popov/cpuminer-opt-gpu
1
0
Fork 0
mirror of https://github.com/JayDDee/cpuminer-opt.git synced 2025-09-17 23:44:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

v3.9.1

Browse Source
This commit is contained in:
Jay D Dee
2019-05-30 16:59:49 -04:00
parent eb3f57bfc7
commit 77c5ae80ab
82 changed files with 6906 additions and 3706 deletions
Download Patch File Download Diff File Expand all files Collapse all files

126
INSTALL_LINUX Normal file
View File

@@ -0,0 +1,126 @@
Requirements:
Intel Core2 or newer, or AMD Steamroller or newer CPU. ARM CPUs are not
supported.
64 bit Linux operating system. Apple is not supported.
Building on linux prerequisites:
It is assumed users know how to install packages on their system and
be able to compile standard source packages. This is basic Linux and
beyond the scope of cpuminer-opt. Regardless compiling is trivial if you
follow the instructions.
Make sure you have the basic development packages installed.
Here is a good start:
http://askubuntu.com/questions/457526/how-to-install-cpuminer-in-ubuntu
Install any additional dependencies needed by cpuminer-opt. The list below
are some of the ones that may not be in the default install and need to
be installed manually. There may be others, read the error messages they
will give a clue as to the missing package.
The following command should install everything you need on Debian based
distributions such as Ubuntu:
sudo apt-get install build-essential libssl-dev libcurl4-openssl-dev libjansson-dev libgmp-dev automake zlib1g-dev
build-essential (Development Tools package group on Fedora)
automake
libjansson-dev
libgmp-dev
libcurl4-openssl-dev
libssl-dev
lib-thread
zlib1g-dev
SHA support on AMD Ryzen CPUs requires gcc version 5 or higher and
openssl 1.1.0e or higher. Add one of the following, depending on the
compiler version, to CFLAGS:
"-march=native" or "-march=znver1" or "-msha".
Due to poor AVX2 performance on Ryzen users should add -DRYZEN_ to CFLAGS
to override multiway AVX2 on algos with sha256, and use SHA instead.
Additional instructions for static compilalation can be found here:
https://lxadm.com/Static_compilation_of_cpuminer
Static builds should only considered in a homogeneous HW and SW environment.
Local builds will always have the best performance and compatibility.
Extract cpuminer source.
tar xvzf cpuminer-opt-x.y.z.tar.gz
cd cpuminer-opt-x.y.z
Run ./build.sh to build on Linux or execute the following commands.
./autogen.sh
CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
make
Start mining.
./cpuminer -a algo -o url -u username -p password
Windows
Precompiled Windows binaries are built on a Linux host using Mingw
with a more recent compiler than the following Windows hosted procedure.
Building on Windows prerequisites:
msys
mingw_w64
Visual C++ redistributable 2008 X64
openssl
Install msys and mingw_w64, only needed once.
Unpack msys into C:\msys or your preferred directory.
Install mingw_w64 from win-builds.
Follow instructions, check "msys or cygwin" and "x86_64" and accept default
existing msys instalation.
Open a msys shell by double clicking on msys.bat.
Note that msys shell uses linux syntax for file specifications, "C:\" is
mounted at "/c/".
Add mingw bin directory to PATH variable
PATH="/c/msys/opt/windows_64/bin/:$PATH"
Instalation complete, compile cpuminer-opt.
Unpack cpuminer-opt source files using tar from msys shell, or using 7zip
or similar Windows program.
In msys shell cd to miner directory.
cd /c/path/to/cpuminer-opt
Run build.sh to build on Windows or execute the following commands.
./autogen.sh
CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
make
Start mining
cpuminer.exe -a algo -o url -u user -p password
The following tips may be useful for older AMD CPUs.
AMD CPUs older than Steamroller, including Athlon x2 and Phenom II x4, are
not supported by cpuminer-opt due to an incompatible implementation of SSE2
on these CPUs. Some algos may crash the miner with an invalid instruction.
Users are recommended to use an unoptimized miner such as cpuminer-multi.
Some users with AMD CPUs without AES_NI have reported problems compiling
with build.sh or "-march=native". Problems have included compile errors
and poor performance. These users are recommended to compile manually
specifying "-march=btver1" on the configure command line.
Support for even older x86_64 without AES_NI or SSE2 is not availble.

173
INSTALL_WINDOWS Normal file
View File

@@ -0,0 +1,173 @@
Instructions for compiling cpuminer-opt for Windows.
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-64.
sudo apt-get install mingw-w64
2. Create a local library directory for packages to be compiled in the next
step. Recommended location is $HOME/usr/lib/
3. Download and build other packages for mingw that don't have a mingw64
version available in the repositories.
Download the following source code packages from their respective and
respected download locations, copy them to ~/usr/lib/ and uncompress them.
openssl
curl
gmp
In most cases the latest vesrion is ok but it's safest to download
the same major and minor version as included in your distribution.
Run the following commands or follow the supplied instructions.
Do not run "make install" unless you are using ~/usr/lib, which isn't
recommended.
Some instructions insist on running "make check". If make check fails
it may still work, YMMV.
You can speed up "make" by using all CPU cores available with "-j n" where
n is the number of CPU threads you want to use.
openssl:
./Configure mingw64 shared --cross-compile-prefix=x86_64-w64-mingw32
make
curl:
./configure --with-winssl --with-winidn --host=x86_64-w64-mingw32
make
gmp:
./configure --host=x86_64-w64-mingw32
make
4. Tweak the environment.
This step is required everytime you login or the commands can be added to
.bashrc.
Define some local variables to point to local library.
export LOCAL_LIB="$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"
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 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 /usr/lib/gcc/x86_64-w64-mingw32/7.3-win32/libstdc++-6.dll release/
cp /usr/lib/gcc/x86_64-w64-mingw32/7.3-win32/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. Prepare to compile
Create a link to the locally compiled version of gmp.h
ln -s $LOCAL_LIB/gmp-version/gmp.h ./gmp.h
Edit configure.ac to fix lipthread package name.
sed -i 's/"-lpthread"/"-lpthreadGC2"/g' configure.ac
7. Compile
you can use the default compile if you intend to use cpuminer-opt on the
same CPU and the virtual machine supports that architecture.
./build.sh
Otherwise you can compile manually while setting options in CFLAGS.
Some common options:
To compile for a specific CPU architecture:
CFLAGS="-O3 -march=znver1 -Wall" ./configure --with-curl
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 --with-curl
This will compile for an older CPU that does not have AVX.
You can find several examples in build-allarch.sh
If you have a CPU with more than 64 threads and Windows 7 or higher you
can enable the CPU Groups feature:
-D_WIN32_WINNT==0x0601
Once you have run configure successfully run make 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.

