = _mm512_add_epi64( state->k, remainingbits );
    state->k = _mm512_sub_epi64( state->k, state->const1536 );
@@ -352,16 +354,16 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
          ctx->uHashSize = 256;
          ctx->uBlockLength = 192;
          ctx->uRounds = 8;
-         ctx->hashsize = m512_const2_64( 0, 0x100 );
+         ctx->hashsize = mm512_bcast128lo_64( 0x100 );
-         ctx->const1536 = m512_const2_64( 0, 0x600 );
+         ctx->const1536 = mm512_bcast128lo_64( 0x600 );
          break;
 
       case 512:
          ctx->uHashSize = 512;
          ctx->uBlockLength = 128;
          ctx->uRounds = 10;
-         ctx->hashsize = m512_const2_64( 0, 0x200 );
+         ctx->hashsize = mm512_bcast128lo_64( 0x200 );
-         ctx->const1536 = m512_const2_64( 0, 0x400 );
+         ctx->const1536 = mm512_bcast128lo_64( 0x400 );
          break;
 
       default:
@@ -388,7 +390,7 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
    {
       echo_4way_compress( ctx, data, 1 );
       ctx->processed_bits = 1024;
-      remainingbits = m512_const2_64( 0, -1024 );
+      remainingbits = mm512_bcast128lo_64( -1024 );
       vlen = 0;
    }
    else
@@ -396,14 +398,14 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
       vlen = databitlen / 128;  // * 4 lanes / 128 bits per lane
       memcpy_512( ctx->buffer, data, vlen );
       ctx->processed_bits += (unsigned int)( databitlen );
-      remainingbits = m512_const2_64( 0, databitlen );
+      remainingbits = mm512_bcast128lo_64( databitlen );
    }
 
-   ctx->buffer[ vlen ] = m512_const2_64( 0, 0x80 );
+   ctx->buffer[ vlen ] = mm512_bcast128lo_64( 0x80 );
    memset_zero_512( ctx->buffer + vlen + 1, vblen - vlen - 2 );
    ctx->buffer[ vblen-2 ] =
-                     m512_const2_64( (uint64_t)ctx->uHashSize << 48, 0 );
+               mm512_bcast128hi_64( (uint64_t)ctx->uHashSize << 48 );
-   ctx->buffer[ vblen-1 ] = m512_const2_64( 0, ctx->processed_bits);
+   ctx->buffer[ vblen-1 ] = mm512_bcast128lo_64( ctx->processed_bits);
 
    ctx->k = _mm512_add_epi64( ctx->k, remainingbits );
    ctx->k = _mm512_sub_epi64( ctx->k, ctx->const1536 );
@@ -425,9 +427,9 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
 
 // AVX2 + VAES
 
-#define mul2mask_2way   m256_const2_64( 0, 0x0000000000001b00 ) 
+#define mul2mask_2way   mm256_bcast128lo_64( 0x0000000000001b00 ) 
 
-#define lsbmask_2way    m256_const1_32( 0x01010101 ) 
+#define lsbmask_2way    _mm256_set1_epi32( 0x01010101 ) 
 
 #define ECHO_SUBBYTES4_2WAY( state, j ) \
    state[0][j] = _mm256_aesenc_epi128( state[0][j], k1 ); \
@@ -467,8 +469,7 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
    t1 = _mm256_and_si256( t1, lsbmask_2way ); \
    t2 = _mm256_shuffle_epi8( mul2mask_2way, t1 ); \
    s2 = _mm256_xor_si256( s2, t2 );\
-   state2[ 0 ][ j ] = _mm256_xor_si256( state2[ 0 ][ j ], \
+   state2[ 0 ][ j ] = mm256_xor3( state2[ 0 ][ j ], s2, state1[ 1 ][ j1 ] ); \
-                              _mm256_xor_si256( s2, state1[ 1 ][ j1 ] ) ); \
    state2[ 1 ][ j ] = _mm256_xor_si256( state2[ 1 ][ j ], s2 ); \
    state2[ 2 ][ j ] = _mm256_xor_si256( state2[ 2 ][ j ], state1[ 1 ][ j1 ] ); \
    state2[ 3 ][ j ] = _mm256_xor_si256( state2[ 3 ][ j ], state1[ 1 ][ j1 ] ); \
@@ -478,8 +479,7 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
    t2 = _mm256_shuffle_epi8( mul2mask_2way, t1 ); \
    s2 = _mm256_xor_si256( s2, t2 ); \
    state2[ 0 ][ j ] = _mm256_xor_si256( state2[ 0 ][ j ], state1[ 2 ][ j2 ] ); \
-   state2[ 1 ][ j ] = _mm256_xor_si256( state2[ 1 ][ j ], \
+   state2[ 1 ][ j ] = mm256_xor3( state2[ 1 ][ j ], s2, state1[ 2 ][ j2 ] ); \
-                            _mm256_xor_si256( s2, state1[ 2 ][ j2 ] ) ); \
    state2[ 2 ][ j ] = _mm256_xor_si256( state2[ 2 ][ j ], s2 ); \
    state2[ 3 ][ j ] = _mm256_xor_si256( state2[ 3][ j ], state1[ 2 ][ j2 ] ); \
    s2 = _mm256_add_epi8( state1[ 3 ][ j3 ], state1[ 3 ][ j3 ] ); \
@@ -489,8 +489,7 @@ int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
    s2 = _mm256_xor_si256( s2, t2 ); \
    state2[ 0 ][ j ] = _mm256_xor_si256( state2[ 0 ][ j ], state1[ 3 ][ j3 ] ); \
    state2[ 1 ][ j ] = _mm256_xor_si256( state2[ 1 ][ j ], state1[ 3 ][ j3 ] ); \
-   state2[ 2 ][ j ] = _mm256_xor_si256( state2[ 2 ][ j ], \
+   state2[ 2 ][ j ] = mm256_xor3( state2[ 2 ][ j ], s2, state1[ 3 ][ j3] ); \
-                            _mm256_xor_si256( s2, state1[ 3 ][ j3] ) ); \
    state2[ 3 ][ j ] = _mm256_xor_si256( state2[ 3 ][ j ], s2 ); \
 } while(0)
 
@@ -679,16 +678,16 @@ int echo_2way_init( echo_2way_context *ctx, int nHashSize )
                         ctx->uHashSize = 256;
                         ctx->uBlockLength = 192;
                         ctx->uRounds = 8;
-                        ctx->hashsize = m256_const2_64( 0, 0x100 );
+                        ctx->hashsize = mm256_bcast128lo_64( 0x100 );
-                        ctx->const1536 = m256_const2_64( 0, 0x600 );
+                        ctx->const1536 = mm256_bcast128lo_64( 0x600 );
                         break;
 
                 case 512:
                         ctx->uHashSize = 512;
                         ctx->uBlockLength = 128;
                         ctx->uRounds = 10;
-                        ctx->hashsize = m256_const2_64( 0, 0x200 );
+                        ctx->hashsize = mm256_bcast128lo_64( 0x200 );
-                        ctx->const1536 = m256_const2_64( 0, 0x400 );
+                        ctx->const1536 = mm256_bcast128lo_64( 0x400 );
                         break;
 
                 default:
@@ -720,20 +719,20 @@ int echo_2way_update_close( echo_2way_context *state, void *hashval,
    {
       echo_2way_compress( state, data, 1 );
       state->processed_bits = 1024;
-      remainingbits = m256_const2_64( 0, -1024 );
+      remainingbits = mm256_bcast128lo_64( -1024 );
       vlen = 0;
    }
    else
    {
       memcpy_256( state->buffer, data, vlen );
       state->processed_bits += (unsigned int)( databitlen );
-      remainingbits = m256_const2_64( 0, databitlen );
+      remainingbits = mm256_bcast128lo_64( databitlen );
    }
 
-   state->buffer[ vlen ] = m256_const2_64( 0, 0x80 );
+   state->buffer[ vlen ] = mm256_bcast128lo_64( 0x80 );
    memset_zero_256( state->buffer + vlen + 1, vblen - vlen - 2 );
-   state->buffer[ vblen-2 ] = m256_const2_64( (uint64_t)state->uHashSize << 48, 0 );
+   state->buffer[ vblen-2 ] = mm256_bcast128hi_64( (uint64_t)state->uHashSize << 48 );
-   state->buffer[ vblen-1 ] = m256_const2_64( 0, state->processed_bits );
+   state->buffer[ vblen-1 ] = mm256_bcast128lo_64( state->processed_bits );
 
    state->k = _mm256_add_epi64( state->k, remainingbits );
    state->k = _mm256_sub_epi64( state->k, state->const1536 );
@@ -766,16 +765,16 @@ int echo_2way_full( echo_2way_context *ctx, void *hashval, int nHashSize,
          ctx->uHashSize = 256;
          ctx->uBlockLength = 192;
          ctx->uRounds = 8;
-         ctx->hashsize = m256_const2_64( 0, 0x100 );
+         ctx->hashsize = mm256_bcast128lo_64( 0x100 );
-         ctx->const1536 = m256_const2_64( 0, 0x600 );
+         ctx->const1536 = mm256_bcast128lo_64( 0x600 );
          break;
 
       case 512:
          ctx->uHashSize = 512;
          ctx->uBlockLength = 128;
          ctx->uRounds = 10;
-         ctx->hashsize = m256_const2_64( 0, 0x200 );
+         ctx->hashsize = mm256_bcast128lo_64( 0x200 );
-         ctx->const1536 = m256_const2_64( 0, 0x400 );
+         ctx->const1536 = mm256_bcast128lo_64( 0x400 );
          break;
 
       default:
@@ -798,7 +797,7 @@ int echo_2way_full( echo_2way_context *ctx, void *hashval, int nHashSize,
    {
       echo_2way_compress( ctx, data, 1 );
       ctx->processed_bits = 1024;
-      remainingbits = m256_const2_64( 0, -1024 );
+      remainingbits = mm256_bcast128lo_64( -1024 );
       vlen = 0;
    }
    else
@@ -806,13 +805,13 @@ int echo_2way_full( echo_2way_context *ctx, void *hashval, int nHashSize,
       vlen = databitlen / 128;  // * 4 lanes / 128 bits per lane
       memcpy_256( ctx->buffer, data, vlen );
       ctx->processed_bits += (unsigned int)( databitlen );
-      remainingbits = m256_const2_64( 0, databitlen );
+      remainingbits = mm256_bcast128lo_64( databitlen );
    }
 
-   ctx->buffer[ vlen ] = m256_const2_64( 0, 0x80 );
+   ctx->buffer[ vlen ] = mm256_bcast128lo_64( 0x80 );
    memset_zero_256( ctx->buffer + vlen + 1, vblen - vlen - 2 );
-   ctx->buffer[ vblen-2 ] = m256_const2_64( (uint64_t)ctx->uHashSize << 48, 0 );
+   ctx->buffer[ vblen-2 ] = mm256_bcast128hi_64( (uint64_t)ctx->uHashSize << 48 );
-   ctx->buffer[ vblen-1 ] = m256_const2_64( 0, ctx->processed_bits );
+   ctx->buffer[ vblen-1 ] = mm256_bcast128lo_64( ctx->processed_bits );
 
    ctx->k = _mm256_add_epi64( ctx->k, remainingbits );
    ctx->k = _mm256_sub_epi64( ctx->k, ctx->const1536 );

algo/echo/sph_echo.c

@@ -73,7 +73,7 @@ extern "C"{
 #endif
 
 #define AES_BIG_ENDIAN   0
-#include "algo/sha/aes_helper.c"
+#include "compat/aes_helper.c"
 
 #if SPH_ECHO_64
 

algo/echo/sph_echo.h

@@ -43,7 +43,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for ECHO-224.

algo/fugue/fugue-aesni.c

@@ -33,11 +33,11 @@ MYALIGN const unsigned long long _supermix4b[]	= {0x07020d08080e0d0d, 0x07070908
 MYALIGN const unsigned long long _supermix4c[]	= {0x0706050403020000, 0x0302000007060504};
 MYALIGN const unsigned long long _supermix7a[]	= {0x010c0b060d080702, 0x0904030e03000104};
 MYALIGN const unsigned long long _supermix7b[]	= {0x8080808080808080, 0x0504070605040f06};
-MYALIGN const unsigned long long _k_n[] = {0x4E4E4E4E4E4E4E4E, 0x1B1B1B1B0E0E0E0E};
+//MYALIGN const unsigned long long _k_n[] = {0x4E4E4E4E4E4E4E4E, 0x1B1B1B1B0E0E0E0E};
-MYALIGN const unsigned char _shift_one_mask[]   = {7, 4, 5, 6, 11, 8, 9, 10, 15, 12, 13, 14, 3, 0, 1, 2};
+//MYALIGN const unsigned char _shift_one_mask[]   = {7, 4, 5, 6, 11, 8, 9, 10, 15, 12, 13, 14, 3, 0, 1, 2};
-MYALIGN const unsigned char _shift_four_mask[]  = {13, 14, 15, 12, 1, 2, 3, 0, 5, 6, 7, 4, 9, 10, 11, 8};
+//MYALIGN const unsigned char _shift_four_mask[]  = {13, 14, 15, 12, 1, 2, 3, 0, 5, 6, 7, 4, 9, 10, 11, 8};
-MYALIGN const unsigned char _shift_seven_mask[] = {10, 11, 8, 9, 14, 15, 12, 13, 2, 3, 0, 1, 6, 7, 4, 5};
+//MYALIGN const unsigned char _shift_seven_mask[] = {10, 11, 8, 9, 14, 15, 12, 13, 2, 3, 0, 1, 6, 7, 4, 5};
-MYALIGN const unsigned char _aes_shift_rows[]   = {0, 5, 10, 15, 4, 9, 14, 3, 8, 13, 2, 7, 12, 1, 6, 11};
+//MYALIGN const unsigned char _aes_shift_rows[]   = {0, 5, 10, 15, 4, 9, 14, 3, 8, 13, 2, 7, 12, 1, 6, 11};
 MYALIGN const unsigned int _inv_shift_rows[] = {0x070a0d00, 0x0b0e0104, 0x0f020508, 0x0306090c};
 MYALIGN const unsigned int _mul2mask[] = {0x1b1b0000, 0x00000000, 0x00000000, 0x00000000};
 MYALIGN const unsigned int _mul4mask[] = {0x2d361b00, 0x00000000, 0x00000000, 0x00000000};
@@ -131,7 +131,7 @@ MYALIGN const unsigned int _IV512[] = {
    t1 = _mm_srli_epi16(t0, 6);\
    t1 = _mm_and_si128(t1, M128(_lsbmask2));\
    t3 = _mm_xor_si128(t3, _mm_shuffle_epi8(M128(_mul2mask), t1));\
-   t0  = _mm_xor_si128(t4, _mm_shuffle_epi8(M128(_mul4mask), t1))
+   t0 = _mm_xor_si128(t4, _mm_shuffle_epi8(M128(_mul4mask), t1))
 
 /*
 #define PRESUPERMIX(x, t1, s1, s2, t2)\

algo/fugue/fugue-aesni.h

@@ -20,7 +20,7 @@
 #error "Unsupported configuration, AES needs SSE4.1. Compile without AES."
 #endif
 
-#include "algo/sha/sha3_common.h"
+#include "compat/sha3_common.h"
 #include "simd-utils.h"
 
 

algo/fugue/sph_fugue.h

@@ -2,7 +2,7 @@
 #define SPH_FUGUE_H__
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 #ifdef __cplusplus
 extern "C"{

algo/gost/sph_gost.h

@@ -41,7 +41,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for GOST-256.

algo/groestl/aes_ni/groestl-intr-aes.h

@@ -139,7 +139,7 @@ static const __m128i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003 };
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m128_const1_64( 0x1b1b1b1b1b1b1b1b );\
+  b1 = _mm_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
   MUL2(a0, b0, b1);\
   a0 = _mm_xor_si128(a0, TEMP0);\
   MUL2(a1, b0, b1);\
@@ -237,7 +237,7 @@ static const __m128i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003 };
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m128_const1_64( 0x1b1b1b1b1b1b1b1b );\
+  b1 = _mm_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
   MUL2(a0, b0, b1);\
   a0 = _mm_xor_si128(a0, TEMP0);\
   MUL2(a1, b0, b1);\

algo/groestl/aes_ni/groestl256-intr-aes.h

@@ -128,7 +128,7 @@ static const __m128i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e };
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m128_const1_64( 0x1b1b1b1b1b1b1b1b );\
+  b1 = _mm_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
   MUL2(a0, b0, b1);\
   a0 = _mm_xor_si128(a0, TEMP0);\
   MUL2(a1, b0, b1);\
@@ -226,7 +226,7 @@ static const __m128i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e };
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m128_const1_64( 0x1b1b1b1b1b1b1b1b );\
+  b1 = _mm_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
   MUL2(a0, b0, b1);\
   a0 = _mm_xor_si128(a0, TEMP0);\
   MUL2(a1, b0, b1);\
@@ -275,7 +275,7 @@ static const __m128i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e };
  */
 #define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
   /* AddRoundConstant */\
-  b1 = m128_const_64( 0xffffffffffffffff, 0 ); \
+  b1 = _mm_set_epi64x( 0xffffffffffffffff, 0 ); \
   a0 = _mm_xor_si128( a0, casti_m128i( round_const_l0, i ) ); \
   a1 = _mm_xor_si128( a1, b1 ); \
   a2 = _mm_xor_si128( a2, b1 ); \

algo/groestl/aes_ni/hash-groestl.c

@@ -31,7 +31,7 @@ HashReturn_gr init_groestl( hashState_groestl* ctx, int hashlen )
   }
 
   // The only non-zero in the IV is len. It can be hard coded.
-  ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
+  ctx->chaining[ 6 ] = _mm_set_epi64x( 0x0200000000000000, 0 );
 
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;
@@ -48,7 +48,7 @@ HashReturn_gr reinit_groestl( hashState_groestl* ctx )
      ctx->chaining[i] = _mm_setzero_si128();
      ctx->buffer[i]   = _mm_setzero_si128();
   }
-  ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
+  ctx->chaining[ 6 ] = _mm_set_epi64x( 0x0200000000000000, 0 );
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;
 
@@ -116,7 +116,7 @@ HashReturn_gr final_groestl( hashState_groestl* ctx, void* output )
    else
    {
        // add first padding
-       ctx->buffer[rem_ptr] = m128_const_64( 0, 0x80 );
+       ctx->buffer[rem_ptr] = _mm_set_epi64x( 0, 0x80 );
        // add zero padding
        for ( i = rem_ptr + 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = _mm_setzero_si128();
@@ -148,7 +148,7 @@ int groestl512_full( hashState_groestl* ctx, void* output,
       ctx->chaining[i] = _mm_setzero_si128();
       ctx->buffer[i]   = _mm_setzero_si128();
    }
-   ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
+   ctx->chaining[ 6 ] = _mm_set_epi64x( 0x0200000000000000, 0 );
    ctx->buf_ptr = 0;
 
    // --- update ---
@@ -182,7 +182,7 @@ int groestl512_full( hashState_groestl* ctx, void* output,
    else
    {
        // add first padding
-       ctx->buffer[i] = m128_const_64( 0, 0x80 );
+       ctx->buffer[i] = _mm_set_epi64x( 0, 0x80 );
        // add zero padding
        for ( i += 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = _mm_setzero_si128();
@@ -239,7 +239,7 @@ HashReturn_gr update_and_final_groestl( hashState_groestl* ctx, void* output,
    else
    {
        // add first padding
-       ctx->buffer[i] = m128_const_64( 0, 0x80 );
+       ctx->buffer[i] = _mm_set_epi64x( 0, 0x80 );
        // add zero padding
        for ( i += 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = _mm_setzero_si128();

algo/groestl/aes_ni/hash-groestl.h

@@ -20,8 +20,8 @@
 #define LENGTH (512)
 
 #include "brg_endian.h"
-#define NEED_UINT_64T
+//#define NEED_UINT_64T
-#include "algo/sha/brg_types.h"
+#include "compat/brg_types.h"
 
 /* some sizes (number of bytes) */
 #define ROWS (8)

algo/groestl/aes_ni/hash-groestl256.c

@@ -46,7 +46,7 @@ HashReturn_gr reinit_groestl256(hashState_groestl256* ctx)
      ctx->buffer[i]   = _mm_setzero_si128();
   }
 
-  ctx->chaining[ 3 ] = m128_const_64( 0, 0x0100000000000000 );
+  ctx->chaining[ 3 ] = _mm_set_epi64x( 0, 0x0100000000000000 );
 
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;

algo/groestl/aes_ni/hash-groestl256.h

@@ -34,8 +34,7 @@ typedef crypto_uint64 u64;
 //#define LENGTH (512)
 
 #include "brg_endian.h"
-#define NEED_UINT_64T
+#include "compat/brg_types.h"
-#include "algo/sha/brg_types.h"
 
 #ifdef IACA_TRACE
   #include IACA_MARKS

algo/groestl/groestl-gate.c

@@ -17,7 +17,7 @@ bool register_dmd_gr_algo( algo_gate_t *gate )
 bool register_groestl_algo( algo_gate_t* gate )
 {
     register_dmd_gr_algo( gate );
-    gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
+    gate->gen_merkle_root = (void*)&sha256_gen_merkle_root;
     return true;
 };
 

algo/groestl/groestl256-hash-4way.c

@@ -33,8 +33,7 @@ int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )
   }
 
   // The only non-zero in the IV is len. It can be hard coded.
-  ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
+  ctx->chaining[ 3 ] = mm512_bcast128lo_64( 0x0100000000000000 );
-
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;
 
@@ -51,9 +50,6 @@ int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
    __m512i* in = (__m512i*)input;
    int i;
 
-//  if (ctx->chaining == NULL || ctx->buffer == NULL)
-//    return 1;
-
   for ( i = 0; i < SIZE256; i++ )
   {
      ctx->chaining[i] = m512_zero;
@@ -61,7 +57,7 @@ int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
   }
 
   // The only non-zero in the IV is len. It can be hard coded.
-  ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
+  ctx->chaining[ 3 ] = mm512_bcast128lo_64( 0x0100000000000000 );
   ctx->buf_ptr = 0;
    
    // --- update ---
@@ -83,18 +79,18 @@ int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
    if ( i == SIZE256 - 1 )
    {        
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 ); 
+       ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 ); 
    }   
    else
    {
        // add first padding
-       ctx->buffer[i] = m512_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
        // add zero padding
        for ( i += 1; i < SIZE256 - 1; i++ )
            ctx->buffer[i] = m512_zero;
 
        // add length padding, second last byte is zero unless blocks > 255
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
    }
 
    // digest final padding block and do output transform
@@ -140,18 +136,18 @@ int groestl256_4way_update_close( groestl256_4way_context* ctx, void* output,
    if ( i == SIZE256 - 1 )
    {
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
+       ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
    }
    else
    {
        // add first padding
-       ctx->buffer[i] = m512_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
        // add zero padding
        for ( i += 1; i < SIZE256 - 1; i++ )
            ctx->buffer[i] = m512_zero;
 
        // add length padding, second last byte is zero unless blocks > 255
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
    }
 
 // digest final padding block and do output transform
@@ -186,7 +182,7 @@ int groestl256_2way_init( groestl256_2way_context* ctx, uint64_t hashlen )
   }
 
   // The only non-zero in the IV is len. It can be hard coded.
-  ctx->chaining[ 3 ] = m256_const2_64( 0, 0x0100000000000000 );
+  ctx->chaining[ 3 ] = mm256_bcast128lo_64( 0x0100000000000000 );
 
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;
@@ -211,7 +207,7 @@ int groestl256_2way_full( groestl256_2way_context* ctx, void* output,
    }
 
    // The only non-zero in the IV is len. It can be hard coded.
-   ctx->chaining[ 3 ] = m256_const2_64( 0, 0x0100000000000000 );
+   ctx->chaining[ 3 ] = mm256_bcast128lo_64( 0x0100000000000000 );
    ctx->buf_ptr = 0;
 
    // --- update ---
@@ -233,18 +229,18 @@ int groestl256_2way_full( groestl256_2way_context* ctx, void* output,
    if ( i == SIZE256 - 1 )
    {
        // only 1 vector left in buffer, all padding at once
-      ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
+      ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
    }
    else
    {
        // add first padding
-       ctx->buffer[i] = m256_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
        // add zero padding
        for ( i += 1; i < SIZE256 - 1; i++ )
            ctx->buffer[i] = m256_zero;
 
        // add length padding, second last byte is zero unless blocks > 255
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
    }
 
    // digest final padding block and do output transform
@@ -289,23 +285,22 @@ int groestl256_2way_update_close( groestl256_2way_context* ctx, void* output,
    if ( i == SIZE256 - 1 )
    {
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
+       ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
    }
    else
    {
        // add first padding
-       ctx->buffer[i] = m256_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
        // add zero padding
        for ( i += 1; i < SIZE256 - 1; i++ )
            ctx->buffer[i] = m256_zero;
 
        // add length padding, second last byte is zero unless blocks > 255
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
    }
 
 // digest final padding block and do output transform
    TF512_2way( ctx->chaining, ctx->buffer );
-
    OF512_2way( ctx->chaining );
 
    // store hash result in output 

algo/groestl/groestl256-hash-4way.h

@@ -22,10 +22,6 @@
 
 #define LENGTH (256)
 
-//#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)

algo/groestl/groestl256-intr-4way.h

@@ -165,7 +165,7 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b ); \
+  b1 = _mm512_set1_epi64( 0x1b1b1b1b1b1b1b1b ); \
   MUL2( a0, b0, b1 ); \
   a0 = _mm512_xor_si512( a0, TEMP0 ); \
   MUL2( a1, b0, b1 ); \
@@ -205,116 +205,18 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
   b1 = _mm512_xor_si512( b1, a4 ); \
 }/*MixBytes*/
 
-
+#define MASK_NOT( a )  _mm512_mask_ternarylogic_epi64( a, 0xaa, a, a, 1 )
-#if 0
-#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 = m512_const1_64( 0x1b1b1b1b1b1b1b1b );\
-  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*/
-#endif
 
 #define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
   /* AddRoundConstant */\
-  b1 = m512_const2_64( 0xffffffffffffffff, 0 ); \
+  a0 = _mm512_xor_si512( a0, mm512_bcast_m128( round_const_l0[i] ) );\
-  a0 = _mm512_xor_si512( a0, m512_const1_128( round_const_l0[i] ) );\
+  a1 = MASK_NOT( a1 ); \
-  a1 = _mm512_xor_si512( a1, b1 );\
+  a2 = MASK_NOT( a2 ); \
-  a2 = _mm512_xor_si512( a2, b1 );\
+  a3 = MASK_NOT( a3 ); \
-  a3 = _mm512_xor_si512( a3, b1 );\
+  a4 = MASK_NOT( a4 ); \
-  a4 = _mm512_xor_si512( a4, b1 );\
+  a5 = MASK_NOT( a5 ); \
-  a5 = _mm512_xor_si512( a5, b1 );\
+  a6 = MASK_NOT( a6 ); \
-  a6 = _mm512_xor_si512( a6, b1 );\
+  a7 = _mm512_xor_si512( a7, mm512_bcast_m128( round_const_l7[i] ) );\
-  a7 = _mm512_xor_si512( a7, m512_const1_128( round_const_l7[i] ) );\
   \
   /* ShiftBytes + SubBytes (interleaved) */\
   b0 = _mm512_xor_si512( b0, b0 );\
@@ -450,7 +352,7 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
  * outputs: (i0-7) = (0|S)
  */
 #define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
-  t0 = _mm512_xor_si512( t0, t0 );\
+  t0 = m512_zero;\
   i1 = i0;\
   i3 = i2;\
   i5 = i4;\
@@ -481,11 +383,11 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
 
 void TF512_4way( __m512i* chaining, __m512i* message )
 {
-  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m512i TEMP0;
+  __m512i TEMP0;
-  static __m512i TEMP1;
+  __m512i TEMP1;
-  static __m512i TEMP2;
+  __m512i TEMP2;
 
   /* load message into registers xmm12 - xmm15 */
   xmm12 = message[0];
@@ -547,11 +449,11 @@ void TF512_4way( __m512i* chaining, __m512i* message )
 
 void OF512_4way( __m512i* chaining )
 {
-  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m512i TEMP0;
+  __m512i TEMP0;
-  static __m512i TEMP1;
+  __m512i TEMP1;
-  static __m512i TEMP2;
+  __m512i TEMP2;
 
   /* load CV into registers xmm8, xmm10, xmm12, xmm14 */
   xmm8 = chaining[0];
@@ -637,7 +539,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   j = _mm256_cmpgt_epi8(j, i );\
   i = _mm256_add_epi8(i, i);\
   j = _mm256_and_si256(j, k);\
-  i = _mm256_xor_si256(i, j);\
+  i = mm256_xorand( i, j, k );\
 }
 
 #define MixBytes_2way(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
@@ -648,7 +550,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   b0 = a2;\
   a1 = _mm256_xor_si256(a1, a2);\
   b1 = a3;\
-  a2 = _mm256_xor_si256(a2, a3);\
+  TEMP2 = _mm256_xor_si256(a2, a3);\
   b2 = a4;\
   a3 = _mm256_xor_si256(a3, a4);\
   b3 = a5;\
@@ -660,34 +562,20 @@ static const __m256i SUBSH_MASK7_2WAY =
   a7 = _mm256_xor_si256(a7, b6);\
   \
   /* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
-  b0 = _mm256_xor_si256(b0, a4);\
+  TEMP0 = mm256_xor3( b0, a4, a6 ); \
-  b6 = _mm256_xor_si256(b6, a4);\
+  TEMP1 = mm256_xor3( b1, a5, a7 ); \
-  b1 = _mm256_xor_si256(b1, a5);\
+  b2 = mm256_xor3( b2, a6, a0 ); \
-  b7 = _mm256_xor_si256(b7, a5);\
+  b0 = a0; \
-  b2 = _mm256_xor_si256(b2, a6);\
+  b3 = mm256_xor3( b3, a7, a1 ); \
-  b0 = _mm256_xor_si256(b0, a6);\
+  b1 = a1; \
-  /* spill values y_4, y_5 to memory */\
+  b6 = mm256_xor3( b6, a4, TEMP2 ); \
-  TEMP0 = b0;\
+  b4 = mm256_xor3( b4, a0, TEMP2 ); \
-  b3 = _mm256_xor_si256(b3, a7);\
+  b7 = mm256_xor3( b7, a5, a3 ); \
-  b1 = _mm256_xor_si256(b1, a7);\
+  b5 = mm256_xor3( b5, a1, a3 ); \
-  TEMP1 = b1;\
-  b4 = _mm256_xor_si256(b4, a0);\
-  b2 = _mm256_xor_si256(b2, a0);\
-  /* save values t0, t1, t2 to xmm8, xmm9 and memory */\
-  b0 = a0;\
-  b5 = _mm256_xor_si256(b5, a1);\
-  b3 = _mm256_xor_si256(b3, a1);\
-  b1 = a1;\
-  b6 = _mm256_xor_si256(b6, a2);\
-  b4 = _mm256_xor_si256(b4, a2);\
-  TEMP2 = a2;\
-  b7 = _mm256_xor_si256(b7, a3);\
-  b5 = _mm256_xor_si256(b5, a3);\
-  \
   /* compute x_i = t_i + t_{i+3} */\
   a0 = _mm256_xor_si256(a0, a3);\
   a1 = _mm256_xor_si256(a1, a4);\
-  a2 = _mm256_xor_si256(a2, a5);\
+  a2 = _mm256_xor_si256( TEMP2, a5);\
   a3 = _mm256_xor_si256(a3, a6);\
   a4 = _mm256_xor_si256(a4, a7);\
   a5 = _mm256_xor_si256(a5, b0);\
@@ -696,7 +584,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m256_const1_64( 0x1b1b1b1b1b1b1b1b );\
+  b1 = _mm256_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
   MUL2_2WAY(a0, b0, b1);\
   a0 = _mm256_xor_si256(a0, TEMP0);\
   MUL2_2WAY(a1, b0, b1);\
@@ -738,15 +626,15 @@ static const __m256i SUBSH_MASK7_2WAY =
 
 #define ROUND_2WAY(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
   /* AddRoundConstant */\
-  b1 = m256_const2_64( 0xffffffffffffffff, 0 ); \
+  b1 = mm256_bcast_m128( mm128_mask_32( m128_neg1, 0x3 ) ); \
-  a0 = _mm256_xor_si256( a0, m256_const1_128( round_const_l0[i] ) );\
+  a0 = _mm256_xor_si256( a0, mm256_bcast_m128( round_const_l0[i] ) );\
   a1 = _mm256_xor_si256( a1, b1 );\
   a2 = _mm256_xor_si256( a2, b1 );\
   a3 = _mm256_xor_si256( a3, b1 );\
   a4 = _mm256_xor_si256( a4, b1 );\
   a5 = _mm256_xor_si256( a5, b1 );\
   a6 = _mm256_xor_si256( a6, b1 );\
-  a7 = _mm256_xor_si256( a7, m256_const1_128( round_const_l7[i] ) );\
+  a7 = _mm256_xor_si256( a7, mm256_bcast_m128( round_const_l7[i] ) );\
   \
   /* ShiftBytes + SubBytes (interleaved) */\
   b0 = _mm256_xor_si256( b0, b0 );\
@@ -769,7 +657,6 @@ static const __m256i SUBSH_MASK7_2WAY =
   \
   /* MixBytes */\
   MixBytes_2way(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
-\
 }
 
 /* 10 rounds, P and Q in parallel */
@@ -850,7 +737,7 @@ static const __m256i SUBSH_MASK7_2WAY =
 }/**/
 
 #define Matrix_Transpose_O_B_2way(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
-  t0 = _mm256_xor_si256( t0, t0 );\
+  t0 = m256_zero;\
   i1 = i0;\
   i3 = i2;\
   i5 = i4;\
@@ -874,11 +761,11 @@ static const __m256i SUBSH_MASK7_2WAY =
 
 void TF512_2way( __m256i* chaining, __m256i* message )
 {
-  static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m256i TEMP0;
+  __m256i TEMP0;
-  static __m256i TEMP1;
+  __m256i TEMP1;
-  static __m256i TEMP2;
+  __m256i TEMP2;
 
   /* load message into registers xmm12 - xmm15 */
   xmm12 = message[0];
@@ -940,11 +827,11 @@ void TF512_2way( __m256i* chaining, __m256i* message )
   
 void OF512_2way( __m256i* chaining )
 {
-  static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m256i TEMP0;
+  __m256i TEMP0;
-  static __m256i TEMP1;
+  __m256i TEMP1;
-  static __m256i TEMP2;
+  __m256i TEMP2;
 
   /* load CV into registers xmm8, xmm10, xmm12, xmm14 */
   xmm8 = chaining[0];

algo/groestl/groestl512-hash-4way.c

@@ -25,8 +25,7 @@ int groestl512_4way_init( groestl512_4way_context* ctx, uint64_t hashlen )
   memset_zero_512( ctx->buffer, SIZE512 );
 
   // The only non-zero in the IV is len. It can be hard coded.
-  ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );
+  ctx->chaining[ 6 ] = mm512_bcast128hi_64( 0x0200000000000000 );
-
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;
 
@@ -61,14 +60,14 @@ int groestl512_4way_update_close( groestl512_4way_context* ctx, void* output,
    if ( i == SIZE512 - 1 )
    {        
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
+       ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
    }   
    else
    {
-       ctx->buffer[i] = m512_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
        for ( i += 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = m512_zero;
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
    }
 
    TF1024_4way( ctx->chaining, ctx->buffer );
@@ -94,7 +93,7 @@ int groestl512_4way_full( groestl512_4way_context* ctx, void* output,
 
    memset_zero_512( ctx->chaining, SIZE512 );
    memset_zero_512( ctx->buffer, SIZE512 );
-   ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );
+   ctx->chaining[ 6 ] = mm512_bcast128hi_64( 0x0200000000000000 );
    ctx->buf_ptr = 0;
 
    // --- update ---
@@ -113,14 +112,14 @@ int groestl512_4way_full( groestl512_4way_context* ctx, void* output,
    if ( i == SIZE512 - 1 )
    {
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
+       ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
    }
    else
    {
-       ctx->buffer[i] = m512_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
        for ( i += 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = m512_zero;
-       ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
    }
 
    TF1024_4way( ctx->chaining, ctx->buffer );
@@ -143,7 +142,7 @@ int groestl512_2way_init( groestl512_2way_context* ctx, uint64_t hashlen )
   memset_zero_256( ctx->buffer, SIZE512 );
 
   // The only non-zero in the IV is len. It can be hard coded.
-  ctx->chaining[ 6 ] = m256_const2_64( 0x0200000000000000, 0 );
+  ctx->chaining[ 6 ] = mm256_bcast128hi_64( 0x0200000000000000 );
 
   ctx->buf_ptr = 0;
   ctx->rem_ptr = 0;
@@ -179,14 +178,14 @@ int groestl512_2way_update_close( groestl512_2way_context* ctx, void* output,
    if ( i == SIZE512 - 1 )
    {
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
+       ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
    }
    else
    {
-       ctx->buffer[i] = m256_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
        for ( i += 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = m256_zero;
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
    }
 
    TF1024_2way( ctx->chaining, ctx->buffer );
@@ -212,7 +211,7 @@ int groestl512_2way_full( groestl512_2way_context* ctx, void* output,
 
    memset_zero_256( ctx->chaining, SIZE512 );
    memset_zero_256( ctx->buffer, SIZE512 );
-   ctx->chaining[ 6 ] = m256_const2_64( 0x0200000000000000, 0 );
+   ctx->chaining[ 6 ] = mm256_bcast128hi_64( 0x0200000000000000 );
    ctx->buf_ptr = 0;
 
    // --- update ---
@@ -231,14 +230,14 @@ int groestl512_2way_full( groestl512_2way_context* ctx, void* output,
    if ( i == SIZE512 - 1 )
    {
        // only 1 vector left in buffer, all padding at once
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
+       ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
    }
    else
    {
-       ctx->buffer[i] = m256_const2_64( 0, 0x80 );
+       ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
        for ( i += 1; i < SIZE512 - 1; i++ )
            ctx->buffer[i] = m256_zero;
-       ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
+       ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
    }
 
    TF1024_2way( ctx->chaining, ctx->buffer );

algo/groestl/groestl512-intr-4way.h

@@ -174,7 +174,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b ); \
+  b1 = _mm512_set1_epi64( 0x1b1b1b1b1b1b1b1b ); \
   MUL2( a0, b0, b1 ); \
   a0 = _mm512_xor_si512( a0, TEMP0 ); \
   MUL2( a1, b0, b1 ); \
@@ -238,7 +238,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
   for ( round_counter = 0; round_counter < 14; round_counter += 2 ) \
   { \
     /* AddRoundConstant P1024 */\
-    xmm8 = _mm512_xor_si512( xmm8, m512_const1_128( \
+    xmm8 = _mm512_xor_si512( xmm8, mm512_bcast_m128( \
              casti_m128i( round_const_p, round_counter ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm8  = _mm512_shuffle_epi8( xmm8,  SUBSH_MASK0 ); \
@@ -253,7 +253,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
     SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
     \
      /* AddRoundConstant P1024 */\
-    xmm0 = _mm512_xor_si512( xmm0, m512_const1_128( \
+    xmm0 = _mm512_xor_si512( xmm0, mm512_bcast_m128( \
              casti_m128i( round_const_p, round_counter+1 ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK0 );\
@@ -282,7 +282,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
     xmm12 = _mm512_xor_si512( xmm12, xmm1 );\
     xmm13 = _mm512_xor_si512( xmm13, xmm1 );\
     xmm14 = _mm512_xor_si512( xmm14, xmm1 );\
-    xmm15 = _mm512_xor_si512( xmm15, m512_const1_128( \
+    xmm15 = _mm512_xor_si512( xmm15, mm512_bcast_m128( \
                  casti_m128i( round_const_q, round_counter ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm8  = _mm512_shuffle_epi8( xmm8,  SUBSH_MASK1 );\
@@ -305,7 +305,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
     xmm4 = _mm512_xor_si512( xmm4, xmm9 );\
     xmm5 = _mm512_xor_si512( xmm5, xmm9 );\
     xmm6 = _mm512_xor_si512( xmm6, xmm9 );\
-    xmm7 = _mm512_xor_si512( xmm7, m512_const1_128( \
+    xmm7 = _mm512_xor_si512( xmm7, mm512_bcast_m128( \
              casti_m128i( round_const_q, round_counter+1 ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK1 );\
@@ -471,8 +471,8 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
 
 void INIT_4way( __m512i* chaining )
 {
-  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
 
   /* load IV into registers xmm8 - xmm15 */
   xmm8 = chaining[0];
@@ -500,12 +500,12 @@ void INIT_4way( __m512i* chaining )
 
 void TF1024_4way( __m512i* chaining, const __m512i* message )
 {
-  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m512i QTEMP[8];
+  __m512i QTEMP[8];
-  static __m512i TEMP0;
+  __m512i TEMP0;
-  static __m512i TEMP1;
+  __m512i TEMP1;
-  static __m512i TEMP2;
+  __m512i TEMP2;
 
   /* load message into registers xmm8 - xmm15 (Q = message) */
   xmm8 = message[0];
@@ -606,11 +606,11 @@ void TF1024_4way( __m512i* chaining, const __m512i* message )
 
 void OF1024_4way( __m512i* chaining )
 {
-  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m512i TEMP0;
+  __m512i TEMP0;
-  static __m512i TEMP1;
+  __m512i TEMP1;
-  static __m512i TEMP2;
+  __m512i TEMP2;
 
   /* load CV into registers xmm8 - xmm15 */
   xmm8 = chaining[0];
@@ -710,7 +710,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   b0 = a2;\
   a1 = _mm256_xor_si256(a1, a2);\
   b1 = a3;\
-  a2 = _mm256_xor_si256(a2, a3);\
+  TEMP2 = _mm256_xor_si256(a2, a3);\
   b2 = a4;\
   a3 = _mm256_xor_si256(a3, a4);\
   b3 = a5;\
@@ -722,34 +722,23 @@ static const __m256i SUBSH_MASK7_2WAY =
   a7 = _mm256_xor_si256(a7, b6);\
   \
   /* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
-  b0 = _mm256_xor_si256(b0, a4);\
+  TEMP0 = mm256_xor3( b0, a4, a6 ); \
-  b6 = _mm256_xor_si256(b6, a4);\
-  b1 = _mm256_xor_si256(b1, a5);\
-  b7 = _mm256_xor_si256(b7, a5);\
-  b2 = _mm256_xor_si256(b2, a6);\
-  b0 = _mm256_xor_si256(b0, a6);\
   /* spill values y_4, y_5 to memory */\
-  TEMP0 = b0;\
+  TEMP1 = mm256_xor3( b1, a5, a7 ); \
-  b3 = _mm256_xor_si256(b3, a7);\
+  b2 = mm256_xor3( b2, a6, a0 ); \
-  b1 = _mm256_xor_si256(b1, a7);\
-  TEMP1 = b1;\
-  b4 = _mm256_xor_si256(b4, a0);\
-  b2 = _mm256_xor_si256(b2, a0);\
   /* save values t0, t1, t2 to xmm8, xmm9 and memory */\
-  b0 = a0;\
+  b0 = a0; \
-  b5 = _mm256_xor_si256(b5, a1);\
+  b3 = mm256_xor3( b3, a7, a1 ); \
-  b3 = _mm256_xor_si256(b3, a1);\
+  b1 = a1; \
-  b1 = a1;\
+  b6 = mm256_xor3( b6, a4, TEMP2 ); \
-  b6 = _mm256_xor_si256(b6, a2);\
+  b4 = mm256_xor3( b4, a0, TEMP2 ); \
-  b4 = _mm256_xor_si256(b4, a2);\
+  b7 = mm256_xor3( b7, a5, a3 ); \
-  TEMP2 = a2;\
+  b5 = mm256_xor3( b5, a1, a3 ); \
-  b7 = _mm256_xor_si256(b7, a3);\
-  b5 = _mm256_xor_si256(b5, a3);\
   \
   /* compute x_i = t_i + t_{i+3} */\
   a0 = _mm256_xor_si256(a0, a3);\
   a1 = _mm256_xor_si256(a1, a4);\
-  a2 = _mm256_xor_si256(a2, a5);\
+  a2 = _mm256_xor_si256( TEMP2, a5);\
   a3 = _mm256_xor_si256(a3, a6);\
   a4 = _mm256_xor_si256(a4, a7);\
   a5 = _mm256_xor_si256(a5, b0);\
@@ -758,7 +747,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   \
   /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
   /* compute w_i : add y_{i+4} */\
-  b1 = m256_const1_64( 0x1b1b1b1b1b1b1b1b );\
+  b1 = _mm256_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
   MUL2_2WAY(a0, b0, b1);\
   a0 = _mm256_xor_si256(a0, TEMP0);\
   MUL2_2WAY(a1, b0, b1);\
@@ -822,7 +811,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   for ( round_counter = 0; round_counter < 14; round_counter += 2 ) \
   { \
     /* AddRoundConstant P1024 */\
-    xmm8 = _mm256_xor_si256( xmm8, m256_const1_128( \
+    xmm8 = _mm256_xor_si256( xmm8, mm256_bcast_m128( \
              casti_m128i( round_const_p, round_counter ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm8  = _mm256_shuffle_epi8( xmm8,  SUBSH_MASK0_2WAY ); \
@@ -837,7 +826,7 @@ static const __m256i SUBSH_MASK7_2WAY =
     SUBMIX_2WAY(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
     \
      /* AddRoundConstant P1024 */\
-    xmm0 = _mm256_xor_si256( xmm0, m256_const1_128( \
+    xmm0 = _mm256_xor_si256( xmm0, mm256_bcast_m128( \
              casti_m128i( round_const_p, round_counter+1 ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm0 = _mm256_shuffle_epi8( xmm0, SUBSH_MASK0_2WAY );\
@@ -866,7 +855,7 @@ static const __m256i SUBSH_MASK7_2WAY =
     xmm12 = _mm256_xor_si256( xmm12, xmm1 );\
     xmm13 = _mm256_xor_si256( xmm13, xmm1 );\
     xmm14 = _mm256_xor_si256( xmm14, xmm1 );\
-    xmm15 = _mm256_xor_si256( xmm15, m256_const1_128( \
+    xmm15 = _mm256_xor_si256( xmm15, mm256_bcast_m128( \
                  casti_m128i( round_const_q, round_counter ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm8  = _mm256_shuffle_epi8( xmm8,  SUBSH_MASK1_2WAY );\
@@ -889,7 +878,7 @@ static const __m256i SUBSH_MASK7_2WAY =
     xmm4 = _mm256_xor_si256( xmm4, xmm9 );\
     xmm5 = _mm256_xor_si256( xmm5, xmm9 );\
     xmm6 = _mm256_xor_si256( xmm6, xmm9 );\
-    xmm7 = _mm256_xor_si256( xmm7, m256_const1_128( \
+    xmm7 = _mm256_xor_si256( xmm7, mm256_bcast_m128( \
              casti_m128i( round_const_q, round_counter+1 ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm0 = _mm256_shuffle_epi8( xmm0, SUBSH_MASK1_2WAY );\
@@ -1040,8 +1029,8 @@ static const __m256i SUBSH_MASK7_2WAY =
 
 void INIT_2way( __m256i *chaining )
 {
-  static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
 
   /* load IV into registers xmm8 - xmm15 */
   xmm8 = chaining[0];
@@ -1069,12 +1058,12 @@ void INIT_2way( __m256i *chaining )
 
 void TF1024_2way( __m256i *chaining, const __m256i *message )
 {
-  static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m256i QTEMP[8];
+  __m256i QTEMP[8];
-  static __m256i TEMP0;
+  __m256i TEMP0;
-  static __m256i TEMP1;
+  __m256i TEMP1;
-  static __m256i TEMP2;
+  __m256i TEMP2;
 
   /* load message into registers xmm8 - xmm15 (Q = message) */
   xmm8 = message[0];
@@ -1175,11 +1164,11 @@ void TF1024_2way( __m256i *chaining, const __m256i *message )
 
 void OF1024_2way( __m256i* chaining )
 {
-  static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m256i TEMP0;
+  __m256i TEMP0;
-  static __m256i TEMP1;
+  __m256i TEMP1;
-  static __m256i TEMP2;
+  __m256i TEMP2;
 
   /* load CV into registers xmm8 - xmm15 */
   xmm8 = chaining[0];

algo/groestl/myrgr-4way.c

@@ -4,7 +4,7 @@
 #include <stdint.h>
 #include <string.h>
 #include "aes_ni/hash-groestl.h"
-#include "algo/sha/sha-hash-4way.h"
+#include "algo/sha/sha256-hash.h"
 #if defined(__VAES__)
   #include "groestl512-hash-4way.h"
 #endif

algo/groestl/sph_groestl.h

@@ -40,7 +40,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 #if !defined(__AES__)   
 /**

algo/hamsi/hamsi-hash-4way.h

@@ -36,44 +36,64 @@
 #define HAMSI_4WAY_H__
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
 
 #if defined (__AVX2__)
 
 #include "simd-utils.h"
 
-#ifdef __cplusplus
+// Hamsi-512 4x64
-extern "C"{
-#endif
-
-#define SPH_SIZE_hamsi512   512
 
 // Partial is only scalar but needs pointer ref for hamsi-helper
 // deprecate partial_len
-typedef struct {
+typedef struct
+{
    __m256i h[8];
    __m256i buf[1];
    size_t partial_len;
-   sph_u32 count_high, count_low;
+   uint32_t count_high, count_low;
 } hamsi_4way_big_context;
-
 typedef hamsi_4way_big_context hamsi512_4way_context;
 
 void hamsi512_4way_init( hamsi512_4way_context *sc );
 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 );
 
+#define hamsi512_4x64_context   hamsi512_4way_context
+#define hamsi512_4x64_init      hamsi512_4way_init
+#define hamsi512_4x64_update    hamsi512_4way_update
+#define hamsi512_4x64_close     hamsi512_4way_close
+
+// Hamsi-512 8x32
+
+typedef struct
+{
+   __m256i h[16];
+   __m256i buf[2];
+   size_t partial_len;
+   uint32_t count_high, count_low;
+} hamsi_8x32_big_context;
+typedef hamsi_8x32_big_context hamsi512_8x32_context;
+
+void hamsi512_8x32_init( hamsi512_8x32_context *sc );
+void hamsi512_8x32_update( hamsi512_8x32_context *sc, const void *data,
+      size_t len );
+void hamsi512_8x32_close( hamsi512_8x32_context *sc, void *dst );
+void hamsi512_8x32_full( hamsi512_8x32_context *sc, void *dst, const void *data,
+      size_t len );
+
+#endif
+
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
+// Hamsi-512 8x64
+
 typedef struct {
    __m512i h[8];
    __m512i buf[1];
    size_t partial_len;
-   sph_u32 count_high, count_low;
+   uint32_t count_high, count_low;
 } hamsi_8way_big_context;
-
 typedef hamsi_8way_big_context hamsi512_8way_context;
 
 void hamsi512_8way_init( hamsi512_8way_context *sc );
@@ -81,15 +101,29 @@ void hamsi512_8way_update( hamsi512_8way_context *sc, const void *data,
                            size_t len );
 void hamsi512_8way_close( hamsi512_8way_context *sc, void *dst );
 