18
Makefile.am
View File

@@ -42,10 +42,11 @@ cpuminer_SOURCES = \
algo/argon2/argon2d/argon2d/argon2.c \
algo/argon2/argon2d/argon2d/core.c \
algo/argon2/argon2d/argon2d/opt.c \
algo/argon2/argon2d/argon2d/thread.c \
algo/argon2/argon2d/argon2d/argon2d_thread.c \
algo/argon2/argon2d/argon2d/encoding.c \
algo/blake/sph_blake.c \
algo/blake/blake-hash-4way.c \
algo/blake/blake256-hash-4way.c \
algo/blake/blake512-hash-4way.c \
algo/blake/blake-gate.c \
algo/blake/blake.c \
algo/blake/blake-4way.c \
@@ -136,6 +137,8 @@ cpuminer_SOURCES = \
algo/nist5/nist5-4way.c \
algo/nist5/nist5.c \
algo/nist5/zr5.c \
algo/panama/sph_panama.c \
algo/radiogatun/sph_radiogatun.c \
algo/pluck.c \
algo/quark/quark-gate.c \
algo/quark/quark.c \
@@ -167,6 +170,7 @@ cpuminer_SOURCES = \
algo/shabal/shabal-hash-4way.c \
algo/shavite/sph_shavite.c \
algo/shavite/sph-shavite-aesni.c \
algo/shavite/shavite-hash-2way.c \
algo/shavite/shavite.c \
algo/simd/sph_simd.c \
algo/simd/nist.c \
@@ -240,16 +244,20 @@ cpuminer_SOURCES = \
algo/x15/x15-gate.c \
algo/x15/x15.c \
algo/x15/x15-4way.c \
algo/x16/x16r-gate.c \
algo/x16/x16r.c \
algo/x16/x16r-4way.c \
algo/x17/x17-gate.c \
algo/x17/x17.c \
algo/x17/x17-4way.c \
algo/x17/xevan-gate.c \
algo/x17/xevan.c \
algo/x17/xevan-4way.c \
algo/x17/x16r-gate.c \
algo/x17/x16r.c \
algo/x17/x16r-4way.c \
algo/x17/hmq1725.c \
algo/x17/sonoa-gate.c \
algo/x17/sonoa-4way.c \
algo/x17/sonoa.c \
algo/x20/x20r.c \
algo/yescrypt/yescrypt.c \
algo/yescrypt/sha256_Y.c \
algo/yescrypt/yescrypt-best.c \

7
README.md
View File

@@ -16,7 +16,8 @@ https://bitcointalk.org/index.php?topic=1326803.0
mailto://jayddee246@gmail.com
See file RELEASE_NOTES for change log and compile instructions.
See file RELEASE_NOTES for change log and INSTALL_LINUX or INSTALL_WINDOWS
for compile instructions.
Requirements
------------
@@ -78,6 +79,7 @@ Supported Algorithms
lyra2h Hppcoin
lyra2re lyra2
lyra2rev2 lyra2v2, Vertcoin
lyra2rev3 lyrav2v3, Vertcoin
lyra2z Zcoin (XZC)
lyra2z330 Lyra2 330 rows, Zoin (ZOI)
m7m Magi (XMG)
@@ -100,6 +102,7 @@ Supported Algorithms
skein Skein+Sha (Skeincoin)
skein2 Double Skein (Woodcoin)
skunk Signatum (SIGT)
sonoa Sono
timetravel Machinecoin (MAC)
timetravel10 Bitcore
tribus Denarius (DNR)
@@ -130,6 +133,8 @@ Supported Algorithms
Errata
------
Cryptonight and variants are no longer supported, use another miner.
Neoscrypt crashes on Windows, use legacy version.
AMD CPUs older than Piledriver, including Athlon x2 and Phenom II x4, are not

11
README.txt
View File

@@ -15,7 +15,8 @@ the features listed at cpuminer startup to ensure you are mining at
optimum speed using all the available features.
Architecture names and compile options used are only provided for Intel
Core series. Pentium and Celeron often have fewer features.
Core series. Even the newest Pentium and Celeron CPUs are often missing
features.
AMD CPUs older than Piledriver, including Athlon x2 and Phenom II x4, are not
supported by cpuminer-opt due to an incompatible implementation of SSE2 on
@@ -27,17 +28,13 @@ Changes in v3.8.4 may have improved compatibility with some of these CPUs.
Exe name Compile flags Arch name
cpuminer-sse2.exe "-msse2" Core2, Nehalem
cpuminer-aes-sse42.exe "-march=westmere" Westmere, Sandy-Ivybridge
cpuminer-aes-sse42.exe "-march=westmere" Westmere
cpuminer-avx.exe "-march=corei7-avx" Sandy-Ivybridge
cpuminer-avx2.exe "-march=core-avx2" Haswell, Sky-Kaby-Coffeelake
cpuminer-avx2-sha.exe "-march=core-avx2 -msha" Ryzen
cpuminer-zen "-march=znver1 -DRYZEN_" Ryzen
If you like this software feel free to donate:
BTC: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT
ETH: 0x72122edabcae9d3f57eab0729305a425f6fef6d0
LTC: LdUwoHJnux9r9EKqFWNvAi45kQompHk6e8
BCH: 1QKYkB6atn4P7RFozyziAXLEnurwnUM1cQ
BTG: GVUyECtRHeC5D58z9F3nGGfVQndwnsPnHQ

184
RELEASE_NOTES
View File

@@ -1,11 +1,11 @@
puminer-opt now supports HW SHA acceleration available on AMD Ryzen CPUs.
cpuminer-opt is a console program run from the command line using the
keyboard, not the mouse.
cpuminer-opt now supports HW SHA acceleration available on AMD Ryzen CPUs.
This feature requires recent SW including GCC version 5 or higher and
openssl version 1.1 or higher. It may also require using "-march=znver1"
compile flag.
HW SHA support is only available when compiled from source, Windows binaries
are not yet available.
cpuminer-opt is a console program, if you're using a mouse you're doing it
wrong.
@@ -25,140 +25,62 @@ required.
Compile Instructions
--------------------
Requirements:
See INSTALL_LINUX or INSTALL_WINDOWS fror compile instruuctions
Requirements
------------
Intel Core2 or newer, or AMD Steamroller or newer CPU. ARM CPUs are not
supported.
64 bit Linux or Windows operating system. Apple is not supported.
Building on linux prerequisites:
It is assumed users know how to install packages on their system and
be able to compile standard source packages. This is basic Linux and
beyond the scope of cpuminer-opt.
Make sure you have the basic development packages installed.
Here is a good start:
http://askubuntu.com/questions/457526/how-to-install-cpuminer-in-ubuntu
Install any additional dependencies needed by cpuminer-opt. The list below
are some of the ones that may not be in the default install and need to
be installed manually. There may be others, read the error messages they
will give a clue as to the missing package.
The following command should install everything you need on Debian based
distributions such as Ubuntu:
sudo apt-get install build-essential libssl-dev libcurl4-openssl-dev libjansson-dev libgmp-dev automake zlib1g-dev
build-essential (for Ubuntu, Development Tools package group on Fedora)
automake
libjansson-dev
libgmp-dev
libcurl4-openssl-dev
libssl-dev
pthreads
zlib
SHA support on AMD Ryzen CPUs requires gcc version 5 or higher and openssl 1.1
or higher. Reports of improved performiance on Ryzen when using openssl 1.0.2
have been due to AVX and AVX2 optimizations added to that version.
Additional improvements are expected on Ryzen with openssl 1.1.
"-march-znver1" or "-msha".
Additional instructions for static compilalation can be found here:
https://lxadm.com/Static_compilation_of_cpuminer
Static builds should only considered in a homogeneous HW and SW environment.
Local builds will always have the best performance and compatibility.
Extract cpuminer source.
tar xvzf cpuminer-opt-x.y.z.tar.gz
cd cpuminer-opt-x.y.z
Run ./build.sh to build on Linux or execute the following commands.
./autogen.sh
CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
make
Additional optional compile flags, add the following to CFLAGS to activate:
-DUSE_SPH_SHA (deprecated)
SPH may give slightly better performance on algos that use sha256 when using
openssl 1.0.1 or older. Openssl 1.0.2 adds AVX2 and 1.1 adds SHA and perform
better than SPH. This option is ignored when 4-way is used, even for CPUs
with SHA.
Start mining.
./cpuminer -a algo -o url -u username -p password
Windows
Precompiled Windows binaries are built on a Linux host using Mingw
with a more recent compiler than the following Windows hosted procedure.
Building on Windows prerequisites:
msys
mingw_w64
Visual C++ redistributable 2008 X64
openssl
Install msys and mingw_w64, only needed once.
Unpack msys into C:\msys or your preferred directory.
Install mingw_w64 from win-builds.
Follow instructions, check "msys or cygwin" and "x86_64" and accept default
existing msys instalation.
Open a msys shell by double clicking on msys.bat.
Note that msys shell uses linux syntax for file specifications, "C:\" is
mounted at "/c/".
Add mingw bin directory to PATH variable
PATH="/c/msys/opt/windows_64/bin/:$PATH"
Instalation complete, compile cpuminer-opt.
Unpack cpuminer-opt source files using tar from msys shell, or using 7zip
or similar Windows program.
In msys shell cd to miner directory.
cd /c/path/to/cpuminer-opt
Run build.sh to build on Windows or execute the following commands.
./autogen.sh
CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
make
Start mining
cpuminer.exe -a algo -o url -u user -p password
The following tips may be useful for older AMD CPUs.
AMD CPUs older than Steamroller, including Athlon x2 and Phenom II x4, are
not supported by cpuminer-opt due to an incompatible implementation of SSE2
on these CPUs. Some algos may crash the miner with an invalid instruction.
Users are recommended to use an unoptimized miner such as cpuminer-multi.
Some users with AMD CPUs without AES_NI have reported problems compiling
with build.sh or "-march=native". Problems have included compile errors
and poor performance. These users are recommended to compile manually
specifying "-march=btver1" on the configure command line.
Support for even older x86_64 without AES_NI or SSE2 is not availble.
Change Log
----------
v3.9.1
Fixed AVX2 version of anime algo.
Added sonoa algo.
Added "-DRYZEN_" compile option for Ryzen to override 4-way hashing when algo
contains sha256 and use SHA instead. This is due to a combination of
the introduction of HW SHA support combined with the poor performance
of AVX2 on Ryzen. The Windows binaries package replaces cpuminer-avx2-sha
with cpuminer-zen compiled with the override. Refer to the build instructions
for more information.
Ongoing restructuring to streamline the process, reduce latency,
reduce memory usage and unnecessary copying of data. Most of these
will not result in a notoceably higher reported hashrate as the
change simply reduces the time wasted that wasn't factored into the
hash rate reported by the miner. In short, less dead time resulting in
a higher net hashrate.
One of these measures to reduce latency also results in an enhanced
share submission message including the share number*, the CPU thread,
and the vector lane that found the solution. The time difference between
the share submission and acceptance (or rejection) response indicates
network ltatency. One other effect of this change is a reduction in hash
meter messages because the scan function no longer exits when a share is
found. Scan cycles will go longer and submit multiple shares per cycle.
*the share number is antcipated and includes both accepted and rejected
shares. Because the share is antipated and not synchronized it may be
incorrect in time of very rapid share submission. Under most conditions
it should be easy to match the submission with the corresponding response.
Removed "-DUSE_SPH_SHA" option, all users should have a recent version of
openssl installed: v1.0.2 (Ubuntu 16.04) or better. Ryzen SHA requires
v1.1.0 or better. Ryzen SHA is not used when hashing multi-way parallel.
Ryzen SHA is available in the Windows binaries release package.
Improved compile instructions, now in seperate files: INSTALL_LINUX and
INSTALL_WINDOWS. The Windows instructions are used to build the binaries
release package. It's built on a Linux system either running as a virtual
machine or a seperate computer. At this time there is no known way to
build natively on a Windows system.
v3.9.0.1
Isolate Windows CPU groups code when CPU groups support not explicitly defined.
@@ -171,6 +93,7 @@ Prep work for AVX512.
Added lyra2rev3 for the vertcoin algo change.
Added yespower, yespowerr16 (Yenten)
Added phi2 algo for LUX
Discontinued support for cryptonight and variants.
v3.8.8.1
@@ -354,6 +277,7 @@ Changed default sha256 and sha512 to openssl. This should be used when
compiling with openssl 1.0.2 or higher (Ubuntu 16.04).
This should increase the hashrate for yescrypt, yescryptr16, m7m, xevan, skein,
myr-gr & others when openssl 1.0.2 is installed.
Note: -DUSE_SPH_SHA has been removed in v3.9.1.
Users with openssl 1.0.1 (Ubuntu 14.04) may get better perforance by adding
"-DUSE_SPH_SHA" to CLAGS.
Windows binaries are compiled with -DUSE_SPH_SHA and won't get the speedup.