+#define hamsi512_8x64_context   hamsi512_8way_context
+#define hamsi512_8x64_init      hamsi512_8way_init
+#define hamsi512_8x64_update    hamsi512_8way_update
+#define hamsi512_8x64_close     hamsi512_8way_close
 
-
+// Hamsi-512 16x32
-#endif
+
-
+typedef struct
-
+{
-#ifdef __cplusplus
+   __m512i h[16];
-}
+   __m512i buf[2];
-#endif
+   size_t partial_len;
-
+   uint32_t count_high, count_low;
-#endif
+} hamsi_16x32_big_context;
+typedef hamsi_16x32_big_context hamsi512_16x32_context;
+
+void hamsi512_16x32_init( hamsi512_16x32_context *sc );
+void hamsi512_16x32_update( hamsi512_16x32_context *sc, const void *data,
+                           size_t len );
+void hamsi512_16way_close( hamsi512_16x32_context *sc, void *dst );
+void hamsi512_16x32_full( hamsi512_16x32_context *sc, void *dst,
+                          const void *data, size_t len );
+
+#endif   // AVX512
 
 #endif

algo/hamsi/sph_hamsi.h

@@ -36,7 +36,7 @@
 #define SPH_HAMSI_H__
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 #ifdef __cplusplus
 extern "C"{

algo/haval/haval-16way-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), _16way_update)
+( haval_16way_context *sc, const void *data, size_t len )
+{
+   __m512i *vdata = (__m512i*)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_512( sc->buf + (current>>2), vdata, clen>>2 );
+      vdata += clen>>2;
+      current += clen;
+      len -= clen;
+      if ( current == 128U )
+      {
+         DSTATE_16W;
+         IN_PREPARE_16W(sc->buf);
+         RSTATE_16W;
+         SPH_XCAT(CORE_16W, PASSES)(INW_16W);
+         WSTATE_16W;
+         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), _16way_close)( haval_16way_context *sc,
+                                                void *dst)
+{
+   unsigned current;
+   DSTATE_16W;
+
+   current = (unsigned)sc->count_low & 127UL;
+
+   sc->buf[ current>>2 ] = v512_32( 1 );
+   current += 4;   
+   RSTATE_16W;
+   if ( current > 116UL )
+   {
+      memset_zero_512( sc->buf + ( current>>2 ), (128UL-current) >> 2 );
+      do
+      {
+         IN_PREPARE_16W(sc->buf);
+         SPH_XCAT(CORE_16W, PASSES)(INW_16W);
+      } while (0);
+      current = 0;
+   }
+
+   uint32_t t1, t2;
+   memset_zero_512( sc->buf + ( current>>2 ), (116UL-current) >> 2 );
+   t1 = 0x01 | (PASSES << 3);
+   t2 = sc->olen << 3;
+   sc->buf[ 116>>2 ] = v512_32( ( t1 << 16 ) | ( t2 << 24 ) );
+   sc->buf[ 120>>2 ] = v512_32( sc->count_low << 3 );
+   sc->buf[ 124>>2 ] = v512_32( (sc->count_high << 3)
+                                     | (sc->count_low >> 29) );
+   do
+   {
+      IN_PREPARE_16W(sc->buf);
+      SPH_XCAT(CORE_16W, PASSES)(INW_16W);
+   } while (0);
+   WSTATE_16W;
+   haval_16way_out( sc, dst );
+}

algo/haval/haval-4way-helper.c

@@ -48,7 +48,7 @@ SPH_XCAT(SPH_XCAT(haval, PASSES), _4way_update)
    while ( len > 0 )
    {
       unsigned clen;
-      sph_u32 clow, clow2;
+      uint32_t clow, clow2;
 
       clen = 128U - current;
       if ( clen > len )
@@ -67,7 +67,7 @@ SPH_XCAT(SPH_XCAT(haval, PASSES), _4way_update)
          current = 0;
       }
       clow = sc->count_low;
-      clow2 = SPH_T32(clow + clen);
+      clow2 = clow + clen;
       sc->count_low = clow2;
       if ( clow2 < clow )
          sc->count_high ++;
@@ -83,7 +83,7 @@ SPH_XCAT(SPH_XCAT(haval, PASSES), _4way_close)( haval_4way_context *sc,
 
    current = (unsigned)sc->count_low & 127UL;
 
-   sc->buf[ current>>2 ] = m128_one_32;
+   sc->buf[ current>>2 ] = v128_32( 1 );
    current += 4;   
    RSTATE;
    if ( current > 116UL )

algo/haval/haval-8way-helper.c

@@ -83,7 +83,7 @@ SPH_XCAT(SPH_XCAT(haval, PASSES), _8way_close)( haval_8way_context *sc,
 
    current = (unsigned)sc->count_low & 127UL;
 
-   sc->buf[ current>>2 ] = m256_one_32;
+   sc->buf[ current>>2 ] = v256_32( 1 );
    current += 4;   
    RSTATE_8W;
    if ( current > 116UL )
@@ -101,9 +101,9 @@ SPH_XCAT(SPH_XCAT(haval, PASSES), _8way_close)( haval_8way_context *sc,
    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[ 116>>2 ] = v256_32( ( t1 << 16 ) | ( t2 << 24 ) );
-   sc->buf[ 120>>2 ] = _mm256_set1_epi32( sc->count_low << 3 );
+   sc->buf[ 120>>2 ] = v256_32( sc->count_low << 3 );
-   sc->buf[ 124>>2 ] = _mm256_set1_epi32( (sc->count_high << 3)
+   sc->buf[ 124>>2 ] = v256_32( (sc->count_high << 3)
                                      | (sc->count_low >> 29) );
    do
    {

algo/haval/haval-hash-4way.c

@@ -52,6 +52,56 @@ extern "C"{
 #define SPH_SMALL_FOOTPRINT_HAVAL   1
 //#endif
 
+#if defined(__AVX512VL__)
+
+// ( ~( a ^ b ) ) & c
+#define mm128_andnotxor( a, b, c ) \
+   _mm_ternarylogic_epi32( a, b, c, 0x82  )
+
+#else
+
+#define mm128_andnotxor( a, b, c ) \
+   _mm_andnot_si128( _mm_xor_si128( a, b ), c )
+
+#endif
+
+#define F1(x6, x5, x4, x3, x2, x1, x0) \
+ mm128_xor3( x0, mm128_andxor( x1, x0, x4 ), \
+                 _mm_xor_si128( _mm_and_si128( x2, x5 ), \
+                                _mm_and_si128( x3, x6 ) ) ) \
+
+#define F2(x6, x5, x4, x3, x2, x1, x0) \
+   mm128_xor3( mm128_andxor( x2, _mm_andnot_si128( x3, x1 ), \
+                       mm128_xor3( _mm_and_si128( x4, x5 ), x6, x0 )  ), \
+               mm128_andxor( x4, x1, x5 ), \
+               mm128_xorand( x0, x3, x5 ) ) \
+
+#define F3(x6, x5, x4, x3, x2, x1, x0) \
+  mm128_xor3( x0, \
+              _mm_and_si128( x3, \
+                         mm128_xor3( _mm_and_si128( x1, x2 ), x6, x0 ) ), \
+              _mm_xor_si128( _mm_and_si128( x1, x4 ), \
+                             _mm_and_si128( x2, x5 ) ) )
+
+#define F4(x6, x5, x4, x3, x2, x1, x0) \
+  mm128_xor3( \
+      mm128_andxor( x3, x5, \
+                    _mm_xor_si128( _mm_and_si128( x1, x2 ), \
+                                      _mm_or_si128( x4, x6 ) ) ), \
+      _mm_and_si128( x4, \
+                        mm128_xor3( x0, _mm_andnot_si128( x2, x5 ), \
+                                    _mm_xor_si128( x1, x6 ) ) ), \
+      mm128_xorand( x0, x2, x6 ) )
+
+#define F5(x6, x5, x4, x3, x2, x1, x0) \
+   _mm_xor_si128( \
+         mm128_andnotxor( mm128_and3( x1, x2, x3 ), x5, x0 ), \
+         mm128_xor3( _mm_and_si128( x1, x4 ), \
+                     _mm_and_si128( x2, x5 ), \
+                     _mm_and_si128( x3, x6 ) ) )
+  
+
+/*
 #define F1(x6, x5, x4, x3, x2, x1, x0) \
    _mm_xor_si128( x0, \
        _mm_xor_si128( _mm_and_si128(_mm_xor_si128( x0, x4 ), x1 ), \
@@ -96,6 +146,7 @@ extern "C"{
       _mm_xor_si128( _mm_xor_si128( _mm_and_si128( x1, x4 ), \
                                     _mm_and_si128( x2, x5 ) ), \
                                     _mm_and_si128( x3, x6 ) ) )
+*/
 
 /*
  * The macros below integrate the phi() permutations, depending on the
@@ -138,7 +189,7 @@ do { \
    __m128i t = FP ## n ## _ ## p(x6, x5, x4, x3, x2, x1, x0); \
    x7 = _mm_add_epi32( _mm_add_epi32( mm128_ror_32( t, 7 ), \
                                       mm128_ror_32( x7, 11 ) ), \
-                       _mm_add_epi32( w, _mm_set1_epi32( c ) ) ); \
+                       _mm_add_epi32( w, v128_32( c ) ) ); \
 } while (0)
 
 #define STEP1(n, p, x7, x6, x5, x4, x3, x2, x1, x0, w) \
@@ -241,7 +292,9 @@ static const unsigned MP5[32] = {
 	 2, 23, 16, 22,  4,  1, 25, 15
 };
 
-static const sph_u32 RK2[32] = {
+#define SPH_C32(x) (x)
+
+static const uint32_t RK2[32] = {
 	SPH_C32(0x452821E6), SPH_C32(0x38D01377),
 	SPH_C32(0xBE5466CF), SPH_C32(0x34E90C6C),
 	SPH_C32(0xC0AC29B7), SPH_C32(0xC97C50DD),
@@ -260,7 +313,7 @@ static const sph_u32 RK2[32] = {
 	SPH_C32(0x7B54A41D), SPH_C32(0xC25A59B5)
 };
 
-static const sph_u32 RK3[32] = {
+static const uint32_t RK3[32] = {
 	SPH_C32(0x9C30D539), SPH_C32(0x2AF26013),
 	SPH_C32(0xC5D1B023), SPH_C32(0x286085F0),
 	SPH_C32(0xCA417918), SPH_C32(0xB8DB38EF),
@@ -279,7 +332,7 @@ static const sph_u32 RK3[32] = {
 	SPH_C32(0xAFD6BA33), SPH_C32(0x6C24CF5C)
 };
 
-static const sph_u32 RK4[32] = {
+static const uint32_t RK4[32] = {
 	SPH_C32(0x7A325381), SPH_C32(0x28958677),
 	SPH_C32(0x3B8F4898), SPH_C32(0x6B4BB9AF),
 	SPH_C32(0xC4BFE81B), SPH_C32(0x66282193),
@@ -298,7 +351,7 @@ static const sph_u32 RK4[32] = {
 	SPH_C32(0x6EEF0B6C), SPH_C32(0x137A3BE4)
 };
 
-static const sph_u32 RK5[32] = {
+static const uint32_t RK5[32] = {
 	SPH_C32(0xBA3BF050), SPH_C32(0x7EFB2A98),
 	SPH_C32(0xA1F1651D), SPH_C32(0x39AF0176),
 	SPH_C32(0x66CA593E), SPH_C32(0x82430E88),
@@ -418,14 +471,14 @@ do { \
 static void
 haval_4way_init( haval_4way_context *sc, unsigned olen, unsigned passes )
 {
-   sc->s0 = _mm_set1_epi32( 0x243F6A88UL );
+   sc->s0 = v128_32( 0x243F6A88UL );
-   sc->s1 = _mm_set1_epi32( 0x85A308D3UL );
+   sc->s1 = v128_32( 0x85A308D3UL );
-   sc->s2 = _mm_set1_epi32( 0x13198A2EUL );
+   sc->s2 = v128_32( 0x13198A2EUL );
-   sc->s3 = _mm_set1_epi32( 0x03707344UL );
+   sc->s3 = v128_32( 0x03707344UL );
-   sc->s4 = _mm_set1_epi32( 0xA4093822UL );
+   sc->s4 = v128_32( 0xA4093822UL );
-   sc->s5 = _mm_set1_epi32( 0x299F31D0UL );
+   sc->s5 = v128_32( 0x299F31D0UL );
-   sc->s6 = _mm_set1_epi32( 0x082EFA98UL );
+   sc->s6 = v128_32( 0x082EFA98UL );
-   sc->s7 = _mm_set1_epi32( 0xEC4E6C89UL );
+   sc->s7 = v128_32( 0xEC4E6C89UL );
    sc->olen = olen;
    sc->passes = passes;
    sc->count_high = 0;
@@ -609,7 +662,7 @@ 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 ) ) ); \
+                       _mm256_add_epi32( w, v256_32( c ) ) ); \
 } while (0)
 
 #define STEP1_8W(n, p, x7, x6, x5, x4, x3, x2, x1, x0, w) \
@@ -740,14 +793,14 @@ do { \
 static void
 haval_8way_init( haval_8way_context *sc, unsigned olen, unsigned passes )
 {
-   sc->s0 = m256_const1_32( 0x243F6A88UL );
+   sc->s0 = v256_32( 0x243F6A88UL );
-   sc->s1 = m256_const1_32( 0x85A308D3UL );
+   sc->s1 = v256_32( 0x85A308D3UL );
-   sc->s2 = m256_const1_32( 0x13198A2EUL );
+   sc->s2 = v256_32( 0x13198A2EUL );
-   sc->s3 = m256_const1_32( 0x03707344UL );
+   sc->s3 = v256_32( 0x03707344UL );
-   sc->s4 = m256_const1_32( 0xA4093822UL );
+   sc->s4 = v256_32( 0xA4093822UL );
-   sc->s5 = m256_const1_32( 0x299F31D0UL );
+   sc->s5 = v256_32( 0x299F31D0UL );
-   sc->s6 = m256_const1_32( 0x082EFA98UL );
+   sc->s6 = v256_32( 0x082EFA98UL );
-   sc->s7 = m256_const1_32( 0xEC4E6C89UL );
+   sc->s7 = v256_32( 0xEC4E6C89UL );
    sc->olen = olen;
    sc->passes = passes;
    sc->count_high = 0;
@@ -826,10 +879,300 @@ do { \
 
 #define INMSG_8W(i)   msg[i]
 
-
-
 #endif // AVX2
 
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// ( ~( a ^ b ) ) & c
+#define mm512_andnotxor( a, b, c ) \
+   _mm512_ternarylogic_epi32( a, b, c, 0x82  )
+
+#define F1_16W(x6, x5, x4, x3, x2, x1, x0) \
+ mm512_xor3( x0, mm512_andxor( x1, x0, x4 ), \
+                 _mm512_xor_si512( _mm512_and_si512( x2, x5 ), \
+                                   _mm512_and_si512( x3, x6 ) ) ) \
+
+#define F2_16W(x6, x5, x4, x3, x2, x1, x0) \
+   mm512_xor3( mm512_andxor( x2, _mm512_andnot_si512( x3, x1 ), \
+                       mm512_xor3( _mm512_and_si512( x4, x5 ), x6, x0 )  ), \
+               mm512_andxor( x4, x1, x5 ), \
+               mm512_xorand( x0, x3, x5 ) ) \
+
+#define F3_16W(x6, x5, x4, x3, x2, x1, x0) \
+  mm512_xor3( x0, \
+              _mm512_and_si512( x3, \
+                         mm512_xor3( _mm512_and_si512( x1, x2 ), x6, x0 ) ), \
+              _mm512_xor_si512( _mm512_and_si512( x1, x4 ), \
+                                _mm512_and_si512( x2, x5 ) ) )
+
+#define F4_16W(x6, x5, x4, x3, x2, x1, x0) \
+  mm512_xor3( \
+      mm512_andxor( x3, x5, \
+                    _mm512_xor_si512( _mm512_and_si512( x1, x2 ), \
+                                      _mm512_or_si512( x4, x6 ) ) ), \
+      _mm512_and_si512( x4, \
+                        mm512_xor3( x0, _mm512_andnot_si512( x2, x5 ), \
+                                    _mm512_xor_si512( x1, x6 ) ) ), \
+      mm512_xorand( x0, x2, x6 ) )
+
+#define F5_16W(x6, x5, x4, x3, x2, x1, x0) \
+   _mm512_xor_si512( \
+         mm512_andnotxor( mm512_and3( x1, x2, x3 ), x5, x0 ), \
+         mm512_xor3( _mm512_and_si512( x1, x4 ), \
+                     _mm512_and_si512( x2, x5 ), \
+                     _mm512_and_si512( x3, x6 ) ) )
+
+#define FP3_1_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F1_16W(x1, x0, x3, x5, x6, x2, x4)
+#define FP3_2_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F2_16W(x4, x2, x1, x0, x5, x3, x6)
+#define FP3_3_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F3_16W(x6, x1, x2, x3, x4, x5, x0)
+
+#define FP4_1_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F1_16W(x2, x6, x1, x4, x5, x3, x0)
+#define FP4_2_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F2_16W(x3, x5, x2, x0, x1, x6, x4)
+#define FP4_3_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F3_16W(x1, x4, x3, x6, x0, x2, x5)
+#define FP4_4_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F4_16W(x6, x4, x0, x5, x2, x1, x3)
+
+#define FP5_1_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F1_16W(x3, x4, x1, x0, x5, x2, x6)
+#define FP5_2_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F2_16W(x6, x2, x1, x0, x3, x4, x5)
+#define FP5_3_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F3_16W(x2, x6, x0, x4, x3, x1, x5)
+#define FP5_4_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F4_16W(x1, x5, x3, x2, x0, x4, x6)
+#define FP5_5_16W(x6, x5, x4, x3, x2, x1, x0) \
+   F5_16W(x2, x5, x0, x6, x4, x3, x1)
+
+#define STEP_16W(n, p, x7, x6, x5, x4, x3, x2, x1, x0, w, c) \
+do { \
+   __m512i t = FP ## n ## _ ## p ## _16W(x6, x5, x4, x3, x2, x1, x0); \
+   x7 = _mm512_add_epi32( _mm512_add_epi32( mm512_ror_32( t, 7 ), \
+                                      mm512_ror_32( x7, 11 ) ), \
+                       _mm512_add_epi32( w, v512_32( c ) ) ); \
+} while (0)
+   
+#define STEP1_16W(n, p, x7, x6, x5, x4, x3, x2, x1, x0, w) \
+do { \
+   __m512i t = FP ## n ## _ ## p ## _16W(x6, x5, x4, x3, x2, x1, x0); \
+   x7 = _mm512_add_epi32( _mm512_add_epi32( mm512_ror_32( t, 7 ), \
+                                      mm512_ror_32( x7, 11 ) ), w ); \
+} while (0)
+
+#define PASS1_16W(n, in)   do { \
+      unsigned pass_count; \
+      for (pass_count = 0; pass_count < 32; pass_count += 8) { \
+         STEP1_16W(n, 1, s7, s6, s5, s4, s3, s2, s1, s0, \
+            in(pass_count + 0) ); \
+         STEP1_16W(n, 1, s6, s5, s4, s3, s2, s1, s0, s7, \
+            in(pass_count + 1) ); \
+         STEP1_16W(n, 1, s5, s4, s3, s2, s1, s0, s7, s6, \
+            in(pass_count + 2) ); \
+         STEP1_16W(n, 1, s4, s3, s2, s1, s0, s7, s6, s5, \
+            in(pass_count + 3) ); \
+         STEP1_16W(n, 1, s3, s2, s1, s0, s7, s6, s5, s4, \
+            in(pass_count + 4) ); \
+         STEP1_16W(n, 1, s2, s1, s0, s7, s6, s5, s4, s3, \
+            in(pass_count + 5) ); \
+         STEP1_16W(n, 1, s1, s0, s7, s6, s5, s4, s3, s2, \
+            in(pass_count + 6) ); \
+         STEP1_16W(n, 1, s0, s7, s6, s5, s4, s3, s2, s1, \
+            in(pass_count + 7) ); \
+         } \
+   } while (0)
+
+#define PASSG_16W(p, n, in)   do { \
+      unsigned pass_count; \
+      for (pass_count = 0; pass_count < 32; pass_count += 8) { \
+         STEP_16W(n, p, s7, s6, s5, s4, s3, s2, s1, s0, \
+            in(MP ## p[pass_count + 0]), \
+            RK ## p[pass_count + 0]); \
+         STEP_16W(n, p, s6, s5, s4, s3, s2, s1, s0, s7, \
+            in(MP ## p[pass_count + 1]), \
+            RK ## p[pass_count + 1]); \
+         STEP_16W(n, p, s5, s4, s3, s2, s1, s0, s7, s6, \
+            in(MP ## p[pass_count + 2]), \
+            RK ## p[pass_count + 2]); \
+         STEP_16W(n, p, s4, s3, s2, s1, s0, s7, s6, s5, \
+            in(MP ## p[pass_count + 3]), \
+            RK ## p[pass_count + 3]); \
+         STEP_16W(n, p, s3, s2, s1, s0, s7, s6, s5, s4, \
+            in(MP ## p[pass_count + 4]), \
+            RK ## p[pass_count + 4]); \
+         STEP_16W(n, p, s2, s1, s0, s7, s6, s5, s4, s3, \
+            in(MP ## p[pass_count + 5]), \
+            RK ## p[pass_count + 5]); \
+         STEP_16W(n, p, s1, s0, s7, s6, s5, s4, s3, s2, \
+            in(MP ## p[pass_count + 6]), \
+            RK ## p[pass_count + 6]); \
+         STEP_16W(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_16W(n, in)    PASSG_16W(2, n, in)
+#define PASS3_16W(n, in)    PASSG_16W(3, n, in)
+#define PASS4_16W(n, in)    PASSG_16W(4, n, in)
+#define PASS5_16W(n, in)    PASSG_16W(5, n, in)
+
+#define SAVE_STATE_16W \
+   __m512i 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_16W \
+do { \
+   s0 = _mm512_add_epi32( s0, u0 ); \
+   s1 = _mm512_add_epi32( s1, u1 ); \
+   s2 = _mm512_add_epi32( s2, u2 ); \
+   s3 = _mm512_add_epi32( s3, u3 ); \
+   s4 = _mm512_add_epi32( s4, u4 ); \
+   s5 = _mm512_add_epi32( s5, u5 ); \
+   s6 = _mm512_add_epi32( s6, u6 ); \
+   s7 = _mm512_add_epi32( s7, u7 ); \
+} while (0)
+
+#define CORE_16W5(in)  do { \
+      SAVE_STATE_16W; \
+      PASS1_16W(5, in); \
+      PASS2_16W(5, in); \
+      PASS3_16W(5, in); \
+      PASS4_16W(5, in); \
+      PASS5_16W(5, in); \
+      UPDATE_STATE_16W; \
+   } while (0)
+
+#define DSTATE_16W   __m512i s0, s1, s2, s3, s4, s5, s6, s7
+
+#define RSTATE_16W \
+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_16W \
+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_16way_init( haval_16way_context *sc, unsigned olen, unsigned passes )
+{
+   sc->s0 = v512_32( 0x243F6A88UL );
+   sc->s1 = v512_32( 0x85A308D3UL );
+   sc->s2 = v512_32( 0x13198A2EUL );
+   sc->s3 = v512_32( 0x03707344UL );
+   sc->s4 = v512_32( 0xA4093822UL );
+   sc->s5 = v512_32( 0x299F31D0UL );
+   sc->s6 = v512_32( 0x082EFA98UL );
+   sc->s7 = v512_32( 0xEC4E6C89UL );
+   sc->olen = olen;
+   sc->passes = passes;
+   sc->count_high = 0;
+   sc->count_low = 0;
+
+}
+#define IN_PREPARE_16W(indata) const __m512i *const load_ptr_16w = (indata)
+
+#define INW_16W(i)   load_ptr_16w[ i ] 
+
+static void
+haval_16way_out( haval_16way_context *sc, void *dst )
+{
+   __m512i *buf = (__m512i*)dst;
+   DSTATE_16W;
+   RSTATE_16W;
+
+   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-16way-helper.c"
+
+#define API_16W(xxx, y) \
+void \
+haval ## xxx ## _ ## y ## _16way_init(void *cc) \
+{ \
+   haval_16way_init(cc, xxx >> 5, y); \
+} \
+ \
+void \
+haval ## xxx ## _ ## y ## _16way_update (void *cc, const void *data, size_t len) \
+{ \
+   haval ## y ## _16way_update(cc, data, len); \
+} \
+ \
+void \
+haval ## xxx ## _ ## y ## _16way_close(void *cc, void *dst) \
+{ \
+   haval ## y ## _16way_close(cc, dst); \
+} \
+
+API_16W(256, 5)
+
+#define RVAL_16W \
+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_16W \
+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_16W(i)   msg[i]
+
+#endif
+
 #ifdef __cplusplus
 }
 #endif	

algo/haval/haval-hash-4way.h

@@ -68,7 +68,6 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
 #include "simd-utils.h"
 
 #define SPH_SIZE_haval256_5   256
@@ -77,7 +76,7 @@ typedef struct {
    __m128i buf[32];
    __m128i s0, s1, s2, s3, s4, s5, s6, s7;
    unsigned olen, passes;
-   sph_u32 count_high, count_low;
+   uint32_t count_high, count_low;
 } haval_4way_context;
 
 typedef haval_4way_context haval256_5_4way_context;
@@ -108,6 +107,25 @@ void haval256_5_8way_close( void *cc, void *dst );
 
 #endif // AVX2
 
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+typedef struct {
+   __m512i buf[32];
+   __m512i s0, s1, s2, s3, s4, s5, s6, s7;
+   unsigned olen, passes;
+   uint32_t count_high, count_low;
+} haval_16way_context __attribute__ ((aligned (64)));
+
+typedef haval_16way_context haval256_5_16way_context;
+
+void haval256_5_16way_init( void *cc );
+
+void haval256_5_16way_update( void *cc, const void *data, size_t len );
+
+void haval256_5_16way_close( void *cc, void *dst );
+
+#endif // AVX512
+
 #ifdef __cplusplus
 }
 #endif

algo/haval/sph-haval.h

@@ -66,7 +66,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for HAVAL-128/3.

algo/jh/jh-hash-4way.c

@@ -76,19 +76,31 @@ do { \
 
 #endif
 
+#if defined(__AVX512VL__)
+//TODO enable for AVX10_256, not used with AVX512VL
+
+#define notxorandnot( a, b, c ) \
+   _mm256_ternarylogic_epi64( a, b, c, 0x2d )
+
+#else
+
+#define notxorandnot( a, b, c ) \
+   _mm256_xor_si256( mm256_not( a ), _mm256_andnot_si256( b, c ) )
+
+#endif
+
 #define Sb(x0, x1, x2, x3, c) \
 do { \
-   const __m256i cc = _mm256_set1_epi64x( c ); \
+    const __m256i cc = _mm256_set1_epi64x( c ); \
-    x3 = mm256_not( x3 ); \
+    x0 = mm256_xorandnot( x0, x2, cc ); \
-    x0 = _mm256_xor_si256( x0, _mm256_andnot_si256( x2, cc ) ); \
+    tmp = mm256_xorand( cc, x0, x1 ); \
-    tmp = _mm256_xor_si256( cc, _mm256_and_si256( x0, x1 ) ); \
+    x0 = mm256_xorandnot( x0, x3, x2 ); \
-    x0 = _mm256_xor_si256( x0, _mm256_and_si256( x2, x3 ) ); \
+    x3 = notxorandnot( x3, x1, x2 ); \
-    x3 = _mm256_xor_si256( x3, _mm256_andnot_si256( x1, x2 ) ); \
+    x1 = mm256_xorand( x1, x0, x2 ); \
-    x1 = _mm256_xor_si256( x1, _mm256_and_si256( x0, x2 ) ); \
+    x2 = mm256_xorandnot( x2, x3, x0 ); \
-    x2 = _mm256_xor_si256( x2, _mm256_andnot_si256( x3, x0 ) ); \
+    x0 = mm256_xoror( x0, x1, x3 ); \
-    x0 = _mm256_xor_si256( x0, _mm256_or_si256( x1, x3 ) ); \
+    x3 = mm256_xorand( x3, x1, x2 ); \
-    x3 = _mm256_xor_si256( x3, _mm256_and_si256( x1, x2 ) ); \
+    x1 = mm256_xorand( x1, tmp, x0 ); \
-    x1 = _mm256_xor_si256( x1, _mm256_and_si256( tmp, x0 ) ); \
     x2 = _mm256_xor_si256( x2, tmp ); \
 } while (0)
 
@@ -96,11 +108,11 @@ do { \
 do { \
     x4 = _mm256_xor_si256( x4, x1 ); \
     x5 = _mm256_xor_si256( x5, x2 ); \
-    x6 = _mm256_xor_si256( x6, _mm256_xor_si256( x3, x0 ) ); \
+    x6 = mm256_xor3( x6, x3, x0 ); \
     x7 = _mm256_xor_si256( x7, x0 ); \
     x0 = _mm256_xor_si256( x0, x5 ); \
     x1 = _mm256_xor_si256( x1, x6 ); \
-    x2 = _mm256_xor_si256( x2, _mm256_xor_si256( x7, x4 ) ); \
+    x2 = mm256_xor3( x2, x7, x4 ); \
     x3 = _mm256_xor_si256( x3, x4 ); \
 } while (0)
 
@@ -323,12 +335,12 @@ do { \
 } while (0)
 
 
-#define W80(x)   Wz_8W(x, m512_const1_64( 0x5555555555555555 ),  1 )
+#define W80(x)   Wz_8W(x, _mm512_set1_epi64( 0x5555555555555555 ),  1 )
-#define W81(x)   Wz_8W(x, m512_const1_64( 0x3333333333333333 ),  2 )
+#define W81(x)   Wz_8W(x, _mm512_set1_epi64( 0x3333333333333333 ),  2 )
-#define W82(x)   Wz_8W(x, m512_const1_64( 0x0F0F0F0F0F0F0F0F ),  4 )
+#define W82(x)   Wz_8W(x, _mm512_set1_epi64( 0x0F0F0F0F0F0F0F0F ),  4 )
-#define W83(x)   Wz_8W(x, m512_const1_64( 0x00FF00FF00FF00FF ),  8 ) 
+#define W83(x)   Wz_8W(x, _mm512_set1_epi64( 0x00FF00FF00FF00FF ),  8 ) 
-#define W84(x)   Wz_8W(x, m512_const1_64( 0x0000FFFF0000FFFF ), 16 )
+#define W84(x)   Wz_8W(x, _mm512_set1_epi64( 0x0000FFFF0000FFFF ), 16 )
-#define W85(x)   Wz_8W(x, m512_const1_64( 0x00000000FFFFFFFF ), 32 )
+#define W85(x)   Wz_8W(x, _mm512_set1_epi64( 0x00000000FFFFFFFF ), 32 )
 #define W86(x) \
 do { \
    __m512i t = x ## h; \
@@ -352,12 +364,12 @@ do { \
    x ## l = _mm256_or_si256( _mm256_and_si256((x ## l >> (n)), (c)), t ); \
 } while (0)
 
-#define W0(x)   Wz(x, m256_const1_64( 0x5555555555555555 ),  1 )
+#define W0(x)   Wz(x, _mm256_set1_epi64x( 0x5555555555555555 ),  1 )
-#define W1(x)   Wz(x, m256_const1_64( 0x3333333333333333 ),  2 )
+#define W1(x)   Wz(x, _mm256_set1_epi64x( 0x3333333333333333 ),  2 )
-#define W2(x)   Wz(x, m256_const1_64( 0x0F0F0F0F0F0F0F0F ),  4 )
+#define W2(x)   Wz(x, _mm256_set1_epi64x( 0x0F0F0F0F0F0F0F0F ),  4 )
-#define W3(x)   Wz(x, m256_const1_64( 0x00FF00FF00FF00FF ),  8 ) 
+#define W3(x)   Wz(x, _mm256_set1_epi64x( 0x00FF00FF00FF00FF ),  8 ) 
-#define W4(x)   Wz(x, m256_const1_64( 0x0000FFFF0000FFFF ), 16 )
+#define W4(x)   Wz(x, _mm256_set1_epi64x( 0x0000FFFF0000FFFF ), 16 )
-#define W5(x)   Wz(x, m256_const1_64( 0x00000000FFFFFFFF ), 32 )
+#define W5(x)   Wz(x, _mm256_set1_epi64x( 0x00000000FFFFFFFF ), 32 )
 #define W6(x) \
 do { \
    __m256i t = x ## h; \
@@ -624,22 +636,22 @@ static const sph_u64 IV512[] = {
 void jh256_8way_init( jh_8way_context *sc )
 {
     // bswapped IV256
-    sc->H[ 0] = m512_const1_64( 0xebd3202c41a398eb );
+    sc->H[ 0] = _mm512_set1_epi64( 0xebd3202c41a398eb );
-    sc->H[ 1] = m512_const1_64( 0xc145b29c7bbecd92 );
+    sc->H[ 1] = _mm512_set1_epi64( 0xc145b29c7bbecd92 );
-    sc->H[ 2] = m512_const1_64( 0xfac7d4609151931c );
+    sc->H[ 2] = _mm512_set1_epi64( 0xfac7d4609151931c );
-    sc->H[ 3] = m512_const1_64( 0x038a507ed6820026 );
+    sc->H[ 3] = _mm512_set1_epi64( 0x038a507ed6820026 );
-    sc->H[ 4] = m512_const1_64( 0x45b92677269e23a4 );
+    sc->H[ 4] = _mm512_set1_epi64( 0x45b92677269e23a4 );
-    sc->H[ 5] = m512_const1_64( 0x77941ad4481afbe0 );
+    sc->H[ 5] = _mm512_set1_epi64( 0x77941ad4481afbe0 );
-    sc->H[ 6] = m512_const1_64( 0x7a176b0226abb5cd );
+    sc->H[ 6] = _mm512_set1_epi64( 0x7a176b0226abb5cd );
-    sc->H[ 7] = m512_const1_64( 0xa82fff0f4224f056 );
+    sc->H[ 7] = _mm512_set1_epi64( 0xa82fff0f4224f056 );
-    sc->H[ 8] = m512_const1_64( 0x754d2e7f8996a371 );
+    sc->H[ 8] = _mm512_set1_epi64( 0x754d2e7f8996a371 );
-    sc->H[ 9] = m512_const1_64( 0x62e27df70849141d );
+    sc->H[ 9] = _mm512_set1_epi64( 0x62e27df70849141d );
-    sc->H[10] = m512_const1_64( 0x948f2476f7957627 );
+    sc->H[10] = _mm512_set1_epi64( 0x948f2476f7957627 );
-    sc->H[11] = m512_const1_64( 0x6c29804757b6d587 );
+    sc->H[11] = _mm512_set1_epi64( 0x6c29804757b6d587 );
-    sc->H[12] = m512_const1_64( 0x6c0d8eac2d275e5c );
+    sc->H[12] = _mm512_set1_epi64( 0x6c0d8eac2d275e5c );
-    sc->H[13] = m512_const1_64( 0x0f7a0557c6508451 );
+    sc->H[13] = _mm512_set1_epi64( 0x0f7a0557c6508451 );
-    sc->H[14] = m512_const1_64( 0xea12247067d3e47b );
+    sc->H[14] = _mm512_set1_epi64( 0xea12247067d3e47b );
-    sc->H[15] = m512_const1_64( 0x69d71cd313abe389 );
+    sc->H[15] = _mm512_set1_epi64( 0x69d71cd313abe389 );
     sc->ptr = 0;
     sc->block_count = 0;
 }
@@ -647,22 +659,22 @@ void jh256_8way_init( jh_8way_context *sc )
 void jh512_8way_init( jh_8way_context *sc )
 {
     // bswapped IV512
-    sc->H[ 0] = m512_const1_64( 0x17aa003e964bd16f );
+    sc->H[ 0] = _mm512_set1_epi64( 0x17aa003e964bd16f );
-    sc->H[ 1] = m512_const1_64( 0x43d5157a052e6a63 );
+    sc->H[ 1] = _mm512_set1_epi64( 0x43d5157a052e6a63 );
-    sc->H[ 2] = m512_const1_64( 0x0bef970c8d5e228a );
+    sc->H[ 2] = _mm512_set1_epi64( 0x0bef970c8d5e228a );
-    sc->H[ 3] = m512_const1_64( 0x61c3b3f2591234e9 );
+    sc->H[ 3] = _mm512_set1_epi64( 0x61c3b3f2591234e9 );
-    sc->H[ 4] = m512_const1_64( 0x1e806f53c1a01d89 );
+    sc->H[ 4] = _mm512_set1_epi64( 0x1e806f53c1a01d89 );
-    sc->H[ 5] = m512_const1_64( 0x806d2bea6b05a92a );
+    sc->H[ 5] = _mm512_set1_epi64( 0x806d2bea6b05a92a );
-    sc->H[ 6] = m512_const1_64( 0xa6ba7520dbcc8e58 );
+    sc->H[ 6] = _mm512_set1_epi64( 0xa6ba7520dbcc8e58 );
-    sc->H[ 7] = m512_const1_64( 0xf73bf8ba763a0fa9 );
+    sc->H[ 7] = _mm512_set1_epi64( 0xf73bf8ba763a0fa9 );
-    sc->H[ 8] = m512_const1_64( 0x694ae34105e66901 );
+    sc->H[ 8] = _mm512_set1_epi64( 0x694ae34105e66901 );
-    sc->H[ 9] = m512_const1_64( 0x5ae66f2e8e8ab546 );
+    sc->H[ 9] = _mm512_set1_epi64( 0x5ae66f2e8e8ab546 );
-    sc->H[10] = m512_const1_64( 0x243c84c1d0a74710 );
+    sc->H[10] = _mm512_set1_epi64( 0x243c84c1d0a74710 );
-    sc->H[11] = m512_const1_64( 0x99c15a2db1716e3b );
+    sc->H[11] = _mm512_set1_epi64( 0x99c15a2db1716e3b );
-    sc->H[12] = m512_const1_64( 0x56f8b19decf657cf );
+    sc->H[12] = _mm512_set1_epi64( 0x56f8b19decf657cf );
-    sc->H[13] = m512_const1_64( 0x56b116577c8806a7 );
+    sc->H[13] = _mm512_set1_epi64( 0x56b116577c8806a7 );
-    sc->H[14] = m512_const1_64( 0xfb1785e6dffcc2e3 );
+    sc->H[14] = _mm512_set1_epi64( 0xfb1785e6dffcc2e3 );
-    sc->H[15] = m512_const1_64( 0x4bdd8ccc78465a54 );
+    sc->H[15] = _mm512_set1_epi64( 0x4bdd8ccc78465a54 );
     sc->ptr = 0;
     sc->block_count = 0;
 }
@@ -721,7 +733,7 @@ jh_8way_close( jh_8way_context *sc, unsigned ub, unsigned n, void *dst,
    size_t numz, u;
    uint64_t l0, l1;
 
-   buf[0] = m512_const1_64( 0x80ULL );
+   buf[0] = _mm512_set1_epi64( 0x80ULL );
 
    if ( sc->ptr == 0 )
        numz = 48;
@@ -772,22 +784,22 @@ jh512_8way_close(void *cc, void *dst)
 void jh256_4way_init( jh_4way_context *sc )
 {
     // bswapped IV256
-    sc->H[ 0] = m256_const1_64( 0xebd3202c41a398eb );
+    sc->H[ 0] = _mm256_set1_epi64x( 0xebd3202c41a398eb );
-    sc->H[ 1] = m256_const1_64( 0xc145b29c7bbecd92 );
+    sc->H[ 1] = _mm256_set1_epi64x( 0xc145b29c7bbecd92 );
-    sc->H[ 2] = m256_const1_64( 0xfac7d4609151931c );
+    sc->H[ 2] = _mm256_set1_epi64x( 0xfac7d4609151931c );
-    sc->H[ 3] = m256_const1_64( 0x038a507ed6820026 );
+    sc->H[ 3] = _mm256_set1_epi64x( 0x038a507ed6820026 );
-    sc->H[ 4] = m256_const1_64( 0x45b92677269e23a4 );
+    sc->H[ 4] = _mm256_set1_epi64x( 0x45b92677269e23a4 );
-    sc->H[ 5] = m256_const1_64( 0x77941ad4481afbe0 );
+    sc->H[ 5] = _mm256_set1_epi64x( 0x77941ad4481afbe0 );
-    sc->H[ 6] = m256_const1_64( 0x7a176b0226abb5cd );
+    sc->H[ 6] = _mm256_set1_epi64x( 0x7a176b0226abb5cd );
-    sc->H[ 7] = m256_const1_64( 0xa82fff0f4224f056 );
+    sc->H[ 7] = _mm256_set1_epi64x( 0xa82fff0f4224f056 );
-    sc->H[ 8] = m256_const1_64( 0x754d2e7f8996a371 );
+    sc->H[ 8] = _mm256_set1_epi64x( 0x754d2e7f8996a371 );
-    sc->H[ 9] = m256_const1_64( 0x62e27df70849141d );
+    sc->H[ 9] = _mm256_set1_epi64x( 0x62e27df70849141d );
-    sc->H[10] = m256_const1_64( 0x948f2476f7957627 );
+    sc->H[10] = _mm256_set1_epi64x( 0x948f2476f7957627 );
-    sc->H[11] = m256_const1_64( 0x6c29804757b6d587 );
+    sc->H[11] = _mm256_set1_epi64x( 0x6c29804757b6d587 );
-    sc->H[12] = m256_const1_64( 0x6c0d8eac2d275e5c );
+    sc->H[12] = _mm256_set1_epi64x( 0x6c0d8eac2d275e5c );
-    sc->H[13] = m256_const1_64( 0x0f7a0557c6508451 );
+    sc->H[13] = _mm256_set1_epi64x( 0x0f7a0557c6508451 );
-    sc->H[14] = m256_const1_64( 0xea12247067d3e47b );
+    sc->H[14] = _mm256_set1_epi64x( 0xea12247067d3e47b );
-    sc->H[15] = m256_const1_64( 0x69d71cd313abe389 );
+    sc->H[15] = _mm256_set1_epi64x( 0x69d71cd313abe389 );
     sc->ptr = 0;
     sc->block_count = 0;
 }
@@ -795,22 +807,22 @@ void jh256_4way_init( jh_4way_context *sc )
 void jh512_4way_init( jh_4way_context *sc )
 {
     // bswapped IV512
-    sc->H[ 0] = m256_const1_64( 0x17aa003e964bd16f );
+    sc->H[ 0] = _mm256_set1_epi64x( 0x17aa003e964bd16f );
-    sc->H[ 1] = m256_const1_64( 0x43d5157a052e6a63 );
+    sc->H[ 1] = _mm256_set1_epi64x( 0x43d5157a052e6a63 );
-    sc->H[ 2] = m256_const1_64( 0x0bef970c8d5e228a );
+    sc->H[ 2] = _mm256_set1_epi64x( 0x0bef970c8d5e228a );
-    sc->H[ 3] = m256_const1_64( 0x61c3b3f2591234e9 );
+    sc->H[ 3] = _mm256_set1_epi64x( 0x61c3b3f2591234e9 );
-    sc->H[ 4] = m256_const1_64( 0x1e806f53c1a01d89 );
+    sc->H[ 4] = _mm256_set1_epi64x( 0x1e806f53c1a01d89 );
-    sc->H[ 5] = m256_const1_64( 0x806d2bea6b05a92a );
+    sc->H[ 5] = _mm256_set1_epi64x( 0x806d2bea6b05a92a );
-    sc->H[ 6] = m256_const1_64( 0xa6ba7520dbcc8e58 );
+    sc->H[ 6] = _mm256_set1_epi64x( 0xa6ba7520dbcc8e58 );
-    sc->H[ 7] = m256_const1_64( 0xf73bf8ba763a0fa9 );
+    sc->H[ 7] = _mm256_set1_epi64x( 0xf73bf8ba763a0fa9 );
-    sc->H[ 8] = m256_const1_64( 0x694ae34105e66901 );
+    sc->H[ 8] = _mm256_set1_epi64x( 0x694ae34105e66901 );
-    sc->H[ 9] = m256_const1_64( 0x5ae66f2e8e8ab546 );
+    sc->H[ 9] = _mm256_set1_epi64x( 0x5ae66f2e8e8ab546 );
-    sc->H[10] = m256_const1_64( 0x243c84c1d0a74710 );
+    sc->H[10] = _mm256_set1_epi64x( 0x243c84c1d0a74710 );
-    sc->H[11] = m256_const1_64( 0x99c15a2db1716e3b );
+    sc->H[11] = _mm256_set1_epi64x( 0x99c15a2db1716e3b );
-    sc->H[12] = m256_const1_64( 0x56f8b19decf657cf );
+    sc->H[12] = _mm256_set1_epi64x( 0x56f8b19decf657cf );
-    sc->H[13] = m256_const1_64( 0x56b116577c8806a7 );
+    sc->H[13] = _mm256_set1_epi64x( 0x56b116577c8806a7 );
-    sc->H[14] = m256_const1_64( 0xfb1785e6dffcc2e3 );
+    sc->H[14] = _mm256_set1_epi64x( 0xfb1785e6dffcc2e3 );
-    sc->H[15] = m256_const1_64( 0x4bdd8ccc78465a54 );
+    sc->H[15] = _mm256_set1_epi64x( 0x4bdd8ccc78465a54 );
     sc->ptr = 0;
     sc->block_count = 0;
 }
@@ -869,7 +881,7 @@ jh_4way_close( jh_4way_context *sc, unsigned ub, unsigned n, void *dst,
    size_t numz, u;
    uint64_t l0, l1;
 
-   buf[0] = m256_const1_64( 0x80ULL );
+   buf[0] = _mm256_set1_epi64x( 0x80ULL );
 
    if ( sc->ptr == 0 )
        numz = 48;

algo/jh/jha-4way.c

@@ -6,7 +6,7 @@
 
 #if defined(JHA_4WAY)
 
-#include "algo/blake/blake-hash-4way.h"
+#include "algo/blake/blake512-hash.h"
 #include "algo/skein/skein-hash-4way.h"
 #include "algo/jh/jh-hash-4way.h"
 #include "algo/keccak/keccak-hash-4way.h"

algo/jh/sph_jh.h

@@ -41,7 +41,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for JH-224.

algo/keccak/keccak-4way.c

@@ -2,7 +2,6 @@
 #include <stdlib.h>
 #include <string.h>
 #include <stdint.h>
-#include "sph_keccak.h"
 #include "keccak-hash-4way.h"
 
 #if defined(KECCAK_8WAY)
@@ -49,7 +48,7 @@ int scanhash_keccak_8way( struct work *work, uint32_t max_nonce,
           }
       }
       *noncev = _mm512_add_epi32( *noncev,
-                                  m512_const1_64( 0x0000000800000000 ) );
+                                  _mm512_set1_epi64( 0x0000000800000000 ) );
       n += 8;
 
    } while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
@@ -101,7 +100,7 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
           }
       }
       *noncev = _mm256_add_epi32( *noncev,
-                                  m256_const1_64( 0x0000000400000000 ) );
+                                  _mm256_set1_epi64x( 0x0000000400000000 ) );
       n += 4;
    } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
    pdata[19] = n;

algo/keccak/keccak-gate.c

@@ -9,7 +9,7 @@ int hard_coded_eb = 1;
 bool register_keccak_algo( algo_gate_t* gate )
 {
   gate->optimizations = AVX2_OPT | AVX512_OPT;
-  gate->gen_merkle_root = (void*)&SHA256_gen_merkle_root;
+  gate->gen_merkle_root = (void*)&sha256_gen_merkle_root;
   opt_target_factor = 128.0;
 #if defined (KECCAK_8WAY)
   gate->scanhash  = (void*)&scanhash_keccak_8way;

algo/keccak/keccak-hash-4way.c

@@ -180,15 +180,15 @@ static void keccak64_8way_close( keccak64_ctx_m512i *kc, void *dst,
     if ( kc->ptr == (lim - 8) )
     {
         const uint64_t t = eb | 0x8000000000000000;
-        u.tmp[0] = m512_const1_64( t );
+        u.tmp[0] = _mm512_set1_epi64( t );
         j = 8;
     }
     else
     {
         j = lim - kc->ptr;
-        u.tmp[0] = m512_const1_64( eb );
+        u.tmp[0] = _mm512_set1_epi64( eb );
         memset_zero_512( u.tmp + 1, (j>>3) - 2 );
-        u.tmp[ (j>>3) - 1] = m512_const1_64( 0x8000000000000000 );
+        u.tmp[ (j>>3) - 1] = _mm512_set1_epi64( 0x8000000000000000 );
     }
     keccak64_8way_core( kc, u.tmp, j, lim );
     /* Finalize the "lane complement" */
@@ -264,8 +264,8 @@ keccak512_8way_close(void *cc, void *dst)
 #define OR64(d, a, b)      (d = _mm256_or_si256(a,b))
 #define NOT64(d, s)        (d = mm256_not( s ) )
 #define ROL64(d, v, n)     (d = mm256_rol_64(v, n))
-#define XOROR(d, a, b, c)  (d = _mm256_xor_si256(a, _mm256_or_si256(b, c)))
+#define XOROR(d, a, b, c)  (d = mm256_xoror( a, b, c ) )
-#define XORAND(d, a, b, c) (d = _mm256_xor_si256(a, _mm256_and_si256(b, c)))
+#define XORAND(d, a, b, c) (d = mm256_xorand( a, b, c ) )
 #define XOR3( d, a, b, c ) (d = mm256_xor3( a, b, c ))
 
 #include "keccak-macros.c"
@@ -368,15 +368,15 @@ static void keccak64_close( keccak64_ctx_m256i *kc, void *dst, size_t byte_len,
     if ( kc->ptr == (lim - 8) )
     {
         const uint64_t t = eb | 0x8000000000000000;
-        u.tmp[0] = m256_const1_64( t );
+        u.tmp[0] = _mm256_set1_epi64x( t );
         j = 8;
     }
     else
     {
         j = lim - kc->ptr;
-        u.tmp[0] = m256_const1_64( eb );
+        u.tmp[0] = _mm256_set1_epi64x( eb );
         memset_zero_256( u.tmp + 1, (j>>3) - 2 );
-        u.tmp[ (j>>3) - 1] = m256_const1_64( 0x8000000000000000 );
+        u.tmp[ (j>>3) - 1] = _mm256_set1_epi64x( 0x8000000000000000 );
     }
     keccak64_core( kc, u.tmp, j, lim );
     /* Finalize the "lane complement" */

algo/keccak/keccak-hash-4way.h

@@ -1,45 +1,6 @@
-/* $Id: sph_keccak.h 216 2010-06-08 09:46:57Z tp $ */
-/**
- * Keccak interface. This is the interface for Keccak with the
- * recommended parameters for SHA-3, with output lengths 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_keccak.h
- * @author   Thomas Pornin <thomas.pornin@cryptolog.com>
- */
-
 #ifndef KECCAK_HASH_4WAY_H__
 #define KECCAK_HASH_4WAY_H__
 