7
algo-gate-api.c
View File

@@ -214,6 +214,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
case ALGO_SKEIN: register_skein_algo ( gate ); break;
case ALGO_SKEIN2: register_skein2_algo ( gate ); break;
case ALGO_SKUNK: register_skunk_algo ( gate ); break;
case ALGO_SONOA: register_sonoa_algo ( gate ); break;
case ALGO_TIMETRAVEL: register_timetravel_algo ( gate ); break;
case ALGO_TIMETRAVEL10: register_timetravel10_algo ( gate ); break;
case ALGO_TRIBUS: register_tribus_algo ( gate ); break;
@@ -266,6 +267,10 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
// override std defaults with jr2 defaults
bool register_json_rpc2( algo_gate_t *gate )
{
applog(LOG_WARNING,"\nCryptonight algorithm and variants are no longer");
applog(LOG_WARNING,"supported by cpuminer-opt. Shares submitted will");
applog(LOG_WARNING,"likely be rejected. Proceed at your own risk.\n");
gate->wait_for_diff = (void*)&do_nothing;
gate->get_new_work = (void*)&jr2_get_new_work;
gate->get_nonceptr = (void*)&jr2_get_nonceptr;
@@ -354,3 +359,5 @@ void get_algo_alias( char** algo_or_alias )
}
}
#undef ALIAS
#undef PROPER

14
algo-gate-api.h
View File

@@ -109,8 +109,15 @@ inline bool set_excl ( set_t a, set_t b ) { return (a & b) == 0; }
typedef struct
{
// special case, only one target, provides a callback for scanhash to
// submit work with less overhead.
// bool (*submit_work ) ( struct thr_info*, const struct work* );
// mandatory functions, must be overwritten
int ( *scanhash ) ( int, struct work*, uint32_t, uint64_t* );
// Added a 5th arg for the thread_info structure to replace the int thr id
// in the first arg. Both will co-exist during the trasition.
//int ( *scanhash ) ( int, struct work*, uint32_t, uint64_t* );
int ( *scanhash ) ( int, struct work*, uint32_t, uint64_t*, struct thr_info* );
// optional unsafe, must be overwritten if algo uses function
void ( *hash ) ( void*, const void*, uint32_t ) ;
@@ -188,6 +195,11 @@ void four_way_not_tested();
// allways returns failure
int null_scanhash();
// The one and only, a callback for scanhash.
bool submit_work( struct thr_info *thr, const struct work *work_in );
// displays warning
void null_hash ();
void null_hash_suw();

2
algo/argon2/argon2d/argon2d/thread.c → algo/argon2/argon2d/argon2d/argon2d_thread.c
View File

@@ -17,7 +17,7 @@
#if !defined(ARGON2_NO_THREADS)
#include "thread.h"
#include "argon2d_thread.h"
#if defined(_WIN32)
#include <windows.h>
#endif

0
algo/argon2/argon2d/argon2d/thread.h → algo/argon2/argon2d/argon2d/argon2d_thread.h
View File

2
algo/argon2/argon2d/argon2d/core.c
View File

@@ -30,7 +30,7 @@
#include <string.h>
#include "core.h"
#include "thread.h"
#include "argon2d_thread.h"
#include "../blake2/blake2.h"
#include "../blake2/blake2-impl.h"

27
algo/blake/blake-hash-4way.h
View File

@@ -37,7 +37,7 @@
#ifndef __BLAKE_HASH_4WAY__
#define __BLAKE_HASH_4WAY__ 1
#ifdef __SSE4_2__
//#ifdef __SSE4_2__
#ifdef __cplusplus
extern "C"{
@@ -57,19 +57,22 @@ extern "C"{
// Blake-256 4 way
typedef struct {
__m128i buf[16] __attribute__ ((aligned (64)));
__m128i H[8];
__m128i S[4];
unsigned char buf[64<<2];
uint32_t H[8<<2];
uint32_t S[4<<2];
// __m128i buf[16] __attribute__ ((aligned (64)));
// __m128i H[8];
// __m128i S[4];
size_t ptr;
sph_u32 T0, T1;
uint32_t T0, T1;
int rounds; // 14 for blake, 8 for blakecoin & vanilla
} blake_4way_small_context;
} blake_4way_small_context __attribute__ ((aligned (64)));
// Default 14 rounds
typedef blake_4way_small_context blake256_4way_context;
void blake256_4way_init(void *cc);
void blake256_4way(void *cc, const void *data, size_t len);
void blake256_4way_close(void *cc, void *dst);
void blake256_4way_init(void *ctx);
void blake256_4way(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;
@@ -132,12 +135,10 @@ void blake512_4way_close(void *cc, void *dst);
void blake512_4way_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
#endif
#endif // AVX2
#ifdef __cplusplus
}
#endif
#endif
#endif
#endif // BLAKE_HASH_4WAY_H__

798
algo/blake/blake-hash-4way.c → algo/blake/blake256-hash-4way.c
View File

@@ -30,9 +30,10 @@
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#if defined (__SSE4_2__)
//#if defined (__SSE4_2__)
#include <stddef.h>
#include <stdint.h>
#include <string.h>
#include <limits.h>
@@ -60,26 +61,12 @@ extern "C"{
// Blake-256
static const sph_u32 IV256[8] = {
SPH_C32(0x6A09E667), SPH_C32(0xBB67AE85),
SPH_C32(0x3C6EF372), SPH_C32(0xA54FF53A),
SPH_C32(0x510E527F), SPH_C32(0x9B05688C),
SPH_C32(0x1F83D9AB), SPH_C32(0x5BE0CD19)
static const uint32_t IV256[8] =
{
0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19
};
#if defined (__AVX2__)
// Blake-512
static const sph_u64 IV512[8] = {
SPH_C64(0x6A09E667F3BCC908), SPH_C64(0xBB67AE8584CAA73B),
SPH_C64(0x3C6EF372FE94F82B), SPH_C64(0xA54FF53A5F1D36F1),
SPH_C64(0x510E527FADE682D1), SPH_C64(0x9B05688C2B3E6C1F),
SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
};
#endif
#if SPH_COMPACT_BLAKE_32 || SPH_COMPACT_BLAKE_64
// Blake-256 4 & 8 way, Blake-512 4 way
@@ -317,47 +304,6 @@ static const sph_u32 CS[16] = {
#endif
#if defined(__AVX2__)
// Blake-512 4 way
#define CBx(r, i) CBx_(Z ## r ## i)
#define CBx_(n) CBx__(n)
#define CBx__(n) CB ## n
#define CB0 SPH_C64(0x243F6A8885A308D3)
#define CB1 SPH_C64(0x13198A2E03707344)
#define CB2 SPH_C64(0xA4093822299F31D0)
#define CB3 SPH_C64(0x082EFA98EC4E6C89)
#define CB4 SPH_C64(0x452821E638D01377)
#define CB5 SPH_C64(0xBE5466CF34E90C6C)
#define CB6 SPH_C64(0xC0AC29B7C97C50DD)
#define CB7 SPH_C64(0x3F84D5B5B5470917)
#define CB8 SPH_C64(0x9216D5D98979FB1B)
#define CB9 SPH_C64(0xD1310BA698DFB5AC)
#define CBA SPH_C64(0x2FFD72DBD01ADFB7)
#define CBB SPH_C64(0xB8E1AFED6A267E96)
#define CBC SPH_C64(0xBA7C9045F12C7F99)
#define CBD SPH_C64(0x24A19947B3916CF7)
#define CBE SPH_C64(0x0801F2E2858EFC16)
#define CBF SPH_C64(0x636920D871574E69)
#if SPH_COMPACT_BLAKE_64
// not used
static const sph_u64 CB[16] = {
SPH_C64(0x243F6A8885A308D3), SPH_C64(0x13198A2E03707344),
SPH_C64(0xA4093822299F31D0), SPH_C64(0x082EFA98EC4E6C89),
SPH_C64(0x452821E638D01377), SPH_C64(0xBE5466CF34E90C6C),
SPH_C64(0xC0AC29B7C97C50DD), SPH_C64(0x3F84D5B5B5470917),
SPH_C64(0x9216D5D98979FB1B), SPH_C64(0xD1310BA698DFB5AC),
SPH_C64(0x2FFD72DBD01ADFB7), SPH_C64(0xB8E1AFED6A267E96),
SPH_C64(0xBA7C9045F12C7F99), SPH_C64(0x24A19947B3916CF7),
SPH_C64(0x0801F2E2858EFC16), SPH_C64(0x636920D871574E69)
};
#endif
#endif
#define GS_4WAY( m0, m1, c0, c1, a, b, c, d ) \
do { \
@@ -411,125 +357,41 @@ do { \
#endif
#if defined (__AVX2__)
// Blake-256 8 way
#define GS_8WAY( m0, m1, c0, c1, a, b, c, d ) \
do { \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set1_epi32( c1 ), m0 ), b ), a ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set1_epi32( c0 ), m1 ), b ), a ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
} while (0)
#define ROUND_S_8WAY(r) do { \
GS_8WAY(Mx(r, 0), Mx(r, 1), CSx(r, 0), CSx(r, 1), V0, V4, V8, VC); \
GS_8WAY(Mx(r, 2), Mx(r, 3), CSx(r, 2), CSx(r, 3), V1, V5, V9, VD); \
GS_8WAY(Mx(r, 4), Mx(r, 5), CSx(r, 4), CSx(r, 5), V2, V6, VA, VE); \
GS_8WAY(Mx(r, 6), Mx(r, 7), CSx(r, 6), CSx(r, 7), V3, V7, VB, VF); \
GS_8WAY(Mx(r, 8), Mx(r, 9), CSx(r, 8), CSx(r, 9), V0, V5, VA, VF); \
GS_8WAY(Mx(r, A), Mx(r, B), CSx(r, A), CSx(r, B), V1, V6, VB, VC); \
GS_8WAY(Mx(r, C), Mx(r, D), CSx(r, C), CSx(r, D), V2, V7, V8, VD); \
GS_8WAY(Mx(r, E), Mx(r, F), CSx(r, E), CSx(r, F), V3, V4, V9, VE); \
} while (0)
// Blake-512 4 way
#define GB_4WAY(m0, m1, c0, c1, a, b, c, d) do { \
a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
_mm256_set_epi64x( c1, c1, c1, c1 ), m0 ), b ), a ); \
d = mm256_ror_64( _mm256_xor_si256( d, a ), 32 ); \
c = _mm256_add_epi64( c, d ); \
b = mm256_ror_64( _mm256_xor_si256( b, c ), 25 ); \
a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
_mm256_set_epi64x( c0, c0, c0, c0 ), m1 ), b ), a ); \
d = mm256_ror_64( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi64( c, d ); \
b = mm256_ror_64( _mm256_xor_si256( b, c ), 11 ); \
} while (0)
#if SPH_COMPACT_BLAKE_64
// not used
#define ROUND_B_4WAY(r) do { \
GB_4WAY(M[sigma[r][0x0]], M[sigma[r][0x1]], \
CB[sigma[r][0x0]], CB[sigma[r][0x1]], V0, V4, V8, VC); \
GB_4WAY(M[sigma[r][0x2]], M[sigma[r][0x3]], \
CB[sigma[r][0x2]], CB[sigma[r][0x3]], V1, V5, V9, VD); \
GB_4WAY(M[sigma[r][0x4]], M[sigma[r][0x5]], \
CB[sigma[r][0x4]], CB[sigma[r][0x5]], V2, V6, VA, VE); \
GB_4WAY(M[sigma[r][0x6]], M[sigma[r][0x7]], \
CB[sigma[r][0x6]], CB[sigma[r][0x7]], V3, V7, VB, VF); \
GB_4WAY(M[sigma[r][0x8]], M[sigma[r][0x9]], \
CB[sigma[r][0x8]], CB[sigma[r][0x9]], V0, V5, VA, VF); \
GB_4WAY(M[sigma[r][0xA]], M[sigma[r][0xB]], \
CB[sigma[r][0xA]], CB[sigma[r][0xB]], V1, V6, VB, VC); \
GB_4WAY(M[sigma[r][0xC]], M[sigma[r][0xD]], \
CB[sigma[r][0xC]], CB[sigma[r][0xD]], V2, V7, V8, VD); \
GB_4WAY(M[sigma[r][0xE]], M[sigma[r][0xF]], \
CB[sigma[r][0xE]], CB[sigma[r][0xF]], V3, V4, V9, VE); \
} while (0)
#else