-#ifdef __cplusplus
-extern "C"{
-#endif
-
 #ifdef  __AVX2__
 
 #include <stddef.h>
@@ -100,8 +61,4 @@ void keccak512_4way_addbits_and_close(
 
 #endif
 
-#ifdef __cplusplus
-}
-#endif
-
 #endif

algo/keccak/sha3d-4way.c

@@ -2,7 +2,6 @@
 #include <stdlib.h>
 #include <string.h>
 #include <stdint.h>
-#include "sph_keccak.h"
 #include "keccak-hash-4way.h"
 
 #if defined(KECCAK_8WAY)
@@ -56,7 +55,7 @@ int scanhash_sha3d_8way( struct work *work, uint32_t max_nonce,
           }
       }
       *noncev = _mm512_add_epi32( *noncev,
-                                  m512_const1_64( 0x0000000800000000 ) );
+                                  _mm512_set1_epi64( 0x0000000800000000 ) );
       n += 8;
 
    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
@@ -115,7 +114,7 @@ int scanhash_sha3d_4way( struct work *work, uint32_t max_nonce,
           }
       }
       *noncev = _mm256_add_epi32( *noncev,
-                                  m256_const1_64( 0x0000000400000000 ) );
+                                  _mm256_set1_epi64x( 0x0000000400000000 ) );
       n += 4;
    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
    pdata[19] = n;

algo/keccak/sph_keccak.h

@@ -41,7 +41,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for Keccak-224.

algo/lanehash/lane.h

@@ -23,7 +23,6 @@
 #define LANE_H
 
 #include <string.h>
-//#include "algo/sha/sha3-defs.h"
 #include <stdint.h>
 
 typedef unsigned char BitSequence;

algo/luffa/luffa-hash-2way.c

@@ -7,8 +7,10 @@
 
 #include "simd-utils.h"
 
+#define uint32 uint32_t
+
 /* initial values of chaining variables */
-static const uint32 IV[40] __attribute((aligned(64))) = {
+static const uint32_t IV[40] __attribute((aligned(64))) = {
     0xdbf78465,0x4eaa6fb4,0x44b051e0,0x6d251e69,
     0xdef610bb,0xee058139,0x90152df4,0x6e292011,
     0xde099fa3,0x70eee9a0,0xd9d2f256,0xc3b44b95,
@@ -22,7 +24,7 @@ static const uint32 IV[40] __attribute((aligned(64))) = {
 };
 
 /* Round Constants */
-static const uint32 CNS_INIT[128] __attribute((aligned(64))) = {
+static const uint32_t CNS_INIT[128] __attribute((aligned(64))) = {
     0xb213afa5,0xfc20d9d2,0xb6de10ed,0x303994a6,
     0xe028c9bf,0xe25e72c1,0x01685f3d,0xe0337818,
     0xc84ebe95,0x34552e25,0x70f47aae,0xc0e65299,
@@ -60,7 +62,7 @@ static const uint32 CNS_INIT[128] __attribute((aligned(64))) = {
 
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
-#define cns4w(i)  m512_const1_128( ( (__m128i*)CNS_INIT)[i] )
+#define cns4w(i)  mm512_bcast_m128( ( (__m128i*)CNS_INIT)[i] )
 
 #define ADD_CONSTANT4W( a, b, c0, c1 ) \
     a = _mm512_xor_si512( a, c0 ); \
@@ -69,7 +71,7 @@ static const uint32 CNS_INIT[128] __attribute((aligned(64))) = {
 #define MULT24W( a0, a1 ) \
 { \
   __m512i b = _mm512_xor_si512( a0, \
-                     _mm512_maskz_shuffle_epi32( 0xbbbb, a1, 16 ) ); \
+                     _mm512_maskz_shuffle_epi32( 0xbbbb, a1, 0x10 ) ); \
   a0 = _mm512_alignr_epi8( a1,  b, 4 ); \
   a1 = _mm512_alignr_epi8(  b, a1, 4 ); \
 }
@@ -107,58 +109,45 @@ static const uint32 CNS_INIT[128] __attribute((aligned(64))) = {
     ADD_CONSTANT4W( x0, x4, c0, c1 );
 
 #define STEP_PART24W( a0, a1, t0, t1, c0, c1 ) \
-    a1 = _mm512_shuffle_epi32( a1, 147 ); \
+    t0 = _mm512_shuffle_epi32( a1, 147 ); \
-    t0 = _mm512_load_si512( &a1 ); \
+    a1 = _mm512_unpacklo_epi32( t0, a0 ); \
-    a1 = _mm512_unpacklo_epi32( a1, a0 ); \
     t0 = _mm512_unpackhi_epi32( t0, a0 ); \
     t1 = _mm512_shuffle_epi32( t0, 78 ); \
     a0 = _mm512_shuffle_epi32( a1, 78 ); \
     SUBCRUMB4W( t1, t0, a0, a1 ); \
     t0 = _mm512_unpacklo_epi32( t0, t1 ); \
     a1 = _mm512_unpacklo_epi32( a1, a0 ); \
-    a0 = _mm512_load_si512( &a1 ); \
+    a0 = _mm512_unpackhi_epi64( a1, t0 ); \
-    a0 = _mm512_unpackhi_epi64( a0, t0 ); \
     a1 = _mm512_unpacklo_epi64( a1, t0 ); \
     a1 = _mm512_shuffle_epi32( a1, 57 ); \
     MIXWORD4W( a0, a1 ); \
     ADD_CONSTANT4W( a0, a1, c0, c1 );
 
 #define NMLTOM10244W(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3)\
-    s1 = _mm512_load_si512(&r3);\
+    s1 = _mm512_unpackhi_epi32( r3, r2 ); \
-    q1 = _mm512_load_si512(&p3);\
+    q1 = _mm512_unpackhi_epi32( p3, p2 ); \
-    s3 = _mm512_load_si512(&r3);\
+    s3 = _mm512_unpacklo_epi32( r3, r2 ); \
-    q3 = _mm512_load_si512(&p3);\
+    q3 = _mm512_unpacklo_epi32( p3, p2 ); \
-    s1 = _mm512_unpackhi_epi32(s1,r2);\
+    r3 = _mm512_unpackhi_epi32( r1, r0 ); \
-    q1 = _mm512_unpackhi_epi32(q1,p2);\
+    r1 = _mm512_unpacklo_epi32( r1, r0 ); \
-    s3 = _mm512_unpacklo_epi32(s3,r2);\
+    p3 = _mm512_unpackhi_epi32( p1, p0 ); \
-    q3 = _mm512_unpacklo_epi32(q3,p2);\
+    p1 = _mm512_unpacklo_epi32( p1, p0 ); \
-    s0 = _mm512_load_si512(&s1);\
+    s0 = _mm512_unpackhi_epi64( s1, r3 ); \
-    q0 = _mm512_load_si512(&q1);\
+    q0 = _mm512_unpackhi_epi64( q1 ,p3 ); \
-    s2 = _mm512_load_si512(&s3);\
+    s1 = _mm512_unpacklo_epi64( s1, r3 ); \
-    q2 = _mm512_load_si512(&q3);\
+    q1 = _mm512_unpacklo_epi64( q1, p3 ); \
-    r3 = _mm512_load_si512(&r1);\
+    s2 = _mm512_unpackhi_epi64( s3, r1 ); \
-    p3 = _mm512_load_si512(&p1);\
+    q2 = _mm512_unpackhi_epi64( q3, p1 ); \
-    r1 = _mm512_unpacklo_epi32(r1,r0);\
+    s3 = _mm512_unpacklo_epi64( s3, r1 ); \
-    p1 = _mm512_unpacklo_epi32(p1,p0);\
+    q3 = _mm512_unpacklo_epi64( q3, p1 );
-    r3 = _mm512_unpackhi_epi32(r3,r0);\
-    p3 = _mm512_unpackhi_epi32(p3,p0);\
-    s0 = _mm512_unpackhi_epi64(s0,r3);\
-    q0 = _mm512_unpackhi_epi64(q0,p3);\
-    s1 = _mm512_unpacklo_epi64(s1,r3);\
-    q1 = _mm512_unpacklo_epi64(q1,p3);\
-    s2 = _mm512_unpackhi_epi64(s2,r1);\
-    q2 = _mm512_unpackhi_epi64(q2,p1);\
-    s3 = _mm512_unpacklo_epi64(s3,r1);\
-    q3 = _mm512_unpacklo_epi64(q3,p1);
 
 #define MIXTON10244W(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3)\
     NMLTOM10244W(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3);
 
-void rnd512_4way( luffa_4way_context *state, __m512i *msg )
+void rnd512_4way( luffa_4way_context *state, const __m512i *msg )
 {
     __m512i t0, t1;
     __m512i *chainv = state->chainv;
-    __m512i msg0, msg1;
     __m512i x0, x1, x2, x3, x4, x5, x6, x7;
 
     t0 = mm512_xor3( chainv[0], chainv[2], chainv[4] );
@@ -168,9 +157,6 @@ void rnd512_4way( luffa_4way_context *state, __m512i *msg )
 
     MULT24W( t0, t1 );
 
-    msg0 = _mm512_shuffle_epi32( msg[0], 27 );
-    msg1 = _mm512_shuffle_epi32( msg[1], 27 );
-
     chainv[0] = _mm512_xor_si512( chainv[0], t0 );
     chainv[1] = _mm512_xor_si512( chainv[1], t1 );
     chainv[2] = _mm512_xor_si512( chainv[2], t0 );
@@ -202,11 +188,8 @@ void rnd512_4way( luffa_4way_context *state, __m512i *msg )
     chainv[7] = _mm512_xor_si512(chainv[7], chainv[9]);
 
     MULT24W( chainv[8], chainv[9] );
-    chainv[8] = _mm512_xor_si512( chainv[8], t0 );
+    t0 = chainv[8] = _mm512_xor_si512( chainv[8], t0 );
-    chainv[9] = _mm512_xor_si512( chainv[9], t1 );
+    t1 = chainv[9] = _mm512_xor_si512( chainv[9], t1 );
-
-    t0 = chainv[8];
-    t1 = chainv[9];
 
     MULT24W( chainv[8], chainv[9] );
     chainv[8] = _mm512_xor_si512( chainv[8], chainv[6] );
@@ -225,27 +208,36 @@ void rnd512_4way( luffa_4way_context *state, __m512i *msg )
     chainv[3] = _mm512_xor_si512( chainv[3], chainv[1] );
 
     MULT24W( chainv[0], chainv[1] );
-    chainv[0] = mm512_xor3( chainv[0], t0, msg0 );
+    chainv[0] = _mm512_xor_si512( chainv[0], t0 );
-    chainv[1] = mm512_xor3( chainv[1], t1, msg1 );
+    chainv[1] = _mm512_xor_si512( chainv[1], t1 );
 
-    MULT24W( msg0, msg1 );
+    if ( msg )
-    chainv[2] = _mm512_xor_si512( chainv[2], msg0 );
+    {
-    chainv[3] = _mm512_xor_si512( chainv[3], msg1 );
+       __m512i msg0, msg1;
 
-    MULT24W( msg0, msg1 );
+       msg0 = _mm512_shuffle_epi32( msg[0], 27 );
-    chainv[4] = _mm512_xor_si512( chainv[4], msg0 );
+       msg1 = _mm512_shuffle_epi32( msg[1], 27 );
-    chainv[5] = _mm512_xor_si512( chainv[5], msg1 );
 
-    MULT24W( msg0, msg1 );
+       chainv[0] = _mm512_xor_si512( chainv[0], msg0 );
-    chainv[6] = _mm512_xor_si512( chainv[6], msg0 );
+       chainv[1] = _mm512_xor_si512( chainv[1], msg1 );
-    chainv[7] = _mm512_xor_si512( chainv[7], msg1 );
 
-    MULT24W( msg0, msg1);
+       MULT24W( msg0, msg1 );
-    chainv[8] = _mm512_xor_si512( chainv[8], msg0 );
+       chainv[2] = _mm512_xor_si512( chainv[2], msg0 );
-    chainv[9] = _mm512_xor_si512( chainv[9], msg1 );
+       chainv[3] = _mm512_xor_si512( chainv[3], msg1 );
 
-    MULT24W( msg0, msg1 );
+       MULT24W( msg0, msg1 );
+       chainv[4] = _mm512_xor_si512( chainv[4], msg0 );
+       chainv[5] = _mm512_xor_si512( chainv[5], msg1 );
 
+       MULT24W( msg0, msg1 );
+       chainv[6] = _mm512_xor_si512( chainv[6], msg0 );
+       chainv[7] = _mm512_xor_si512( chainv[7], msg1 );
+
+       MULT24W( msg0, msg1);
+       chainv[8] = _mm512_xor_si512( chainv[8], msg0 );
+       chainv[9] = _mm512_xor_si512( chainv[9], msg1 );
+    }
+    
     chainv[3] = _mm512_rol_epi32( chainv[3], 1 );
     chainv[5] = _mm512_rol_epi32( chainv[5], 2 );
     chainv[7] = _mm512_rol_epi32( chainv[7], 3 );
@@ -282,16 +274,11 @@ void finalization512_4way( luffa_4way_context *state, uint32 *b )
     uint32_t hash[8*4] __attribute((aligned(128)));
     __m512i* chainv = state->chainv;
     __m512i t[2];
-    __m512i zero[2];
+    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x( 
-    zero[0] = zero[1] = m512_zero;
+                                  0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
-    const __m512i shuff_bswap32 = m512_const_64(
-                                  0x3c3d3e3f38393a3b, 0x3435363730313233,
-                                  0x2c2d2e2f28292a2b, 0x2425262720212223,
-                                  0x1c1d1e1f18191a1b, 0x1415161710111213,
-                                  0x0c0d0e0f08090a0b, 0x0405060700010203 );
 
     /*---- blank round with m=0 ----*/
-    rnd512_4way( state, zero );
+    rnd512_4way( state, NULL );
     
     t[0] = mm512_xor3( chainv[0], chainv[2], chainv[4] );
     t[1] = mm512_xor3( chainv[1], chainv[3], chainv[5] );
@@ -300,37 +287,30 @@ void finalization512_4way( luffa_4way_context *state, uint32 *b )
     t[0] = _mm512_shuffle_epi32( t[0], 27 );
     t[1] = _mm512_shuffle_epi32( t[1], 27 );
 
-    _mm512_store_si512( (__m512i*)&hash[0], t[0] );
+    _mm512_store_si512( (__m512i*)&hash[ 0], t[0] );
     _mm512_store_si512( (__m512i*)&hash[16], t[1] );
 
-    casti_m512i( b, 0 ) = _mm512_shuffle_epi8(
+    casti_m512i( b,0 ) = _mm512_shuffle_epi8(
-                                  casti_m512i( hash, 0 ), shuff_bswap32 );
+                                  casti_m512i( hash,0 ), shuff_bswap32 );
-    casti_m512i( b, 1 ) = _mm512_shuffle_epi8(
+    casti_m512i( b,1 ) = _mm512_shuffle_epi8(
-                                  casti_m512i( hash, 1 ), shuff_bswap32 );
+                                  casti_m512i( hash,1 ), shuff_bswap32 );
 
-    rnd512_4way( state, zero );
+    rnd512_4way( state, NULL );
-
-    t[0] = chainv[0];
-    t[1] = chainv[1];
-    t[0] = _mm512_xor_si512( t[0], chainv[2] );
-    t[1] = _mm512_xor_si512( t[1], chainv[3] );
-    t[0] = _mm512_xor_si512( t[0], chainv[4] );
-    t[1] = _mm512_xor_si512( t[1], chainv[5] );
-    t[0] = _mm512_xor_si512( t[0], chainv[6] );
-    t[1] = _mm512_xor_si512( t[1], chainv[7] );
-    t[0] = _mm512_xor_si512( t[0], chainv[8] );
-    t[1] = _mm512_xor_si512( t[1], chainv[9] );
 
+    t[0] = mm512_xor3( chainv[0], chainv[2], chainv[4] );
+    t[1] = mm512_xor3( chainv[1], chainv[3], chainv[5] );
+    t[0] = mm512_xor3( t[0], chainv[6], chainv[8] );
+    t[1] = mm512_xor3( t[1], chainv[7], chainv[9] );
     t[0] = _mm512_shuffle_epi32( t[0], 27 );
     t[1] = _mm512_shuffle_epi32( t[1], 27 );
 
-    _mm512_store_si512( (__m512i*)&hash[0], t[0] );
+    _mm512_store_si512( (__m512i*)&hash[ 0], t[0] );
     _mm512_store_si512( (__m512i*)&hash[16], t[1] );
 
-    casti_m512i( b, 2 ) = _mm512_shuffle_epi8(
+    casti_m512i( b,2 ) = _mm512_shuffle_epi8(
-                                  casti_m512i( hash, 0 ), shuff_bswap32 );
+                                  casti_m512i( hash,0 ), shuff_bswap32 );
-    casti_m512i( b, 3 ) = _mm512_shuffle_epi8(
+    casti_m512i( b,3 ) = _mm512_shuffle_epi8(
-                                  casti_m512i( hash, 1 ), shuff_bswap32 );
+                                  casti_m512i( hash,1 ), shuff_bswap32 );
 }
 
 int luffa_4way_init( luffa_4way_context *state, int hashbitlen )
@@ -338,16 +318,16 @@ int luffa_4way_init( luffa_4way_context *state, int hashbitlen )
     state->hashbitlen = hashbitlen;
     __m128i *iv = (__m128i*)IV;
 
-    state->chainv[0] = m512_const1_128( iv[0] );
+    state->chainv[0] = mm512_bcast_m128( iv[0] );
-    state->chainv[1] = m512_const1_128( iv[1] );
+    state->chainv[1] = mm512_bcast_m128( iv[1] );
-    state->chainv[2] = m512_const1_128( iv[2] );
+    state->chainv[2] = mm512_bcast_m128( iv[2] );
-    state->chainv[3] = m512_const1_128( iv[3] );
+    state->chainv[3] = mm512_bcast_m128( iv[3] );
-    state->chainv[4] = m512_const1_128( iv[4] );
+    state->chainv[4] = mm512_bcast_m128( iv[4] );
-    state->chainv[5] = m512_const1_128( iv[5] );
+    state->chainv[5] = mm512_bcast_m128( iv[5] );
-    state->chainv[6] = m512_const1_128( iv[6] );
+    state->chainv[6] = mm512_bcast_m128( iv[6] );
-    state->chainv[7] = m512_const1_128( iv[7] );
+    state->chainv[7] = mm512_bcast_m128( iv[7] );
-    state->chainv[8] = m512_const1_128( iv[8] );
+    state->chainv[8] = mm512_bcast_m128( iv[8] );
-    state->chainv[9] = m512_const1_128( iv[9] );
+    state->chainv[9] = mm512_bcast_m128( iv[9] );
 
     ((__m512i*)state->buffer)[0] = m512_zero;
     ((__m512i*)state->buffer)[1] = m512_zero;
@@ -370,11 +350,8 @@ int luffa_4way_update( luffa_4way_context *state, const void *data,
     __m512i msg[2];
     int i;
     int blocks = (int)len >> 5;
-    const __m512i shuff_bswap32 = m512_const_64( 
+    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x(  
-                                   0x3c3d3e3f38393a3b, 0x3435363730313233,
+                                   0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
-                                   0x2c2d2e2f28292a2b, 0x2425262720212223,
-                                   0x1c1d1e1f18191a1b, 0x1415161710111213,
-                                   0x0c0d0e0f08090a0b, 0x0405060700010203 );
 
     state->rembytes = (int)len & 0x1F;
 
@@ -392,7 +369,7 @@ int luffa_4way_update( luffa_4way_context *state, const void *data,
     {
       // remaining data bytes
       buffer[0] = _mm512_shuffle_epi8( vdata[0], shuff_bswap32 );
-      buffer[1] = m512_const1_i128(  0x0000000080000000 );
+      buffer[1] = mm512_bcast128lo_64( 0x0000000080000000 );
     }
     return 0;
 }
@@ -416,7 +393,7 @@ int luffa_4way_close( luffa_4way_context *state, void *hashval )
       rnd512_4way( state, buffer );
     else
     {     // empty pad block, constant data
-      msg[0] = m512_const1_i128(  0x0000000080000000 );
+      msg[0] = mm512_bcast128lo_64( 0x0000000080000000 );
       msg[1] = m512_zero;
       rnd512_4way( state, msg );
     }
@@ -440,16 +417,16 @@ int luffa512_4way_full( luffa_4way_context *state, void *output,
     state->hashbitlen = 512;
     __m128i *iv = (__m128i*)IV;
 
-    state->chainv[0] = m512_const1_128( iv[0] );
+    state->chainv[0] = mm512_bcast_m128( iv[0] );
-    state->chainv[1] = m512_const1_128( iv[1] );
+    state->chainv[1] = mm512_bcast_m128( iv[1] );
-    state->chainv[2] = m512_const1_128( iv[2] );
+    state->chainv[2] = mm512_bcast_m128( iv[2] );
-    state->chainv[3] = m512_const1_128( iv[3] );
+    state->chainv[3] = mm512_bcast_m128( iv[3] );
-    state->chainv[4] = m512_const1_128( iv[4] );
+    state->chainv[4] = mm512_bcast_m128( iv[4] );
-    state->chainv[5] = m512_const1_128( iv[5] );
+    state->chainv[5] = mm512_bcast_m128( iv[5] );
-    state->chainv[6] = m512_const1_128( iv[6] );
+    state->chainv[6] = mm512_bcast_m128( iv[6] );
-    state->chainv[7] = m512_const1_128( iv[7] );
+    state->chainv[7] = mm512_bcast_m128( iv[7] );
-    state->chainv[8] = m512_const1_128( iv[8] );
+    state->chainv[8] = mm512_bcast_m128( iv[8] );
-    state->chainv[9] = m512_const1_128( iv[9] );
+    state->chainv[9] = mm512_bcast_m128( iv[9] );
 
     ((__m512i*)state->buffer)[0] = m512_zero;
     ((__m512i*)state->buffer)[1] = m512_zero;
@@ -458,11 +435,8 @@ int luffa512_4way_full( luffa_4way_context *state, void *output,
     __m512i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m512i shuff_bswap32 = m512_const_64(
+    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x( 
-                                   0x3c3d3e3f38393a3b, 0x3435363730313233,
+                                   0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
-                                   0x2c2d2e2f28292a2b, 0x2425262720212223,
-                                   0x1c1d1e1f18191a1b, 0x1415161710111213,
-                                   0x0c0d0e0f08090a0b, 0x0405060700010203 );
 
     state->rembytes = inlen & 0x1F;
 
@@ -479,13 +453,13 @@ int luffa512_4way_full( luffa_4way_context *state, void *output,
     {
        // padding of partial block
        msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
-       msg[1] = m512_const1_i128(  0x0000000080000000 );
+       msg[1] = mm512_bcast128lo_64( 0x0000000080000000 );
        rnd512_4way( state, msg );
     }
     else
     {
        // empty pad block
-       msg[0] = m512_const1_i128( 0x0000000080000000 );
+       msg[0] = mm512_bcast128lo_64( 0x0000000080000000 );
        msg[1] = m512_zero;
        rnd512_4way( state, msg );
     }
@@ -506,11 +480,8 @@ int luffa_4way_update_close( luffa_4way_context *state,
     __m512i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m512i shuff_bswap32 = m512_const_64(
+    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x( 
-                                   0x3c3d3e3f38393a3b, 0x3435363730313233,
+                                   0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
-                                   0x2c2d2e2f28292a2b, 0x2425262720212223,
-                                   0x1c1d1e1f18191a1b, 0x1415161710111213,
-                                   0x0c0d0e0f08090a0b, 0x0405060700010203 );
 
     state->rembytes = inlen & 0x1F;
 
@@ -527,13 +498,13 @@ int luffa_4way_update_close( luffa_4way_context *state,
     {
        // padding of partial block
        msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
-       msg[1] = m512_const1_i128( 0x0000000080000000 );
+       msg[1] = mm512_bcast128lo_64( 0x0000000080000000 );
        rnd512_4way( state, msg );
     }
     else
     {
        // empty pad block
-       msg[0] = m512_const1_i128( 0x0000000080000000 );
+       msg[0] = mm512_bcast128lo_64( 0x0000000080000000 );
        msg[1] = m512_zero;
        rnd512_4way( state, msg );
     }
@@ -548,26 +519,45 @@ int luffa_4way_update_close( luffa_4way_context *state,
 
 #endif // AVX512
 
-#define cns(i)  m256_const1_128( ( (__m128i*)CNS_INIT)[i] )
+#define cns(i)  mm256_bcast_m128( ( (__m128i*)CNS_INIT)[i] )
 
 #define ADD_CONSTANT( a, b, c0, c1 ) \
     a = _mm256_xor_si256( a, c0 ); \
     b = _mm256_xor_si256( b, c1 );
 