//current_impl
#define ROUND_B_4WAY(r) do { \
GB_4WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
GB_4WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
GB_4WAY(Mx(r, 4), Mx(r, 5), CBx(r, 4), CBx(r, 5), V2, V6, VA, VE); \
GB_4WAY(Mx(r, 6), Mx(r, 7), CBx(r, 6), CBx(r, 7), V3, V7, VB, VF); \
GB_4WAY(Mx(r, 8), Mx(r, 9), CBx(r, 8), CBx(r, 9), V0, V5, VA, VF); \
GB_4WAY(Mx(r, A), Mx(r, B), CBx(r, A), CBx(r, B), V1, V6, VB, VC); \
GB_4WAY(Mx(r, C), Mx(r, D), CBx(r, C), CBx(r, D), V2, V7, V8, VD); \
GB_4WAY(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
} while (0)
#endif
#endif
// Blake-256 4 way
#define DECL_STATE32_4WAY \
__m128i H0, H1, H2, H3, H4, H5, H6, H7; \
__m128i S0, S1, S2, S3; \
sph_u32 T0, T1;
uint32_t T0, T1;
#define READ_STATE32_4WAY(state) do { \
H0 = (state)->H[0]; \
H1 = (state)->H[1]; \
H2 = (state)->H[2]; \
H3 = (state)->H[3]; \
H4 = (state)->H[4]; \
H5 = (state)->H[5]; \
H6 = (state)->H[6]; \
H7 = (state)->H[7]; \
S0 = (state)->S[0]; \
S1 = (state)->S[1]; \
S2 = (state)->S[2]; \
S3 = (state)->S[3]; \
H0 = casti_m128i( state->H, 0 ); \
H1 = casti_m128i( state->H, 1 ); \
H2 = casti_m128i( state->H, 2 ); \
H3 = casti_m128i( state->H, 3 ); \
H4 = casti_m128i( state->H, 4 ); \
H5 = casti_m128i( state->H, 5 ); \
H6 = casti_m128i( state->H, 6 ); \
H7 = casti_m128i( state->H, 7 ); \
S0 = casti_m128i( state->S, 0 ); \
S1 = casti_m128i( state->S, 1 ); \
S2 = casti_m128i( state->S, 2 ); \
S3 = casti_m128i( state->S, 3 ); \
T0 = (state)->T0; \
T1 = (state)->T1; \
} while (0)
#define WRITE_STATE32_4WAY(state) do { \
(state)->H[0] = H0; \
(state)->H[1] = H1; \
(state)->H[2] = H2; \
(state)->H[3] = H3; \
(state)->H[4] = H4; \
(state)->H[5] = H5; \
(state)->H[6] = H6; \
(state)->H[7] = H7; \
(state)->S[0] = S0; \
(state)->S[1] = S1; \
(state)->S[2] = S2; \
(state)->S[3] = S3; \
casti_m128i( state->H, 0 ) = H0; \
casti_m128i( state->H, 1 ) = H1; \
casti_m128i( state->H, 2 ) = H2; \
casti_m128i( state->H, 3 ) = H3; \
casti_m128i( state->H, 4 ) = H4; \
casti_m128i( state->H, 5 ) = H5; \
casti_m128i( state->H, 6 ) = H6; \
casti_m128i( state->H, 7 ) = H7; \
casti_m128i( state->S, 0 ) = S0; \
casti_m128i( state->S, 1 ) = S1; \
casti_m128i( state->S, 2 ) = S2; \
casti_m128i( state->S, 3 ) = S3; \
(state)->T0 = T0; \
(state)->T1 = T1; \
} while (0)
@@ -616,30 +478,30 @@ do { \
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm_xor_si128( S0, _mm_set_epi32( CS0, CS0, CS0, CS0 ) ); \
V9 = _mm_xor_si128( S1, _mm_set_epi32( CS1, CS1, CS1, CS1 ) ); \
VA = _mm_xor_si128( S2, _mm_set_epi32( CS2, CS2, CS2, CS2 ) ); \
VB = _mm_xor_si128( S3, _mm_set_epi32( CS3, CS3, CS3, CS3 ) ); \
V8 = _mm_xor_si128( S0, _mm_set1_epi32( CS0 ) ); \
V9 = _mm_xor_si128( S1, _mm_set1_epi32( CS1 ) ); \
VA = _mm_xor_si128( S2, _mm_set1_epi32( CS2 ) ); \
VB = _mm_xor_si128( S3, _mm_set1_epi32( CS3 ) ); \
VC = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS4 ) ); \
VD = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS5 ) ); \
VE = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS6 ) ); \
VF = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS7 ) ); \
M0 = mm128_bswap_32( * buf ); \
M1 = mm128_bswap_32( *(buf+1) ); \
M2 = mm128_bswap_32( *(buf+2) ); \
M3 = mm128_bswap_32( *(buf+3) ); \
M4 = mm128_bswap_32( *(buf+4) ); \
M5 = mm128_bswap_32( *(buf+5) ); \
M6 = mm128_bswap_32( *(buf+6) ); \
M7 = mm128_bswap_32( *(buf+7) ); \
M8 = mm128_bswap_32( *(buf+8) ); \
M9 = mm128_bswap_32( *(buf+9) ); \
MA = mm128_bswap_32( *(buf+10) ); \
MB = mm128_bswap_32( *(buf+11) ); \
MC = mm128_bswap_32( *(buf+12) ); \
MD = mm128_bswap_32( *(buf+13) ); \
ME = mm128_bswap_32( *(buf+14) ); \
MF = mm128_bswap_32( *(buf+15) ); \
M0 = mm128_bswap_32( buf[ 0] ); \
M1 = mm128_bswap_32( buf[ 1] ); \
M2 = mm128_bswap_32( buf[ 2] ); \
M3 = mm128_bswap_32( buf[ 3] ); \
M4 = mm128_bswap_32( buf[ 4] ); \
M5 = mm128_bswap_32( buf[ 5] ); \
M6 = mm128_bswap_32( buf[ 6] ); \
M7 = mm128_bswap_32( buf[ 7] ); \
M8 = mm128_bswap_32( buf[ 8] ); \
M9 = mm128_bswap_32( buf[ 9] ); \
MA = mm128_bswap_32( buf[10] ); \
MB = mm128_bswap_32( buf[11] ); \
MC = mm128_bswap_32( buf[12] ); \
MD = mm128_bswap_32( buf[13] ); \
ME = mm128_bswap_32( buf[14] ); \
MF = mm128_bswap_32( buf[15] ); \
ROUND_S_4WAY(0); \
ROUND_S_4WAY(1); \
ROUND_S_4WAY(2); \
@@ -673,6 +535,31 @@ do { \
// Blake-256 8 way
#define GS_8WAY( m0, m1, c0, c1, a, b, c, d ) \
do { \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set1_epi32( c1 ), m0 ), b ), a ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set1_epi32( c0 ), m1 ), b ), a ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
} while (0)
#define ROUND_S_8WAY(r) do { \
GS_8WAY(Mx(r, 0), Mx(r, 1), CSx(r, 0), CSx(r, 1), V0, V4, V8, VC); \
GS_8WAY(Mx(r, 2), Mx(r, 3), CSx(r, 2), CSx(r, 3), V1, V5, V9, VD); \
GS_8WAY(Mx(r, 4), Mx(r, 5), CSx(r, 4), CSx(r, 5), V2, V6, VA, VE); \
GS_8WAY(Mx(r, 6), Mx(r, 7), CSx(r, 6), CSx(r, 7), V3, V7, VB, VF); \
GS_8WAY(Mx(r, 8), Mx(r, 9), CSx(r, 8), CSx(r, 9), V0, V5, VA, VF); \
GS_8WAY(Mx(r, A), Mx(r, B), CSx(r, A), CSx(r, B), V1, V6, VB, VC); \
GS_8WAY(Mx(r, C), Mx(r, D), CSx(r, C), CSx(r, D), V2, V7, V8, VD); \
GS_8WAY(Mx(r, E), Mx(r, F), CSx(r, E), CSx(r, F), V3, V4, V9, VE); \
} while (0)
#define DECL_STATE32_8WAY \
__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
__m256i S0, S1, S2, S3; \
@@ -787,312 +674,136 @@ do { \
S3 ), H7 ); \
} while (0)
// Blake-512 4 way
#define DECL_STATE64_4WAY \
__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
__m256i S0, S1, S2, S3; \
sph_u64 T0, T1;
#define READ_STATE64_4WAY(state) do { \
H0 = (state)->H[0]; \
H1 = (state)->H[1]; \
H2 = (state)->H[2]; \
H3 = (state)->H[3]; \
H4 = (state)->H[4]; \
H5 = (state)->H[5]; \
H6 = (state)->H[6]; \
H7 = (state)->H[7]; \
S0 = (state)->S[0]; \
S1 = (state)->S[1]; \
S2 = (state)->S[2]; \
S3 = (state)->S[3]; \
T0 = (state)->T0; \
T1 = (state)->T1; \
} while (0)
#define WRITE_STATE64_4WAY(state) do { \
(state)->H[0] = H0; \
(state)->H[1] = H1; \
(state)->H[2] = H2; \
(state)->H[3] = H3; \
(state)->H[4] = H4; \
(state)->H[5] = H5; \
(state)->H[6] = H6; \
(state)->H[7] = H7; \
(state)->S[0] = S0; \
(state)->S[1] = S1; \
(state)->S[2] = S2; \
(state)->S[3] = S3; \
(state)->T0 = T0; \
(state)->T1 = T1; \
} while (0)
#if SPH_COMPACT_BLAKE_64
// not used
#define COMPRESS64_4WAY do { \
__m256i M[16]; \
__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
__m256i V8, V9, VA, VB, VC, VD, VE, VF; \
unsigned r; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = H4; \
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm256_xor_si256( S0, _mm256_set_epi64x( CB0, CB0, CB0, CB0 ) ); \
V9 = _mm256_xor_si256( S1, _mm256_set_epi64x( CB1, CB1, CB1, CB1 ) ); \
VA = _mm256_xor_si256( S2, _mm256_set_epi64x( CB2, CB2, CB2, CB2 ) ); \
VB = _mm256_xor_si256( S3, _mm256_set_epi64x( CB3, CB3, CB3, CB3 ) ); \
VC = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB4, CB4, CB4, CB4 ) ); \
VD = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB5, CB5, CB5, CB5 ) ); \
VE = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB6, CB6, CB6, CB6 ) ); \
VF = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB7, CB7, CB7, CB7 ) ); \
M[0x0] = mm256_bswap_64( *(buf+0) ); \
M[0x1] = mm256_bswap_64( *(buf+1) ); \
M[0x2] = mm256_bswap_64( *(buf+2) ); \
M[0x3] = mm256_bswap_64( *(buf+3) ); \
M[0x4] = mm256_bswap_64( *(buf+4) ); \
M[0x5] = mm256_bswap_64( *(buf+5) ); \
M[0x6] = mm256_bswap_64( *(buf+6) ); \
M[0x7] = mm256_bswap_64( *(buf+7) ); \
M[0x8] = mm256_bswap_64( *(buf+8) ); \
M[0x9] = mm256_bswap_64( *(buf+9) ); \
M[0xA] = mm256_bswap_64( *(buf+10) ); \
M[0xB] = mm256_bswap_64( *(buf+11) ); \
M[0xC] = mm256_bswap_64( *(buf+12) ); \
M[0xD] = mm256_bswap_64( *(buf+13) ); \
M[0xE] = mm256_bswap_64( *(buf+14) ); \
M[0xF] = mm256_bswap_64( *(buf+15) ); \
for (r = 0; r < 16; r ++) \
ROUND_B_4WAY(r); \
H0 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V2 ), VA ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V3 ), VB ), H3 ); \
H4 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V4 ), VC ), H4 ); \
H5 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V5 ), VD ), H5 ); \
H6 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V6 ), VE ), H6 ); \
H7 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V7 ), VF ), H7 ); \
} while (0)
#else
//current impl
#define COMPRESS64_4WAY do { \
__m256i M0, M1, M2, M3, M4, M5, M6, M7; \
__m256i M8, M9, MA, MB, MC, MD, ME, MF; \
__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
__m256i V8, V9, VA, VB, VC, VD, VE, VF; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = H4; \
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm256_xor_si256( S0, _mm256_set_epi64x( CB0, CB0, CB0, CB0 ) ); \
V9 = _mm256_xor_si256( S1, _mm256_set_epi64x( CB1, CB1, CB1, CB1 ) ); \
VA = _mm256_xor_si256( S2, _mm256_set_epi64x( CB2, CB2, CB2, CB2 ) ); \
VB = _mm256_xor_si256( S3, _mm256_set_epi64x( CB3, CB3, CB3, CB3 ) ); \
VC = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB4, CB4, CB4, CB4 ) ); \
VD = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB5, CB5, CB5, CB5 ) ); \
VE = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB6, CB6, CB6, CB6 ) ); \
VF = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB7, CB7, CB7, CB7 ) ); \
M0 = mm256_bswap_64( *(buf + 0) ); \
M1 = mm256_bswap_64( *(buf + 1) ); \
M2 = mm256_bswap_64( *(buf + 2) ); \
M3 = mm256_bswap_64( *(buf + 3) ); \
M4 = mm256_bswap_64( *(buf + 4) ); \
M5 = mm256_bswap_64( *(buf + 5) ); \
M6 = mm256_bswap_64( *(buf + 6) ); \
M7 = mm256_bswap_64( *(buf + 7) ); \
M8 = mm256_bswap_64( *(buf + 8) ); \
M9 = mm256_bswap_64( *(buf + 9) ); \
MA = mm256_bswap_64( *(buf + 10) ); \
MB = mm256_bswap_64( *(buf + 11) ); \
MC = mm256_bswap_64( *(buf + 12) ); \
MD = mm256_bswap_64( *(buf + 13) ); \
ME = mm256_bswap_64( *(buf + 14) ); \
MF = mm256_bswap_64( *(buf + 15) ); \