-/*
+//TODO Enable for AVX10_256, not used with AVX512 or AVX10_512
-#define MULT2( a0, a1, mask ) \
+#if defined(__AVX512VL__) 
-do { \
-  __m256i b = _mm256_xor_si256( a0, \
-                   _mm256_shuffle_epi32( _mm256_and_si256(a1,mask), 16 ) ); \
-  a0 = _mm256_or_si256( _mm256_srli_si256(b,4), _mm256_slli_si256(a1,12) ); \
-  a1 = _mm256_or_si256( _mm256_srli_si256(a1,4), _mm256_slli_si256(b,12) );  \
-} while(0)
-*/
 
-#define MULT2( a0, a1, mask ) \
+#define MULT2( a0, a1 ) \
 { \
   __m256i b = _mm256_xor_si256( a0, \
-                 _mm256_shuffle_epi32( _mm256_and_si256( a1, mask ), 16 ) ); \
+                     _mm256_maskz_shuffle_epi32( 0xbb, a1, 0x10 ) ); \
+  a0 = _mm256_alignr_epi8( a1,  b, 4 ); \
+  a1 = _mm256_alignr_epi8(  b, a1, 4 ); \
+}
+
+#define SUBCRUMB( a0, a1, a2, a3 ) \
+{ \
+    __m256i t = a0; \
+    a0 = mm256_xoror( a3, a0, a1 ); \
+    a2 = _mm256_xor_si256( a2, a3 ); \
+    a1 = _mm256_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 xnor (a3 & t) */ \
+    a3 = mm256_xorand( a2, a3, t ); \
+    a2 = mm256_xorand( a1, a2, a0); \
+    a1 = _mm256_or_si256( a1, a3 ); \
+    a3 = _mm256_xor_si256( a3, a2 ); \
+    t  = _mm256_xor_si256( t, a1 ); \
+    a2 = _mm256_and_si256( a2, a1 ); \
+    a1 = mm256_xnor( a1, a0 ); \
+    a0 = t; \
+}
+
+#else
+
+#define MULT2( a0, a1 ) \
+{ \
+  __m256i b = _mm256_xor_si256( a0, _mm256_shuffle_epi32( \
+                         _mm256_blend_epi32( a1, m256_zero, 0xee ), 0x10 ) ); \
   a0 = _mm256_alignr_epi8( a1,  b, 4 ); \
   a1 = _mm256_alignr_epi8(  b, a1, 4 ); \
 }
@@ -593,26 +583,14 @@ do { \
     a0 = t; \
 }
 
+#endif
+
 #define MIXWORD( a, b ) \
-{ \
+    b = _mm256_xor_si256( a, b ); \
-    __m256i t1, t2; \
+    a = _mm256_xor_si256( b, mm256_rol_32( a,  2 ) ); \
-    b  = _mm256_xor_si256( a,b ); \
+    b = _mm256_xor_si256( a, mm256_rol_32( b, 14 ) ); \
-    t1 = _mm256_slli_epi32( a,  2 ); \
+    a = _mm256_xor_si256( b, mm256_rol_32( a, 10 ) ); \
-    t2 = _mm256_srli_epi32( a, 30 ); \
+    b = mm256_rol_32( b, 1 );
-    a  = _mm256_or_si256( t1, t2 ); \
-    a  = _mm256_xor_si256( a, b ); \
-    t1 = _mm256_slli_epi32( b, 14 ); \
-    t2 = _mm256_srli_epi32( b, 18 ); \
-    b  = _mm256_or_si256( t1, t2 ); \
-    b  = _mm256_xor_si256( a, b ); \
-    t1 = _mm256_slli_epi32( a, 10 ); \
-    t2 = _mm256_srli_epi32( a, 22 ); \
-    a  = _mm256_or_si256( t1,t2 ); \
-    a  = _mm256_xor_si256( a,b ); \
-    t1 = _mm256_slli_epi32( b,1 ); \
-    t2 = _mm256_srli_epi32( b,31 ); \
-    b  = _mm256_or_si256( t1, t2 ); \
-}
 
 #define STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, c0, c1 ) \
     SUBCRUMB( x0, x1, x2, x3 ); \
@@ -624,49 +602,37 @@ do { \
     ADD_CONSTANT( x0, x4, c0, c1 );
 
 #define STEP_PART2( a0, a1, t0, t1, c0, c1 ) \
-    a1 = _mm256_shuffle_epi32( a1, 147); \
+    t0 = _mm256_shuffle_epi32( a1, 147 ); \
-    t0 = _mm256_load_si256( &a1 ); \
+    a1 = _mm256_unpacklo_epi32( t0, a0 ); \
-    a1 = _mm256_unpacklo_epi32( a1, a0 ); \
     t0 = _mm256_unpackhi_epi32( t0, a0 ); \
     t1 = _mm256_shuffle_epi32( t0, 78 ); \
     a0 = _mm256_shuffle_epi32( a1, 78 ); \
-    SUBCRUMB( t1, t0, a0, a1 );\
+    SUBCRUMB( t1, t0, a0, a1 ); \
     t0 = _mm256_unpacklo_epi32( t0, t1 ); \
     a1 = _mm256_unpacklo_epi32( a1, a0 ); \
-    a0 = _mm256_load_si256( &a1 ); \
+    a0 = _mm256_unpackhi_epi64( a1, t0 ); \
-    a0 = _mm256_unpackhi_epi64( a0, t0 ); \
     a1 = _mm256_unpacklo_epi64( a1, t0 ); \
     a1 = _mm256_shuffle_epi32( a1, 57 ); \
     MIXWORD( a0, a1 ); \
     ADD_CONSTANT( a0, a1, c0, c1 );
 
 #define NMLTOM1024(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3)\
-    s1 = _mm256_load_si256(&r3);\
+    s1 = _mm256_unpackhi_epi32( r3, r2 ); \
-    q1 = _mm256_load_si256(&p3);\
+    q1 = _mm256_unpackhi_epi32( p3, p2 ); \
-    s3 = _mm256_load_si256(&r3);\
+    s3 = _mm256_unpacklo_epi32( r3, r2 ); \
-    q3 = _mm256_load_si256(&p3);\
+    q3 = _mm256_unpacklo_epi32( p3, p2 ); \
-    s1 = _mm256_unpackhi_epi32(s1,r2);\
+    r3 = _mm256_unpackhi_epi32( r1, r0 ); \
-    q1 = _mm256_unpackhi_epi32(q1,p2);\
+    r1 = _mm256_unpacklo_epi32( r1, r0 ); \
-    s3 = _mm256_unpacklo_epi32(s3,r2);\
+    p3 = _mm256_unpackhi_epi32( p1, p0 ); \
-    q3 = _mm256_unpacklo_epi32(q3,p2);\
+    p1 = _mm256_unpacklo_epi32( p1, p0 ); \
-    s0 = _mm256_load_si256(&s1);\
+    s0 = _mm256_unpackhi_epi64( s1, r3 ); \
-    q0 = _mm256_load_si256(&q1);\
+    q0 = _mm256_unpackhi_epi64( q1 ,p3 ); \
-    s2 = _mm256_load_si256(&s3);\
+    s1 = _mm256_unpacklo_epi64( s1, r3 ); \
-    q2 = _mm256_load_si256(&q3);\
+    q1 = _mm256_unpacklo_epi64( q1, p3 ); \
-    r3 = _mm256_load_si256(&r1);\
+    s2 = _mm256_unpackhi_epi64( s3, r1 ); \
-    p3 = _mm256_load_si256(&p1);\
+    q2 = _mm256_unpackhi_epi64( q3, p1 ); \
-    r1 = _mm256_unpacklo_epi32(r1,r0);\
+    s3 = _mm256_unpacklo_epi64( s3, r1 ); \
-    p1 = _mm256_unpacklo_epi32(p1,p0);\
+    q3 = _mm256_unpacklo_epi64( q3, p1 );
-    r3 = _mm256_unpackhi_epi32(r3,r0);\
-    p3 = _mm256_unpackhi_epi32(p3,p0);\
-    s0 = _mm256_unpackhi_epi64(s0,r3);\
-    q0 = _mm256_unpackhi_epi64(q0,p3);\
-    s1 = _mm256_unpacklo_epi64(s1,r3);\
-    q1 = _mm256_unpacklo_epi64(q1,p3);\
-    s2 = _mm256_unpackhi_epi64(s2,r1);\
-    q2 = _mm256_unpackhi_epi64(q2,p1);\
-    s3 = _mm256_unpacklo_epi64(s3,r1);\
-    q3 = _mm256_unpacklo_epi64(q3,p1);
 
 #define MIXTON1024(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3)\
     NMLTOM1024(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3);
@@ -676,30 +642,18 @@ do { \
 /* Round function         */
 /* state: hash context    */
 
-void rnd512_2way( luffa_2way_context *state, __m256i *msg )
+void rnd512_2way( luffa_2way_context *state, const __m256i *msg )
 {
     __m256i t0, t1;
     __m256i *chainv = state->chainv;
-    __m256i msg0, msg1;
     __m256i x0, x1, x2, x3, x4, x5, x6, x7;
-    const __m256i MASK = m256_const1_i128( 0xffffffff );
 
-    t0 = chainv[0];
+    t0 = mm256_xor3( chainv[0], chainv[2], chainv[4] );
-    t1 = chainv[1];
+    t1 = mm256_xor3( chainv[1], chainv[3], chainv[5] );
+    t0 = mm256_xor3( t0, chainv[6], chainv[8] );
+    t1 = mm256_xor3( t1, chainv[7], chainv[9] );
 
-    t0 = _mm256_xor_si256( t0, chainv[2] );
+    MULT2( t0, t1 );
-    t1 = _mm256_xor_si256( t1, chainv[3] );
-    t0 = _mm256_xor_si256( t0, chainv[4] );
-    t1 = _mm256_xor_si256( t1, chainv[5] );
-    t0 = _mm256_xor_si256( t0, chainv[6] );
-    t1 = _mm256_xor_si256( t1, chainv[7] );
-    t0 = _mm256_xor_si256( t0, chainv[8] );
-    t1 = _mm256_xor_si256( t1, chainv[9] );
-
-    MULT2( t0, t1, MASK );
-
-    msg0 = _mm256_shuffle_epi32( msg[0], 27 );
-    msg1 = _mm256_shuffle_epi32( msg[1], 27 );
 
     chainv[0] = _mm256_xor_si256( chainv[0], t0 );
     chainv[1] = _mm256_xor_si256( chainv[1], t1 );
@@ -715,66 +669,72 @@ void rnd512_2way( luffa_2way_context *state, __m256i *msg )
     t0 = chainv[0];
     t1 = chainv[1];
 
-    MULT2( chainv[0], chainv[1], MASK );
+    MULT2( chainv[0], chainv[1] );
     chainv[0] = _mm256_xor_si256( chainv[0], chainv[2] );
     chainv[1] = _mm256_xor_si256( chainv[1], chainv[3] );
 
-    MULT2( chainv[2], chainv[3], MASK );
+    MULT2( chainv[2], chainv[3] );
     chainv[2] = _mm256_xor_si256(chainv[2], chainv[4]);
     chainv[3] = _mm256_xor_si256(chainv[3], chainv[5]);
 
-    MULT2( chainv[4], chainv[5], MASK );
+    MULT2( chainv[4], chainv[5] );
     chainv[4] = _mm256_xor_si256(chainv[4], chainv[6]);
     chainv[5] = _mm256_xor_si256(chainv[5], chainv[7]);
 
-    MULT2( chainv[6], chainv[7], MASK );
+    MULT2( chainv[6], chainv[7] );
     chainv[6] = _mm256_xor_si256(chainv[6], chainv[8]);
     chainv[7] = _mm256_xor_si256(chainv[7], chainv[9]);
 
-    MULT2( chainv[8], chainv[9], MASK );
+    MULT2( chainv[8], chainv[9] );
-    chainv[8] = _mm256_xor_si256( chainv[8], t0 );
+    t0 = chainv[8] = _mm256_xor_si256( chainv[8], t0 );
-    chainv[9] = _mm256_xor_si256( chainv[9], t1 );
+    t1 = chainv[9] = _mm256_xor_si256( chainv[9], t1 );
 
-    t0 = chainv[8];
+    MULT2( chainv[8], chainv[9] );
-    t1 = chainv[9];
-
-    MULT2( chainv[8], chainv[9], MASK );
     chainv[8] = _mm256_xor_si256( chainv[8], chainv[6] );
     chainv[9] = _mm256_xor_si256( chainv[9], chainv[7] );
 
-    MULT2( chainv[6], chainv[7], MASK );
+    MULT2( chainv[6], chainv[7] );
     chainv[6] = _mm256_xor_si256( chainv[6], chainv[4] );
     chainv[7] = _mm256_xor_si256( chainv[7], chainv[5] );
 
-    MULT2( chainv[4], chainv[5], MASK );
+    MULT2( chainv[4], chainv[5] );
     chainv[4] = _mm256_xor_si256( chainv[4], chainv[2] );
     chainv[5] = _mm256_xor_si256( chainv[5], chainv[3] );
 
-    MULT2( chainv[2], chainv[3], MASK );
+    MULT2( chainv[2], chainv[3] );
     chainv[2] = _mm256_xor_si256( chainv[2], chainv[0] );
     chainv[3] = _mm256_xor_si256( chainv[3], chainv[1] );
 
-    MULT2( chainv[0], chainv[1], MASK );
+    MULT2( chainv[0], chainv[1] );
-    chainv[0] = _mm256_xor_si256( _mm256_xor_si256( chainv[0], t0 ), msg0 );
+    chainv[0] = _mm256_xor_si256( chainv[0], t0 );
-    chainv[1] = _mm256_xor_si256( _mm256_xor_si256( chainv[1], t1 ), msg1 );
+    chainv[1] = _mm256_xor_si256( chainv[1], t1 );
 
-    MULT2( msg0, msg1, MASK );
+    if ( msg )
-    chainv[2] = _mm256_xor_si256( chainv[2], msg0 );
+    {
-    chainv[3] = _mm256_xor_si256( chainv[3], msg1 );
+       __m256i msg0, msg1;
+    
+       msg0 = _mm256_shuffle_epi32( msg[0], 27 );
+       msg1 = _mm256_shuffle_epi32( msg[1], 27 );
 
-    MULT2( msg0, msg1, MASK );
+       chainv[0] = _mm256_xor_si256( chainv[0], msg0 );
-    chainv[4] = _mm256_xor_si256( chainv[4], msg0 );
+       chainv[1] = _mm256_xor_si256( chainv[1], msg1 );
-    chainv[5] = _mm256_xor_si256( chainv[5], msg1 );
+    
+       MULT2( msg0, msg1 );
+       chainv[2] = _mm256_xor_si256( chainv[2], msg0 );
+       chainv[3] = _mm256_xor_si256( chainv[3], msg1 );
 
-    MULT2( msg0, msg1, MASK );
+       MULT2( msg0, msg1 );
-    chainv[6] = _mm256_xor_si256( chainv[6], msg0 );
+       chainv[4] = _mm256_xor_si256( chainv[4], msg0 );
-    chainv[7] = _mm256_xor_si256( chainv[7], msg1 );
+       chainv[5] = _mm256_xor_si256( chainv[5], msg1 );
 
-    MULT2( msg0, msg1, MASK );
+       MULT2( msg0, msg1 );
-    chainv[8] = _mm256_xor_si256( chainv[8], msg0 );
+       chainv[6] = _mm256_xor_si256( chainv[6], msg0 );
-    chainv[9] = _mm256_xor_si256( chainv[9], msg1 );
+       chainv[7] = _mm256_xor_si256( chainv[7], msg1 );
 
-    MULT2( msg0, msg1, MASK );
+       MULT2( msg0, msg1 );
+       chainv[8] = _mm256_xor_si256( chainv[8], msg0 );
+       chainv[9] = _mm256_xor_si256( chainv[9], msg1 );
+    }
 
     chainv[3] = mm256_rol_32( chainv[3], 1 );
     chainv[5] = mm256_rol_32( chainv[5], 2 );
@@ -816,57 +776,40 @@ void finalization512_2way( luffa_2way_context *state, uint32 *b )
 {
     uint32 hash[8*2] __attribute((aligned(64)));
     __m256i* chainv = state->chainv;
-    __m256i t[2];
+    __m256i t0, t1;
-    __m256i zero[2];
+    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-    zero[0] = zero[1] = m256_zero;
-    const __m256i shuff_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b,
-                                                 0x1415161710111213,
-                                                 0x0c0d0e0f08090a0b,
                                                  0x0405060700010203 );
     /*---- blank round with m=0 ----*/
-    rnd512_2way( state, zero );
+    rnd512_2way( state, NULL );
 
-    t[0] = chainv[0];
+    t0 = mm256_xor3( chainv[0], chainv[2], chainv[4] );
-    t[1] = chainv[1];
+    t1 = mm256_xor3( chainv[1], chainv[3], chainv[5] );
+    t0 = mm256_xor3( t0, chainv[6], chainv[8] );
+    t1 = mm256_xor3( t1, chainv[7], chainv[9] );
 
-    t[0] = _mm256_xor_si256( t[0], chainv[2] );
+    t0 = _mm256_shuffle_epi32( t0, 27 );
-    t[1] = _mm256_xor_si256( t[1], chainv[3] );
+    t1 = _mm256_shuffle_epi32( t1, 27 );
-    t[0] = _mm256_xor_si256( t[0], chainv[4] );
-    t[1] = _mm256_xor_si256( t[1], chainv[5] );
-    t[0] = _mm256_xor_si256( t[0], chainv[6] );
-    t[1] = _mm256_xor_si256( t[1], chainv[7] );
-    t[0] = _mm256_xor_si256( t[0], chainv[8] );
-    t[1] = _mm256_xor_si256( t[1], chainv[9] );
 
-    t[0] = _mm256_shuffle_epi32( t[0], 27 );
+    _mm256_store_si256( (__m256i*)&hash[0], t0 );
-    t[1] = _mm256_shuffle_epi32( t[1], 27 );
+    _mm256_store_si256( (__m256i*)&hash[8], t1 );
-
-    _mm256_store_si256( (__m256i*)&hash[0], t[0] );
-    _mm256_store_si256( (__m256i*)&hash[8], t[1] );
 
     casti_m256i( b, 0 ) = _mm256_shuffle_epi8(
                                   casti_m256i( hash, 0 ), shuff_bswap32 );
     casti_m256i( b, 1 ) = _mm256_shuffle_epi8( 
                                   casti_m256i( hash, 1 ), shuff_bswap32 );
 
-    rnd512_2way( state, zero );
+    rnd512_2way( state, NULL );
 
-    t[0] = chainv[0];
+    t0 = mm256_xor3( chainv[0], chainv[2], chainv[4] );
-    t[1] = chainv[1];
+    t1 = mm256_xor3( chainv[1], chainv[3], chainv[5] );
-    t[0] = _mm256_xor_si256( t[0], chainv[2] );
+    t0 = mm256_xor3( t0, chainv[6], chainv[8] );
-    t[1] = _mm256_xor_si256( t[1], chainv[3] );
+    t1 = mm256_xor3( t1, chainv[7], chainv[9] );
-    t[0] = _mm256_xor_si256( t[0], chainv[4] );
+    
-    t[1] = _mm256_xor_si256( t[1], chainv[5] );
+    t0 = _mm256_shuffle_epi32( t0, 27 );
-    t[0] = _mm256_xor_si256( t[0], chainv[6] );
+    t1 = _mm256_shuffle_epi32( t1, 27 );
-    t[1] = _mm256_xor_si256( t[1], chainv[7] );
-    t[0] = _mm256_xor_si256( t[0], chainv[8] );
-    t[1] = _mm256_xor_si256( t[1], chainv[9] );
 
-    t[0] = _mm256_shuffle_epi32( t[0], 27 );
+    _mm256_store_si256( (__m256i*)&hash[0], t0 );
-    t[1] = _mm256_shuffle_epi32( t[1], 27 );
+    _mm256_store_si256( (__m256i*)&hash[8], t1 );
-
-    _mm256_store_si256( (__m256i*)&hash[0], t[0] );
-    _mm256_store_si256( (__m256i*)&hash[8], t[1] );
 
     casti_m256i( b, 2 ) = _mm256_shuffle_epi8( 
                                   casti_m256i( hash, 0 ), shuff_bswap32 );
@@ -879,16 +822,16 @@ int luffa_2way_init( luffa_2way_context *state, int hashbitlen )
     state->hashbitlen = hashbitlen;
     __m128i *iv = (__m128i*)IV;
     
-    state->chainv[0] = m256_const1_128( iv[0] );
+    state->chainv[0] = mm256_bcast_m128( iv[0] );
-    state->chainv[1] = m256_const1_128( iv[1] );
+    state->chainv[1] = mm256_bcast_m128( iv[1] );
-    state->chainv[2] = m256_const1_128( iv[2] );
+    state->chainv[2] = mm256_bcast_m128( iv[2] );
-    state->chainv[3] = m256_const1_128( iv[3] );
+    state->chainv[3] = mm256_bcast_m128( iv[3] );
-    state->chainv[4] = m256_const1_128( iv[4] );
+    state->chainv[4] = mm256_bcast_m128( iv[4] );
-    state->chainv[5] = m256_const1_128( iv[5] );
+    state->chainv[5] = mm256_bcast_m128( iv[5] );
-    state->chainv[6] = m256_const1_128( iv[6] );
+    state->chainv[6] = mm256_bcast_m128( iv[6] );
-    state->chainv[7] = m256_const1_128( iv[7] );
+    state->chainv[7] = mm256_bcast_m128( iv[7] );
-    state->chainv[8] = m256_const1_128( iv[8] );
+    state->chainv[8] = mm256_bcast_m128( iv[8] );
-    state->chainv[9] = m256_const1_128( iv[9] );
+    state->chainv[9] = mm256_bcast_m128( iv[9] );
 
     ((__m256i*)state->buffer)[0] = m256_zero;
     ((__m256i*)state->buffer)[1] = m256_zero;
@@ -906,9 +849,7 @@ int luffa_2way_update( luffa_2way_context *state, const void *data,
     __m256i msg[2];
     int i;
     int blocks = (int)len >> 5;
-    const __m256i shuff_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b,
+    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x1415161710111213,
-                                                 0x0c0d0e0f08090a0b,
                                                  0x0405060700010203 );
     state-> rembytes = (int)len & 0x1F;
 
@@ -926,7 +867,7 @@ int luffa_2way_update( luffa_2way_context *state, const void *data,
     {
       // remaining data bytes
       buffer[0] = _mm256_shuffle_epi8( vdata[0], shuff_bswap32 );
-      buffer[1] = m256_const1_i128( 0x0000000080000000 );
+      buffer[1] = mm256_bcast128lo_64( 0x0000000080000000 );
     }
     return 0;
 }
@@ -942,7 +883,7 @@ int luffa_2way_close( luffa_2way_context *state, void *hashval )
       rnd512_2way( state, buffer );
     else
     {     // empty pad block, constant data
-      msg[0] = m256_const1_i128( 0x0000000080000000 );
+      msg[0] = mm256_bcast128lo_64( 0x0000000080000000 );
       msg[1] = m256_zero;
       rnd512_2way( state, msg );
     }
@@ -959,16 +900,16 @@ int luffa512_2way_full( luffa_2way_context *state, void *output,
     state->hashbitlen = 512;
     __m128i *iv = (__m128i*)IV;
 
-    state->chainv[0] = m256_const1_128( iv[0] );
+    state->chainv[0] = mm256_bcast_m128( iv[0] );
-    state->chainv[1] = m256_const1_128( iv[1] );
+    state->chainv[1] = mm256_bcast_m128( iv[1] );
-    state->chainv[2] = m256_const1_128( iv[2] );
+    state->chainv[2] = mm256_bcast_m128( iv[2] );
-    state->chainv[3] = m256_const1_128( iv[3] );
+    state->chainv[3] = mm256_bcast_m128( iv[3] );
-    state->chainv[4] = m256_const1_128( iv[4] );
+    state->chainv[4] = mm256_bcast_m128( iv[4] );
-    state->chainv[5] = m256_const1_128( iv[5] );
+    state->chainv[5] = mm256_bcast_m128( iv[5] );
-    state->chainv[6] = m256_const1_128( iv[6] );
+    state->chainv[6] = mm256_bcast_m128( iv[6] );
-    state->chainv[7] = m256_const1_128( iv[7] );
+    state->chainv[7] = mm256_bcast_m128( iv[7] );
-    state->chainv[8] = m256_const1_128( iv[8] );
+    state->chainv[8] = mm256_bcast_m128( iv[8] );
-    state->chainv[9] = m256_const1_128( iv[9] );
+    state->chainv[9] = mm256_bcast_m128( iv[9] );
 
     ((__m256i*)state->buffer)[0] = m256_zero;
     ((__m256i*)state->buffer)[1] = m256_zero;
@@ -977,9 +918,7 @@ int luffa512_2way_full( luffa_2way_context *state, void *output,
     __m256i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m256i shuff_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b,
+    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x1415161710111213,
-                                                 0x0c0d0e0f08090a0b,
                                                  0x0405060700010203 );
 
     state->rembytes = inlen & 0x1F;
@@ -997,13 +936,13 @@ int luffa512_2way_full( luffa_2way_context *state, void *output,
     {
        // padding of partial block
        msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
-       msg[1] = m256_const1_i128( 0x0000000080000000 );
+       msg[1] = mm256_bcast128lo_64( 0x0000000080000000 );
        rnd512_2way( state, msg );
     }
     else
     {
        // empty pad block
-       msg[0] = m256_const1_i128( 0x0000000080000000 );
+       msg[0] = mm256_bcast128lo_64( 0x0000000080000000 );
        msg[1] = m256_zero;
        rnd512_2way( state, msg );
     }
@@ -1024,9 +963,7 @@ int luffa_2way_update_close( luffa_2way_context *state,
     __m256i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m256i shuff_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b,
+    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x1415161710111213,
-                                                 0x0c0d0e0f08090a0b,
                                                  0x0405060700010203 );
 
     state->rembytes = inlen & 0x1F;
@@ -1044,13 +981,13 @@ int luffa_2way_update_close( luffa_2way_context *state,
     {
        // padding of partial block
        msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
-       msg[1] = m256_const1_i128( 0x0000000080000000 );
+       msg[1] = mm256_bcast128lo_64( 0x0000000080000000 );
        rnd512_2way( state, msg );
     }
     else
     {
        // empty pad block
-       msg[0] = m256_const1_i128( 0x0000000080000000 );
+       msg[0] = mm256_bcast128lo_64( 0x0000000080000000 );
        msg[1] = m256_zero;
        rnd512_2way( state, msg );
     }

algo/luffa/luffa-hash-2way.h

@@ -23,7 +23,7 @@
 #if defined(__AVX2__)
 
 #include <immintrin.h>
-#include "algo/sha/sha3-defs.h"
+//#include "algo/sha/sha3-defs.h"
 #include "simd-utils.h"
 
 /* The length of digests*/
@@ -54,7 +54,7 @@
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
 typedef struct {
-    uint32 buffer[8*4];
+    uint32_t buffer[8*4];
     __m512i chainv[10];   /* Chaining values */
     int hashbitlen;
     int rembytes;
@@ -82,7 +82,7 @@ int luffa512_4way_update_close( luffa_4way_context *state, void *output,
 #endif
 
 typedef struct {
-    uint32 buffer[8*2];
+    uint32_t buffer[8*2];
     __m256i chainv[10];   /* Chaining values */
     int hashbitlen;
     int rembytes;

algo/luffa/luffa_for_sse2.c

@@ -22,20 +22,28 @@
 #include "simd-utils.h"
 #include "luffa_for_sse2.h"
 
+#define cns(i)  ( ( (__m128i*)CNS_INIT)[i] )
+
+#define ADD_CONSTANT( a, b, c0 ,c1 ) \
+    a = _mm_xor_si128( a, c0 ); \
+    b = _mm_xor_si128( b, c1 ); \
+
 #if defined(__AVX512VL__)
+//TODO enable for AVX10_512 AVX10_256
 
 #define MULT2( a0, a1 ) \
 { \
   __m128i b = _mm_xor_si128( a0, _mm_maskz_shuffle_epi32( 0xb, a1, 0x10 ) ); \
-  a0 = _mm_alignr_epi32( a1, b, 1 ); \
+  a0 = _mm_alignr_epi8( a1, b, 4 ); \
-  a1 = _mm_alignr_epi32( b, a1, 1 ); \
+  a1 = _mm_alignr_epi8( b, a1, 4 ); \
 }
 
 #elif defined(__SSE4_1__)
 
 #define MULT2( a0, a1 ) do \
 { \
-  __m128i b = _mm_xor_si128( a0, _mm_shuffle_epi32( mm128_mask_32( a1, 0xe ), 0x10 ) ); \
+  __m128i b = _mm_xor_si128( a0, \
+                      _mm_shuffle_epi32( mm128_mask_32( a1, 0xe ), 0x10 ) ); \
   a0 = _mm_alignr_epi8( a1, b, 4 ); \
   a1 = _mm_alignr_epi8( b, a1, 4 ); \
 } while(0)
@@ -44,79 +52,88 @@
 
 #define MULT2( a0, a1 ) do \
 { \
-  __m128i b = _mm_xor_si128( a0, _mm_shuffle_epi32( _mm_and_si128( a1, MASK ), 0x10 ) ); \
+  __m128i b = _mm_xor_si128( a0, \
-  a0 = _mm_or_si128( _mm_srli_si128( b, 4 ), _mm_slli_si128( a1, 12 ) ); \
+                      _mm_shuffle_epi32( _mm_and_si128( a1, MASK ), 0x10 ) ); \
-  a1 = _mm_or_si128( _mm_srli_si128( a1, 4 ), _mm_slli_si128( b, 12 ) ); \
+  a0 = _mm_or_si128( _mm_srli_si128(  b, 4 ), _mm_slli_si128( a1, 12 ) ); \
+  a1 = _mm_or_si128( _mm_srli_si128( a1, 4 ), _mm_slli_si128(  b, 12 ) ); \
 } while(0)
 
 #endif
 
-#define STEP_PART(x,c,t)\
+#if defined(__AVX512VL__)
-    SUBCRUMB(*x,*(x+1),*(x+2),*(x+3),*t);\
+//TODO enable for AVX10_512 AVX10_256
-    SUBCRUMB(*(x+5),*(x+6),*(x+7),*(x+4),*t);\
-    MIXWORD(*x,*(x+4),*t,*(t+1));\
-    MIXWORD(*(x+1),*(x+5),*t,*(t+1));\
-    MIXWORD(*(x+2),*(x+6),*t,*(t+1));\
-    MIXWORD(*(x+3),*(x+7),*t,*(t+1));\
-    ADD_CONSTANT(*x, *(x+4), *c, *(c+1));
 
-#define STEP_PART2(a0,a1,t0,t1,c0,c1,tmp0,tmp1)\
+#define SUBCRUMB( a0, a1, a2, a3 ) \
-    a1 = _mm_shuffle_epi32(a1,147);\
+{ \
-    t0 = _mm_load_si128(&a1);\
+    __m128i t = a0; \
-    a1 = _mm_unpacklo_epi32(a1,a0);\
+    a0 = mm128_xoror( a3, a0, a1 ); \
-    t0 = _mm_unpackhi_epi32(t0,a0);\
+    a2 = _mm_xor_si128( a2, a3 ); \
-    t1 = _mm_shuffle_epi32(t0,78);\
+    a1 = _mm_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 xnor (a3 & t) */ \
-    a0 = _mm_shuffle_epi32(a1,78);\
+    a3 = mm128_xorand( a2, a3, t ); \
-    SUBCRUMB(t1,t0,a0,a1,tmp0);\
+    a2 = mm128_xorand( a1, a2, a0 ); \
-    t0 = _mm_unpacklo_epi32(t0,t1);\
+    a1 = _mm_or_si128( a1, a3 ); \
-    a1 = _mm_unpacklo_epi32(a1,a0);\
+    a3 = _mm_xor_si128( a3, a2 ); \
-    a0 = _mm_load_si128(&a1);\
+    t  = _mm_xor_si128( t, a1 ); \
-    a0 = _mm_unpackhi_epi64(a0,t0);\
+    a2 = _mm_and_si128( a2, a1 ); \
-    a1 = _mm_unpacklo_epi64(a1,t0);\
+    a1 = mm128_xnor( a1, a0 ); \
-    a1 = _mm_shuffle_epi32(a1,57);\
+    a0 = t; \
-    MIXWORD(a0,a1,tmp0,tmp1);\
+}
-    ADD_CONSTANT(a0,a1,c0,c1);
 
-#define SUBCRUMB(a0,a1,a2,a3,t)\
+#else
-    t  = _mm_load_si128(&a0);\
-    a0 = _mm_or_si128(a0,a1);\
-    a2 = _mm_xor_si128(a2,a3);\
-    a1 = mm128_not( a1 );\
-    a0 = _mm_xor_si128(a0,a3);\
-    a3 = _mm_and_si128(a3,t);\
-    a1 = _mm_xor_si128(a1,a3);\
-    a3 = _mm_xor_si128(a3,a2);\
-    a2 = _mm_and_si128(a2,a0);\
-    a0 = mm128_not( a0 );\
-    a2 = _mm_xor_si128(a2,a1);\
-    a1 = _mm_or_si128(a1,a3);\
-    t  = _mm_xor_si128(t,a1);\
-    a3 = _mm_xor_si128(a3,a2);\
-    a2 = _mm_and_si128(a2,a1);\
-    a1 = _mm_xor_si128(a1,a0);\
-    a0 = _mm_load_si128(&t);\
 
-#define MIXWORD(a,b,t1,t2)\
+#define SUBCRUMB( a0, a1, a2, a3 ) \
-    b = _mm_xor_si128(a,b);\
+{ \
-    t1 = _mm_slli_epi32(a,2);\
+    __m128i t = a0; \
-    t2 = _mm_srli_epi32(a,30);\
+    a0 = _mm_or_si128( a0, a1 ); \
-    a = _mm_or_si128(t1,t2);\
+    a2 = _mm_xor_si128( a2, a3 ); \
-    a = _mm_xor_si128(a,b);\
+    a1 = mm128_not( a1 ); \
-    t1 = _mm_slli_epi32(b,14);\
+    a0 = _mm_xor_si128( a0, a3 ); \
-    t2 = _mm_srli_epi32(b,18);\
+    a3 = _mm_and_si128( a3, t ); \
-    b = _mm_or_si128(t1,t2);\
+    a1 = _mm_xor_si128( a1, a3 ); \
-    b = _mm_xor_si128(a,b);\
+    a3 = _mm_xor_si128( a3, a2 ); \
-    t1 = _mm_slli_epi32(a,10);\
+    a2 = _mm_and_si128( a2, a0 ); \
-    t2 = _mm_srli_epi32(a,22);\
+    a0 = mm128_not( a0 ); \
-    a = _mm_or_si128(t1,t2);\
+    a2 = _mm_xor_si128( a2, a1 ); \
-    a = _mm_xor_si128(a,b);\
+    a1 = _mm_or_si128(  a1, a3 ); \
-    t1 = _mm_slli_epi32(b,1);\
+    t  = _mm_xor_si128( t , a1 ); \
-    t2 = _mm_srli_epi32(b,31);\
+    a3 = _mm_xor_si128( a3, a2 ); \
-    b = _mm_or_si128(t1,t2);
+    a2 = _mm_and_si128( a2, a1 ); \
+    a1 = _mm_xor_si128( a1, a0 ); \
+    a0 = t; \
+}
 
-#define ADD_CONSTANT(a,b,c0,c1)\
+#endif
-    a = _mm_xor_si128(a,c0);\
+
-    b = _mm_xor_si128(b,c1);\
+#define MIXWORD( a, b ) \
+    b = _mm_xor_si128( a, b ); \
+    a = _mm_xor_si128( b, mm128_rol_32( a, 2 ) ); \
+    b = _mm_xor_si128( a, mm128_rol_32( b, 14 ) ); \
+    a = _mm_xor_si128( b, mm128_rol_32( a, 10 ) ); \
+    b = mm128_rol_32( b, 1 );
+
+#define STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, c0, c1 ) \
+    SUBCRUMB( x0, x1, x2, x3 ); \
+    SUBCRUMB( x5, x6, x7, x4 ); \
+    MIXWORD( x0, x4 ); \
+    MIXWORD( x1, x5 ); \
+    MIXWORD( x2, x6 ); \
+    MIXWORD( x3, x7 ); \
+    ADD_CONSTANT( x0, x4, c0, c1 );
+
+#define STEP_PART2( a0, a1, t0, t1, c0, c1 ) \
+    t0 = _mm_shuffle_epi32( a1, 147 ); \
+    a1 = _mm_unpacklo_epi32( t0, a0 ); \
+    t0 = _mm_unpackhi_epi32( t0, a0 ); \
+    t1 = _mm_shuffle_epi32( t0, 78 ); \
+    a0 = _mm_shuffle_epi32( a1, 78 ); \
+    SUBCRUMB( t1, t0, a0, a1 ); \
+    t0 = _mm_unpacklo_epi32( t0, t1 ); \
+    a1 = _mm_unpacklo_epi32( a1, a0 ); \
+    a0 = _mm_unpackhi_epi64( a1, t0 ); \
+    a1 = _mm_unpacklo_epi64( a1, t0 ); \
+    a1 = _mm_shuffle_epi32( a1, 57 ); \
+    MIXWORD( a0, a1 ); \
+    ADD_CONSTANT( a0, a1, c0, c1 );
 
 #define NMLTOM768(r0,r1,r2,s0,s1,s2,s3,p0,p1,p2,q0,q1,q2,q3)\
     s2 = _mm_load_si128(&r1);\
@@ -177,32 +194,22 @@
     q1 = _mm_load_si128(&p1);\
 
 #define NMLTOM1024(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3)\
-    s1 = _mm_load_si128(&r3);\
+    s1 = _mm_unpackhi_epi32( r3, r2 ); \
-    q1 = _mm_load_si128(&p3);\
+    q1 = _mm_unpackhi_epi32( p3, p2 ); \
-    s3 = _mm_load_si128(&r3);\
+    s3 = _mm_unpacklo_epi32( r3, r2 ); \
-    q3 = _mm_load_si128(&p3);\
+    q3 = _mm_unpacklo_epi32( p3, p2 ); \
-    s1 = _mm_unpackhi_epi32(s1,r2);\
+    r3 = _mm_unpackhi_epi32( r1, r0 ); \
-    q1 = _mm_unpackhi_epi32(q1,p2);\
+    r1 = _mm_unpacklo_epi32( r1, r0 ); \
-    s3 = _mm_unpacklo_epi32(s3,r2);\
+    p3 = _mm_unpackhi_epi32( p1, p0 ); \
-    q3 = _mm_unpacklo_epi32(q3,p2);\
+    p1 = _mm_unpacklo_epi32( p1, p0 ); \
-    s0 = _mm_load_si128(&s1);\
+    s0 = _mm_unpackhi_epi64( s1, r3 ); \
-    q0 = _mm_load_si128(&q1);\
+    q0 = _mm_unpackhi_epi64( q1 ,p3 ); \
-    s2 = _mm_load_si128(&s3);\
+    s1 = _mm_unpacklo_epi64( s1, r3 ); \
-    q2 = _mm_load_si128(&q3);\
+    q1 = _mm_unpacklo_epi64( q1, p3 ); \
-    r3 = _mm_load_si128(&r1);\
+    s2 = _mm_unpackhi_epi64( s3, r1 ); \
-    p3 = _mm_load_si128(&p1);\
+    q2 = _mm_unpackhi_epi64( q3, p1 ); \
-    r1 = _mm_unpacklo_epi32(r1,r0);\
+    s3 = _mm_unpacklo_epi64( s3, r1 ); \
-    p1 = _mm_unpacklo_epi32(p1,p0);\
+    q3 = _mm_unpacklo_epi64( q3, p1 );
-    r3 = _mm_unpackhi_epi32(r3,r0);\
-    p3 = _mm_unpackhi_epi32(p3,p0);\
-    s0 = _mm_unpackhi_epi64(s0,r3);\
-    q0 = _mm_unpackhi_epi64(q0,p3);\
-    s1 = _mm_unpacklo_epi64(s1,r3);\
-    q1 = _mm_unpacklo_epi64(q1,p3);\
-    s2 = _mm_unpackhi_epi64(s2,r1);\
-    q2 = _mm_unpackhi_epi64(q2,p1);\
-    s3 = _mm_unpacklo_epi64(s3,r1);\
-    q3 = _mm_unpacklo_epi64(q3,p1);
 
 #define MIXTON1024(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3)\
     NMLTOM1024(r0,r1,r2,r3,s0,s1,s2,s3,p0,p1,p2,p3,q0,q1,q2,q3);
@@ -306,8 +313,7 @@ HashReturn update_luffa( hashState_luffa *state, const BitSequence *data,
       // remaining data bytes
       casti_m128i( state->buffer, 0 ) = mm128_bswap_32( cast_m128i( data ) );
       // padding of partial block
-      casti_m128i( state->buffer, 1 ) =
+      casti_m128i( state->buffer, 1 ) =  _mm_set_epi32( 0, 0, 0, 0x80000000 );
-            _mm_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0x80,0,0,0 );
     }
 
     return SUCCESS;
@@ -325,8 +331,7 @@ HashReturn final_luffa(hashState_luffa *state, BitSequence *hashval)
     else
     {
       // empty pad block, constant data
-     rnd512( state, _mm_setzero_si128(),
+     rnd512( state, _mm_setzero_si128(), _mm_set_epi32( 0, 0, 0, 0x80000000 ) );
-                       _mm_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0x80,0,0,0 ) );
     }
 
     finalization512(state, (uint32*) hashval);
@@ -354,11 +359,11 @@ HashReturn update_and_final_luffa( hashState_luffa *state, BitSequence* output,
     // 16 byte partial block exists for 80 byte len
     if ( state->rembytes  )
        // padding of partial block
-       rnd512( state, m128_const_i128(  0x80000000 ),
+       rnd512( state, mm128_mov64_128(  0x80000000 ),
                       mm128_bswap_32( cast_m128i( data ) ) );
     else
        // empty pad block
-       rnd512( state, m128_zero, m128_const_i128( 0x80000000 ) );
+       rnd512( state, m128_zero, mm128_mov64_128( 0x80000000 ) );
 
     finalization512( state, (uint32*) output );
     if ( state->hashbitlen > 512 )
@@ -403,11 +408,11 @@ int luffa_full( hashState_luffa *state, BitSequence* output, int hashbitlen,
     // 16 byte partial block exists for 80 byte len
     if ( state->rembytes  )
        // padding of partial block
-       rnd512( state, m128_const_i128( 0x80000000 ),
+       rnd512( state, mm128_mov64_128( 0x80000000 ),
                       mm128_bswap_32( cast_m128i( data ) ) );
     else
        // empty pad block
-       rnd512( state, m128_zero, m128_const_i128( 0x80000000 ) );
+       rnd512( state, m128_zero, mm128_mov64_128( 0x80000000 ) );
 
     finalization512( state, (uint32*) output );
     if ( state->hashbitlen > 512 )
@@ -423,163 +428,119 @@ int luffa_full( hashState_luffa *state, BitSequence* output, int hashbitlen,
 
 static void rnd512( hashState_luffa *state, __m128i msg1, __m128i msg0 )
 {
-    __m128i t[2];
+    __m128i t0, t1;
     __m128i *chainv = state->chainv;
-    __m128i tmp[2];
+    __m128i x0, x1, x2, x3, x4, x5, x6, x7; 
-    __m128i x[8];
 
-    t[0] = chainv[0];
+    t0 = mm128_xor3( chainv[0], chainv[2], chainv[4] );
-    t[1] = chainv[1];
+    t1 = mm128_xor3( chainv[1], chainv[3], chainv[5] );
+    t0 = mm128_xor3( t0, chainv[6], chainv[8] );
+    t1 = mm128_xor3( t1, chainv[7], chainv[9] );
 
-    t[0] = _mm_xor_si128( t[0], chainv[2] );
+    MULT2( t0, t1 );
-    t[1] = _mm_xor_si128( t[1], chainv[3] );
-    t[0] = _mm_xor_si128( t[0], chainv[4] );
-    t[1] = _mm_xor_si128( t[1], chainv[5] );
-    t[0] = _mm_xor_si128( t[0], chainv[6] );
-    t[1] = _mm_xor_si128( t[1], chainv[7] );
-    t[0] = _mm_xor_si128( t[0], chainv[8] );
-    t[1] = _mm_xor_si128( t[1], chainv[9] );
-
-    MULT2( t[0], t[1] );
 
     msg0 = _mm_shuffle_epi32( msg0, 27 );
     msg1 = _mm_shuffle_epi32( msg1, 27 );
 
-    chainv[0] = _mm_xor_si128( chainv[0], t[0] );
+    chainv[0] = _mm_xor_si128( chainv[0], t0 );
-    chainv[1] = _mm_xor_si128( chainv[1], t[1] );
+    chainv[1] = _mm_xor_si128( chainv[1], t1 );
-    chainv[2] = _mm_xor_si128( chainv[2], t[0] );
+    chainv[2] = _mm_xor_si128( chainv[2], t0 );
-    chainv[3] = _mm_xor_si128( chainv[3], t[1] );
+    chainv[3] = _mm_xor_si128( chainv[3], t1 );
-    chainv[4] = _mm_xor_si128( chainv[4], t[0] );
+    chainv[4] = _mm_xor_si128( chainv[4], t0 );
-    chainv[5] = _mm_xor_si128( chainv[5], t[1] );
+    chainv[5] = _mm_xor_si128( chainv[5], t1 );
-    chainv[6] = _mm_xor_si128( chainv[6], t[0] );
+    chainv[6] = _mm_xor_si128( chainv[6], t0 );
-    chainv[7] = _mm_xor_si128( chainv[7], t[1] );
+    chainv[7] = _mm_xor_si128( chainv[7], t1 );
-    chainv[8] = _mm_xor_si128( chainv[8], t[0] );
+    chainv[8] = _mm_xor_si128( chainv[8], t0 );
-    chainv[9] = _mm_xor_si128( chainv[9], t[1] );
+    chainv[9] = _mm_xor_si128( chainv[9], t1 );
 
-    t[0] = chainv[0];
+    t0 = chainv[0];
-    t[1] = chainv[1];
+    t1 = chainv[1];
 
     MULT2( chainv[0], chainv[1]);
-
     chainv[0] = _mm_xor_si128( chainv[0], chainv[2] );
     chainv[1] = _mm_xor_si128( chainv[1], chainv[3] );
 
     MULT2( chainv[2], chainv[3]);
-
     chainv[2] = _mm_xor_si128(chainv[2], chainv[4]);
     chainv[3] = _mm_xor_si128(chainv[3], chainv[5]);
 
     MULT2( chainv[4], chainv[5]);
-
     chainv[4] = _mm_xor_si128(chainv[4], chainv[6]);
     chainv[5] = _mm_xor_si128(chainv[5], chainv[7]);
 
     MULT2( chainv[6], chainv[7]);
-
     chainv[6] = _mm_xor_si128(chainv[6], chainv[8]);
     chainv[7] = _mm_xor_si128(chainv[7], chainv[9]);
 
     MULT2( chainv[8], chainv[9]);
-
+    t0 = chainv[8] = _mm_xor_si128( chainv[8], t0 );
-    chainv[8] = _mm_xor_si128( chainv[8], t[0] );
+    t1 = chainv[9] = _mm_xor_si128( chainv[9], t1 );
-    chainv[9] = _mm_xor_si128( chainv[9], t[1] );
-
-    t[0] = chainv[8];
-    t[1] = chainv[9];
 
     MULT2( chainv[8], chainv[9]);
-
     chainv[8] = _mm_xor_si128( chainv[8], chainv[6] );
     chainv[9] = _mm_xor_si128( chainv[9], chainv[7] );
 
     MULT2( chainv[6], chainv[7]);
-
     chainv[6] = _mm_xor_si128( chainv[6], chainv[4] );
     chainv[7] = _mm_xor_si128( chainv[7], chainv[5] );
 
     MULT2( chainv[4], chainv[5]);
-
     chainv[4] = _mm_xor_si128( chainv[4], chainv[2] );
     chainv[5] = _mm_xor_si128( chainv[5], chainv[3] );
 
     MULT2( chainv[2], chainv[3] );
-
     chainv[2] = _mm_xor_si128( chainv[2], chainv[0] );
     chainv[3] = _mm_xor_si128( chainv[3], chainv[1] );
 
     MULT2( chainv[0], chainv[1] );
-
+    chainv[0] = _mm_xor_si128( _mm_xor_si128( chainv[0], t0 ), msg0 );
-    chainv[0] = _mm_xor_si128( _mm_xor_si128( chainv[0], t[0] ), msg0 );
+    chainv[1] = _mm_xor_si128( _mm_xor_si128( chainv[1], t1 ), msg1 );
-    chainv[1] = _mm_xor_si128( _mm_xor_si128( chainv[1], t[1] ), msg1 );
 
     MULT2( msg0, msg1);
-
     chainv[2] = _mm_xor_si128( chainv[2], msg0 );
     chainv[3] = _mm_xor_si128( chainv[3], msg1 );
 
     MULT2( msg0, msg1);
-
     chainv[4] = _mm_xor_si128( chainv[4], msg0 );
     chainv[5] = _mm_xor_si128( chainv[5], msg1 );
 
     MULT2( msg0, msg1);
-
     chainv[6] = _mm_xor_si128( chainv[6], msg0 );
     chainv[7] = _mm_xor_si128( chainv[7], msg1 );
 
     MULT2( msg0, msg1);
-
     chainv[8] = _mm_xor_si128( chainv[8], msg0 );
     chainv[9] = _mm_xor_si128( chainv[9], msg1 );
 
     MULT2( msg0, msg1);
+    chainv[3] = mm128_rol_32( chainv[3], 1 );    
+    chainv[5] = mm128_rol_32( chainv[5], 2 );
+    chainv[7] = mm128_rol_32( chainv[7], 3 );
+    chainv[9] = mm128_rol_32( chainv[9], 4 );
+    
+    NMLTOM1024( chainv[0], chainv[2], chainv[4], chainv[6], x0, x1, x2, x3,
+                chainv[1], chainv[3], chainv[5], chainv[7], x4, x5, x6, x7 );
 
-    chainv[3] = _mm_or_si128( _mm_slli_epi32(chainv[3], 1),
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns( 0), cns( 1) );
-                              _mm_srli_epi32(chainv[3], 31) );
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns( 2), cns( 3) );
-    chainv[5] = _mm_or_si128( _mm_slli_epi32(chainv[5], 2),
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns( 4), cns( 5) );
-                              _mm_srli_epi32(chainv[5], 30) );
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns( 6), cns( 7) );
-    chainv[7] = _mm_or_si128( _mm_slli_epi32(chainv[7], 3),
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns( 8), cns( 9) );
-                              _mm_srli_epi32(chainv[7], 29) );
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns(10), cns(11) );
-    chainv[9] = _mm_or_si128( _mm_slli_epi32(chainv[9], 4),
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns(12), cns(13) );
-                              _mm_srli_epi32(chainv[9], 28) );
+    STEP_PART( x0, x1, x2, x3, x4, x5, x6, x7, cns(14), cns(15) );
+    
+    MIXTON1024( x0, x1, x2, x3, chainv[0], chainv[2], chainv[4], chainv[6],
+                x4, x5, x6, x7, chainv[1], chainv[3], chainv[5], chainv[7]);
 
-
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(16), cns(17) );
-    NMLTOM1024( chainv[0], chainv[2], chainv[4], chainv[6],
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(18), cns(19) );
-                x[0], x[1], x[2], x[3],
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(20), cns(21) );
-                chainv[1],chainv[3],chainv[5],chainv[7],
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(22), cns(23) );
-                x[4], x[5], x[6], x[7] );
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(24), cns(25) );
-
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(26), cns(27) );
-    STEP_PART( &x[0], &CNS128[ 0], &tmp[0] );
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(28), cns(29) );
-    STEP_PART( &x[0], &CNS128[ 2], &tmp[0] );
+    STEP_PART2( chainv[8], chainv[9], t0, t1, cns(30), cns(31) );
-    STEP_PART( &x[0], &CNS128[ 4], &tmp[0] );
-    STEP_PART( &x[0], &CNS128[ 6], &tmp[0] );
-    STEP_PART( &x[0], &CNS128[ 8], &tmp[0] );
-    STEP_PART( &x[0], &CNS128[10], &tmp[0] );
-    STEP_PART( &x[0], &CNS128[12], &tmp[0] );
-    STEP_PART( &x[0], &CNS128[14], &tmp[0] );
-
-    MIXTON1024( x[0], x[1], x[2], x[3],
-                chainv[0], chainv[2], chainv[4],chainv[6],
-                x[4], x[5], x[6], x[7],
-                chainv[1],chainv[3],chainv[5],chainv[7]);
-
-    /* Process last 256-bit block */
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[16], CNS128[17],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[18], CNS128[19],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[20], CNS128[21],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[22], CNS128[23],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[24], CNS128[25],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[26], CNS128[27],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[28], CNS128[29],
-                tmp[0], tmp[1] );
-    STEP_PART2( chainv[8], chainv[9], t[0], t[1], CNS128[30], CNS128[31],
-                tmp[0], tmp[1] );
 }
 
 
@@ -588,51 +549,6 @@ static void rnd512( hashState_luffa *state, __m128i msg1, __m128i msg0 )
 /* state: hash context    */
 /* b[8]: hash values      */
 
-#if defined (__AVX2__)
-
-static void finalization512( hashState_luffa *state, uint32 *b )
-{
-    uint32   hash[8] __attribute((aligned(64)));
-    __m256i* chainv = (__m256i*)state->chainv;
-    __m256i  t;
-    const __m128i zero = m128_zero;
-    const __m256i shuff_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b,
-                                                 0x1415161710111213,
-                                                 0x0c0d0e0f08090a0b,
-                                                 0x0405060700010203 );
-
-    rnd512( state, zero, zero );
-
-    t = chainv[0];
-    t = _mm256_xor_si256( t, chainv[1] );
-    t = _mm256_xor_si256( t, chainv[2] );
-    t = _mm256_xor_si256( t, chainv[3] );
-    t = _mm256_xor_si256( t, chainv[4] );
-
-    t = _mm256_shuffle_epi32( t, 27 );
-
-    _mm256_store_si256( (__m256i*)hash, t );
-
-    casti_m256i( b, 0 ) = _mm256_shuffle_epi8(
-                                 casti_m256i( hash, 0 ), shuff_bswap32 );
-
-    rnd512( state, zero, zero );
-
-    t = chainv[0];
-    t = _mm256_xor_si256( t, chainv[1] );
-    t = _mm256_xor_si256( t, chainv[2] );
-    t = _mm256_xor_si256( t, chainv[3] );
-    t = _mm256_xor_si256( t, chainv[4] );
-    t = _mm256_shuffle_epi32( t, 27 );
-
-    _mm256_store_si256( (__m256i*)hash, t );
-
-    casti_m256i( b, 1 ) = _mm256_shuffle_epi8( 
-                                 casti_m256i( hash, 0 ), shuff_bswap32 );
-}
-
-#else
-
 static void finalization512( hashState_luffa *state, uint32 *b )
 {
     uint32 hash[8] __attribute((aligned(64)));
@@ -685,6 +601,5 @@ static void finalization512( hashState_luffa *state, uint32 *b )
     casti_m128i( b, 2 ) = mm128_bswap_32( casti_m128i( hash, 0 ) );
     casti_m128i( b, 3 ) = mm128_bswap_32( casti_m128i( hash, 1 ) );
 }
-#endif
 
 /***************************************************/

algo/luffa/luffa_for_sse2.h

@@ -22,7 +22,7 @@
  */
 
 #include <emmintrin.h>
-#include "algo/sha/sha3-defs.h"
+#include "compat/sha3-defs.h"
 /* The length of digests*/
 #define DIGEST_BIT_LEN_224 224
 #define DIGEST_BIT_LEN_256 256

algo/luffa/sph_luffa.h

@@ -41,7 +41,7 @@ extern "C"{
 #endif
 
 #include <stddef.h>
-#include "algo/sha/sph_types.h"
+#include "compat/sph_types.h"
 
 /**
  * Output size (in bits) for Luffa-224.

algo/lyra2/allium-4way.c

@@ -1,7 +1,7 @@
 #include "lyra2-gate.h"
 #include <memory.h>
 #include <mm_malloc.h>
-#include "algo/blake/blake-hash-4way.h"
+#include "algo/blake/blake256-hash.h"
 #include "algo/keccak/keccak-hash-4way.h"
 #include "algo/skein/skein-hash-4way.h"
 #include "algo/cubehash/cubehash_sse2.h"
@@ -48,7 +48,7 @@ static void allium_16way_hash( void *state, const void *midstate_vars,
    uint32_t hash15[8] __attribute__ ((aligned (32)));
    allium_16way_ctx_holder ctx __attribute__ ((aligned (64)));
 
-   blake256_16way_final_rounds_le( vhash, midstate_vars, midhash, block );
+   blake256_16way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
 
    dintrlv_16x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
                   hash8, hash9, hash10, hash11, hash12, hash13, hash14, hash15,
@@ -212,12 +212,12 @@ int scanhash_allium_16way( struct work *work, uint32_t max_nonce,
    const uint32_t last_nonce = max_nonce - 16;
    const int thr_id = mythr->id;
    const bool bench = opt_benchmark;
-   const __m512i sixteen = m512_const1_32( 16 );
+   const __m512i sixteen = _mm512_set1_epi32( 16 );
 
    if ( bench ) ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
    // Prehash first block.
-   blake256_transform_le( phash, pdata, 512, 0 );
+   blake256_transform_le( phash, pdata, 512, 0, 14 );
 
    // Interleave hash for second block prehash.
    block0_hash[0] = _mm512_set1_epi32( phash[0] );
@@ -286,7 +286,7 @@ static void allium_8way_hash( void *hash, const void *midstate_vars,
    uint64_t *hash7 = (uint64_t*)hash+28;
    allium_8way_ctx_holder ctx __attribute__ ((aligned (64))); 
 
-   blake256_8way_final_rounds_le( vhashA, midstate_vars, midhash, block );
+   blake256_8way_final_rounds_le( vhashA, midstate_vars, midhash, block, 14 );
 
    dintrlv_8x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
                  vhashA, 256 );
@@ -398,10 +398,10 @@ int scanhash_allium_8way( struct work *work, uint32_t max_nonce,
    uint32_t n = first_nonce;
    const int thr_id = mythr->id;  
    const bool bench = opt_benchmark;
-   const __m256i eight = m256_const1_32( 8 );
+   const __m256i eight = _mm256_set1_epi32( 8 );
 
    // Prehash first block
-   blake256_transform_le( phash, pdata, 512, 0 );
+   blake256_transform_le( phash, pdata, 512, 0, 14 );
 
    block0_hash[0] = _mm256_set1_epi32( phash[0] );
    block0_hash[1] = _mm256_set1_epi32( phash[1] );

algo/lyra2/lyra2.h

@@ -21,9 +21,8 @@
 #define LYRA2_H_
 
 #include <stdint.h>
-#include "algo/sha/sha3-defs.h"
 
-//typedef unsigned char byte;
+typedef unsigned char byte;
 
 //Block length required so Blake2's Initialization Vector (IV) is not overwritten (THIS SHOULD NOT BE MODIFIED)
 #define BLOCK_LEN_BLAKE2_SAFE_INT64 8                                   //512 bits (=64 bytes, =8 uint64_t)

algo/lyra2/lyra2h-4way.c

@@ -5,8 +5,7 @@
 #include <memory.h>
 #include <mm_malloc.h>
 #include "lyra2.h"
-//#include "algo/blake/sph_blake.h"
+#include "algo/blake/blake256-hash.h"
-#include "algo/blake/blake-hash-4way.h"
 
 __thread uint64_t* lyra2h_4way_matrix;
 

algo/lyra2/lyra2rev2-4way.c

@@ -1,6 +1,6 @@
 #include "lyra2-gate.h"
 #include <memory.h>
-#include "algo/blake/blake-hash-4way.h"
+#include "algo/blake/blake256-hash.h"
 #include "algo/keccak/keccak-hash-4way.h"
 #include "algo/skein/skein-hash-4way.h"
 #include "algo/bmw/bmw-hash-4way.h"
@@ -75,7 +75,7 @@ void lyra2rev2_16way_hash( void *state, const void *input )
    keccak256_8way_close( &ctx.keccak, vhash );
 
    dintrlv_8x64( hash8,  hash9,  hash10,  hash11,
-                 hash12, hash13, hash14, hash5, vhash, 256 );
+                 hash12, hash13, hash14, hash15, vhash, 256 );
 
    cubehash_full( &ctx.cube, (byte*) hash0,  256, (const byte*) hash0,  32 );
    cubehash_full( &ctx.cube, (byte*) hash1,  256, (const byte*) hash1,  32 );
@@ -203,7 +203,7 @@ int scanhash_lyra2rev2_16way( struct work *work, const uint32_t max_nonce,
              submit_solution( work, lane_hash, mythr );
          }
       }
-      *noncev = _mm512_add_epi32( *noncev, m512_const1_32( 16 ) );
+      *noncev = _mm512_add_epi32( *noncev, _mm512_set1_epi32( 16 ) );
       n += 16;
    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
    pdata[19] = n;
@@ -345,7 +345,7 @@ int scanhash_lyra2rev2_8way( struct work *work, const uint32_t max_nonce,
              submit_solution( work, lane_hash, mythr );
          }
       }
-      *noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
+      *noncev = _mm256_add_epi32( *noncev, _mm256_set1_epi32( 8 ) );
       n += 8;
    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
    pdata[19] = n;

algo/lyra2/lyra2rev2.c

@@ -4,7 +4,6 @@
 
 #include <memory.h>
 #include "algo/blake/sph_blake.h"
-#include "algo/cubehash/sph_cubehash.h"
 #include "algo/keccak/sph_keccak.h"
 #include "algo/skein/sph_skein.h"
 #include "algo/bmw/sph_bmw.h"

algo/lyra2/lyra2rev3-4way.c

@@ -1,7 +1,7 @@
 #include "lyra2-gate.h"
 #include <memory.h>
 
-#include "algo/blake/blake-hash-4way.h"
+#include "algo/blake/blake256-hash.h"
 #include "algo/bmw/bmw-hash-4way.h"
 #include "algo/cubehash/cubehash_sse2.h" 
 #include "algo/cubehash/cube-hash-2way.h"
@@ -287,7 +287,7 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
              submit_solution( work, lane_hash, mythr );
          }
       }
-      *noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
+      *noncev = _mm256_add_epi32( *noncev, _mm256_set1_epi32( 8 ) );
       n += 8;
    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
    pdata[19] = n;
@@ -389,7 +389,7 @@ int scanhash_lyra2rev3_4way( struct work *work, const uint32_t max_nonce,
               submit_solution( work, lane_hash, mythr );
 	      }
       }
-      *noncev = _mm_add_epi32( *noncev, m128_const1_32( 4 ) );
+      *noncev = _mm_add_epi32( *noncev, _mm_set1_epi32( 4 ) );
       n += 4;
    } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
    pdata[19] = n;

algo/lyra2/lyra2rev3.c

@@ -4,7 +4,6 @@
 
 #include <memory.h>
 #include "algo/blake/sph_blake.h"
-#include "algo/cubehash/sph_cubehash.h"
 #include "algo/bmw/sph_bmw.h"
 #include "algo/cubehash/cubehash_sse2.h" 
 //#include "lyra2.h"

algo/lyra2/lyra2z-4way.c

@@ -2,8 +2,7 @@
 #include <memory.h>
 #include <mm_malloc.h>
 #include "lyra2.h"
-#include "algo/blake/sph_blake.h"
+#include "algo/blake/blake256-hash.h"
-#include "algo/blake/blake-hash-4way.h"
 
 #if defined(LYRA2Z_16WAY)
 
@@ -35,7 +34,7 @@ static void lyra2z_16way_hash( void *state, const void *midstate_vars,
     uint32_t hash14[8] __attribute__ ((aligned (32)));
     uint32_t hash15[8] __attribute__ ((aligned (32)));
 
-    blake256_16way_final_rounds_le( vhash, midstate_vars, midhash, block );
+    blake256_16way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
 
     dintrlv_16x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
               hash8, hash9, hash10, hash11 ,hash12, hash13, hash14, hash15,
@@ -103,12 +102,12 @@ int scanhash_lyra2z_16way( struct work *work, uint32_t max_nonce,
    const uint32_t last_nonce = max_nonce - 16;
    const int thr_id = mythr->id;
    const bool bench = opt_benchmark;
-   const __m512i sixteen = m512_const1_32( 16 );
+   const __m512i sixteen = _mm512_set1_epi32( 16 );
 
    if ( bench ) ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
    // Prehash first block
-   blake256_transform_le( phash, pdata, 512, 0 );
+   blake256_transform_le( phash, pdata, 512, 0, 14 );
 
    block0_hash[0] = _mm512_set1_epi32( phash[0] );
    block0_hash[1] = _mm512_set1_epi32( phash[1] );
@@ -170,7 +169,7 @@ static void lyra2z_8way_hash( void *state, const void *midstate_vars,
      uint32_t hash7[8] __attribute__ ((aligned (32)));
      uint32_t vhash[8*8] __attribute__ ((aligned (64)));
 
-     blake256_8way_final_rounds_le( vhash, midstate_vars, midhash, block );
+     blake256_8way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
 
      dintrlv_8x32( hash0, hash1, hash2, hash3,
                    hash4, hash5, hash6, hash7, vhash, 256 );
@@ -213,10 +212,10 @@ int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
    uint32_t n = first_nonce;
    const int thr_id = mythr->id;
    const bool bench = opt_benchmark;
-   const __m256i eight = m256_const1_32( 8 );
+   const __m256i eight = _mm256_set1_epi32( 8 );
 
    // Prehash first block
-   blake256_transform_le( phash, pdata, 512, 0 );
+   blake256_transform_le( phash, pdata, 512, 0, 14 );
 
    block0_hash[0] = _mm256_set1_epi32( phash[0] );
    block0_hash[1] = _mm256_set1_epi32( phash[1] );
@@ -328,7 +327,7 @@ int scanhash_lyra2z_4way( struct work *work, uint32_t max_nonce,
            submit_solution( work, lane_hash, mythr );
         }
       }
-      *noncev = _mm_add_epi32( *noncev, m128_const1_32( 4 ) );
+      *noncev = _mm_add_epi32( *noncev, _mm_set1_epi32( 4 ) );
       n += 4;
    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
 

algo/lyra2/sponge-2way.c

@@ -85,10 +85,10 @@ inline void absorbBlockBlake2Safe_2way( uint64_t *State, const uint64_t *In,
 
   state0 = 
   state1 = m512_zero;
-  state2 = m512_const4_64( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL,
+  state2 = _mm512_set4_epi64( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL,
-                           0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
+                              0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
-  state3 = m512_const4_64( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL,
+  state3 = _mm512_set4_epi64( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL,
-                           0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
+                              0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
 
   for ( int i = 0; i < nBlocks; i++ )
   { 

algo/lyra2/sponge.c

@@ -41,17 +41,17 @@
 inline void initState( uint64_t State[/*16*/] )
 {
 
-   /*
+/*
 #if defined (__AVX2__)
 
   __m256i* state = (__m256i*)State;
   const __m256i zero = m256_zero; 
   state[0] = zero;
   state[1] = zero;
-  state[2] = m256_const_64( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL,
+  state[2] = _mm256_set_epi64x( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL,
-                            0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
+                                0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
-  state[3] = m256_const_64( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL,
+  state[3] = _mm256_set_epi64x( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL,
-                            0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
+                                0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
 
 #elif defined (__SSE2__)
 
@@ -62,10 +62,10 @@ inline void initState( uint64_t State[/*16*/] )
   state[1] = zero;
   state[2] = zero;
   state[3] = zero;
-  state[4] = m128_const_64( 0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
+  state[4] = _mm_set_epi64x( 0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
-  state[5] = m128_const_64( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL );
+  state[5] = _mm_set_epi64x( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL );
-  state[6] = m128_const_64( 0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
+  state[6] = _mm_set_epi64x( 0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
-  state[7] = m128_const_64( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL );
+  state[7] = _mm_set_epi64x( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL );
 
 #else
     //First 512 bis are zeros
@@ -271,10 +271,10 @@ inline void absorbBlockBlake2Safe( uint64_t *State, const uint64_t *In,
 
   state0 = 
   state1 = m256_zero;
-  state2 = m256_const_64( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL,
+  state2 = _mm256_set_epi64x( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL,
-                          0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
+                              0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
-  state3 = m256_const_64( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL,
+  state3 = _mm256_set_epi64x( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL,
-                          0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
+                              0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
 
   for ( int i = 0; i < nBlocks; i++ )
   { 
@@ -299,10 +299,10 @@ inline void absorbBlockBlake2Safe( uint64_t *State, const uint64_t *In,
   state1 =
   state2 =
   state3 = m128_zero;
-  state4 = m128_const_64( 0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
+  state4 = _mm_set_epi64x( 0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL );
-  state5 = m128_const_64( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL );
+  state5 = _mm_set_epi64x( 0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL );
-  state6 = m128_const_64( 0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
+  state6 = _mm_set_epi64x( 0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL );
-  state7 = m128_const_64( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL );
+  state7 = _mm_set_epi64x( 0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL );
 
   for ( int i = 0; i < nBlocks; i++ )
   { 

algo/lyra2/sponge.h

@@ -43,27 +43,29 @@ static const uint64_t blake2b_IV[8] =
   0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
 };
 
-/*Blake2b's rotation*/
-static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
-    return ( w >> c ) | ( w << ( 64 - c ) );
-}
-
-// serial data is only 32 bytes so AVX2 is the limit for that dimension.
-// However, 2 way parallel looks trivial to code for AVX512 except for
-// a data dependency with rowa.
-
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
 #define G2W_4X64(a,b,c,d) \
    a = _mm512_add_epi64( a, b ); \
-   d = mm512_ror_64( _mm512_xor_si512( d, a ), 32 ); \
+   d = _mm512_ror_epi64( _mm512_xor_si512( d, a ), 32 ); \
    c = _mm512_add_epi64( c, d ); \
-   b = mm512_ror_64( _mm512_xor_si512( b, c ), 24 ); \
+   b = _mm512_ror_epi64( _mm512_xor_si512( b, c ), 24 ); \
    a = _mm512_add_epi64( a, b ); \
-   d = mm512_ror_64( _mm512_xor_si512( d, a ), 16 ); \
+   d = _mm512_ror_epi64( _mm512_xor_si512( d, a ), 16 ); \
    c = _mm512_add_epi64( c, d ); \
-   b = mm512_ror_64( _mm512_xor_si512( b, c ), 63 );
+   b = _mm512_ror_epi64( _mm512_xor_si512( b, c ), 63 );
 
+#define LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
+   G2W_4X64( s0, s1, s2, s3 ); \
+   s0 = mm512_shufll256_64( s0 ); \
+   s3 = mm512_swap256_128( s3); \
+   s2 = mm512_shuflr256_64( s2 ); \
+   G2W_4X64( s0, s1, s2, s3 ); \
+   s0 = mm512_shuflr256_64( s0 ); \
+   s3 = mm512_swap256_128( s3 ); \
+   s2 = mm512_shufll256_64( s2 ); 
+
+/*
 #define LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
    G2W_4X64( s0, s1, s2, s3 ); \
    s3 = mm512_shufll256_64( s3 ); \
@@ -73,6 +75,7 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
    s3 = mm512_shuflr256_64( s3 ); \
    s1 = mm512_shufll256_64( s1 ); \
    s2 = mm512_swap256_128( s2 ); 
+*/
 
 #define LYRA_12_ROUNDS_2WAY_AVX512( s0, s1, s2, s3 ) \
    LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
@@ -88,13 +91,10 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
    LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
    LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 )
 
-
 #endif  // AVX512
 
-#if defined __AVX2__
+#if defined(__AVX2__)
 
-// process 4 columns in parallel
-// returns void, updates all args
 #define G_4X64(a,b,c,d) \
    a = _mm256_add_epi64( a, b ); \
    d = mm256_swap64_32( _mm256_xor_si256( d, a ) ); \
@@ -105,6 +105,18 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
    c = _mm256_add_epi64( c, d ); \
    b = mm256_ror_64( _mm256_xor_si256( b, c ), 63 );
 
+// Pivot about s1 instead of s0 reduces latency.
+#define LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
+   G_4X64( s0, s1, s2, s3 ); \
+   s0 = mm256_shufll_64( s0 ); \
+   s3 = mm256_swap_128( s3); \
+   s2 = mm256_shuflr_64( s2 ); \
+   G_4X64( s0, s1, s2, s3 ); \
+   s0 = mm256_shuflr_64( s0 ); \
+   s3 = mm256_swap_128( s3 ); \
+   s2 = mm256_shufll_64( s2 );
+
+/*
 #define LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
    G_4X64( s0, s1, s2, s3 ); \
    s3 = mm256_shufll_64( s3 ); \
@@ -114,6 +126,7 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
    s3 = mm256_shuflr_64( s3 ); \
    s1 = mm256_shufll_64( s1 ); \
    s2 = mm256_swap_128( s2 );
+*/
 
 #define LYRA_12_ROUNDS_AVX2( s0, s1, s2, s3 ) \
    LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
@@ -182,8 +195,13 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
 
 #endif // AVX2 else SSE2
 
-// Scalar
+/*
-//Blake2b's G function
+// Scalar, not used.
+
+static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
+    return ( w >> c ) | ( w << ( 64 - c ) );
+}
+
 #define G(r,i,a,b,c,d) \
   do { \
     a = a + b; \
@@ -196,8 +214,6 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
     b = rotr64(b ^ c, 63); \
   } while(0)
 
-
-/*One Round of the Blake2b's compression function*/
 #define ROUND_LYRA(r)  \
     G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
     G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
@@ -207,6 +223,7 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
     G(r,5,v[ 1],v[ 6],v[11],v[12]); \
     G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
     G(r,7,v[ 3],v[ 4],v[ 9],v[14]);
+*/
 
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 

algo/nist5/nist5-4way.c

@@ -3,7 +3,7 @@
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
-#include "algo/blake/blake-hash-4way.h"
+#include "algo/blake/blake512-hash.h"
 #include "algo/skein/skein-hash-4way.h"
 #include "algo/jh/jh-hash-4way.h"
 #include "algo/keccak/keccak-hash-4way.h"

algo/panama/panama-hash-4way.c

@@ -98,7 +98,7 @@ do { \
    ( g ## n0 = _mm_xor_si128( a ## n0, _mm_xor_si128( a ## n1, a ## n4 ) ) )
 
 #define SIGMA_ALL_4W   do { \
-		a0 = _mm_xor_si128( g0, m128_one_32 ); \
+		a0 = _mm_xor_si128( g0, v128_32( 1 ) ); \
 		a1 = _mm_xor_si128( g1, INW2( 0 ) ); \
 		a2 = _mm_xor_si128( g2, INW2( 1 ) ); \
 		a3 = _mm_xor_si128( g3, INW2( 2 ) ); \
@@ -268,7 +268,7 @@ panama_4way_close( void *cc, void *dst )
 
 	sc = cc;
 	current = sc->data_ptr;
-	*(__m128i*)( sc->data + current ) = m128_one_32;
+	*(__m128i*)( sc->data + current ) = v128_32( 1 );
    current++;
    memset_zero_128( (__m128i*)sc->data + current, 32 - current );
    panama_4way_push( sc, sc->data, 1 );
@@ -354,7 +354,7 @@ do { \
 
 
 #define SIGMA_ALL_8W   do { \
-      a0  = _mm256_xor_si256( g0, m256_one_32 ); \
+      a0  = _mm256_xor_si256( g0, v256_32( 1 ) ); \
       a1  = _mm256_xor_si256( g1, INW2( 0 ) ); \
       a2  = _mm256_xor_si256( g2, INW2( 1 ) ); \
       a3  = _mm256_xor_si256( g3, INW2( 2 ) ); \
@@ -521,7 +521,7 @@ panama_8way_close( void *cc, void *dst )
 
    sc = cc;
    current = sc->data_ptr;
-   *(__m256i*)( sc->data + current ) = m256_one_32;
+   *(__m256i*)( sc->data + current ) = v256_32( 1 );
    current++;
    memset_zero_256( (__m256i*)sc->data + current, 32 - current );
    panama_8way_push( sc, sc->data, 1 );