ROUND_B_4WAY(0); \
ROUND_B_4WAY(1); \
ROUND_B_4WAY(2); \
ROUND_B_4WAY(3); \
ROUND_B_4WAY(4); \
ROUND_B_4WAY(5); \
ROUND_B_4WAY(6); \
ROUND_B_4WAY(7); \
ROUND_B_4WAY(8); \
ROUND_B_4WAY(9); \
ROUND_B_4WAY(0); \
ROUND_B_4WAY(1); \
ROUND_B_4WAY(2); \
ROUND_B_4WAY(3); \
ROUND_B_4WAY(4); \
ROUND_B_4WAY(5); \
H0 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V2 ), VA ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V3 ), VB ), H3 ); \
H4 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V4 ), VC ), H4 ); \
H5 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V5 ), VD ), H5 ); \
H6 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V6 ), VE ), H6 ); \
H7 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V7 ), VF ), H7 ); \
} while (0)
#endif
#endif
// Blake-256 4 way
static const sph_u32 salt_zero_4way_small[4] = { 0, 0, 0, 0 };
static const uint32_t salt_zero_4way_small[4] = { 0, 0, 0, 0 };
static void
blake32_4way_init( blake_4way_small_context *sc, const sph_u32 *iv,
const sph_u32 *salt, int rounds )
blake32_4way_init( blake_4way_small_context *ctx, const uint32_t *iv,
const uint32_t *salt, int rounds )
{
int i;
for ( i = 0; i < 8; i++ )
sc->H[i] = _mm_set1_epi32( iv[i] );
for ( i = 0; i < 4; i++ )
sc->S[i] = _mm_set1_epi32( salt[i] );
sc->T0 = sc->T1 = 0;
sc->ptr = 0;
sc->rounds = rounds;
casti_m128i( ctx->H, 0 ) = _mm_set1_epi32( iv[0] );
casti_m128i( ctx->H, 1 ) = _mm_set1_epi32( iv[1] );
casti_m128i( ctx->H, 2 ) = _mm_set1_epi32( iv[2] );
casti_m128i( ctx->H, 3 ) = _mm_set1_epi32( iv[3] );
casti_m128i( ctx->H, 4 ) = _mm_set1_epi32( iv[4] );
casti_m128i( ctx->H, 5 ) = _mm_set1_epi32( iv[5] );
casti_m128i( ctx->H, 6 ) = _mm_set1_epi32( iv[6] );
casti_m128i( ctx->H, 7 ) = _mm_set1_epi32( iv[7] );
casti_m128i( ctx->S, 0 ) = m128_zero;
casti_m128i( ctx->S, 1 ) = m128_zero;
casti_m128i( ctx->S, 2 ) = m128_zero;
casti_m128i( ctx->S, 3 ) = m128_zero;
/*
sc->S[0] = _mm_set1_epi32( salt[0] );
sc->S[1] = _mm_set1_epi32( salt[1] );
sc->S[2] = _mm_set1_epi32( salt[2] );
sc->S[3] = _mm_set1_epi32( salt[3] );
*/
ctx->T0 = ctx->T1 = 0;
ctx->ptr = 0;
ctx->rounds = rounds;
}
static void
blake32_4way( blake_4way_small_context *sc, const void *data, size_t len )
blake32_4way( blake_4way_small_context *ctx, const void *data, size_t len )
{
__m128i *vdata = (__m128i*)data;
__m128i *buf;
size_t ptr;
const int buf_size = 64; // number of elements, sizeof/4
__m128i *buf = (__m128i*)ctx->buf;
size_t bptr = ctx->ptr<<2;
size_t vptr = ctx->ptr >> 2;
size_t blen = len << 2;
DECL_STATE32_4WAY
buf = sc->buf;
ptr = sc->ptr;
if ( len < buf_size - ptr )
if ( blen < (sizeof ctx->buf) - bptr )
{
memcpy_128( buf + (ptr>>2), vdata, len>>2 );
ptr += len;
sc->ptr = ptr;
memcpy( buf + vptr, data, (sizeof ctx->buf) - bptr );
bptr += blen;
ctx->ptr = bptr>>2;
return;
}
READ_STATE32_4WAY(sc);
while ( len > 0 )
READ_STATE32_4WAY( ctx );
while ( blen > 0 )
{
size_t clen;
size_t clen = ( sizeof ctx->buf ) - bptr;
clen = buf_size - ptr;
if ( clen > len )
clen = len;
memcpy_128( buf + (ptr>>2), vdata, clen>>2 );
ptr += clen;
vdata += (clen>>2);
len -= clen;
if ( ptr == buf_size )
if ( clen > blen )
clen = blen;
memcpy( buf + vptr, data, clen );
bptr += clen;
data = (const unsigned char *)data + clen;
blen -= clen;
if ( bptr == ( sizeof ctx->buf ) )
{
if ( ( T0 = SPH_T32(T0 + 512) ) < 512 )
T1 = SPH_T32(T1 + 1);
COMPRESS32_4WAY( sc->rounds );
ptr = 0;
if ( ( T0 = T0 + 512 ) < 512 )
T1 = T1 + 1;
COMPRESS32_4WAY( ctx->rounds );
bptr = 0;
}
}
WRITE_STATE32_4WAY(sc);
sc->ptr = ptr;
WRITE_STATE32_4WAY( ctx );
ctx->ptr = bptr>>2;
}
static void
blake32_4way_close( blake_4way_small_context *sc, unsigned ub, unsigned n,
blake32_4way_close( blake_4way_small_context *ctx, unsigned ub, unsigned n,
void *dst, size_t out_size_w32 )
{
// union {
__m128i buf[16];
// sph_u32 dummy;
// } u;
size_t ptr, k;
unsigned bit_len;
sph_u32 th, tl;
__m128i *out;
ptr = sc->ptr;
bit_len = ((unsigned)ptr << 3);
buf[ptr>>2] = _mm_set1_epi32( 0x80 );
tl = sc->T0 + bit_len;
th = sc->T1;
__m128i buf[16] __attribute__ ((aligned (64)));
size_t ptr = ctx->ptr;
size_t vptr = ctx->ptr>>2;
unsigned bit_len = ( (unsigned)ptr << 3 );
uint32_t tl = ctx->T0 + bit_len;
uint32_t th = ctx->T1;
if ( ptr == 0 )
{
sc->T0 = SPH_C32(0xFFFFFE00UL);
sc->T1 = SPH_C32(0xFFFFFFFFUL);
ctx->T0 = 0xFFFFFE00UL;
ctx->T1 = 0xFFFFFFFFUL;
}
else if ( sc->T0 == 0 )
else if ( ctx->T0 == 0 )
{
sc->T0 = SPH_C32(0xFFFFFE00UL) + bit_len;
sc->T1 = SPH_T32(sc->T1 - 1);
ctx->T0 = 0xFFFFFE00UL + bit_len;
ctx->T1 = ctx->T1 - 1;
}
else
sc->T0 -= 512 - bit_len;
ctx->T0 -= 512 - bit_len;
if ( ptr <= 52 )
buf[vptr] = _mm_set1_epi32( 0x80 );
if ( vptr < 12 )
{
memset_zero_128( buf + (ptr>>2) + 1, (52 - ptr) >> 2 );
if (out_size_w32 == 8)
buf[52>>2] = _mm_or_si128( buf[52>>2],
_mm_set1_epi32( 0x01000000UL ) );
*(buf+(56>>2)) = mm128_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( sc, buf + (ptr>>2), 64 - ptr );
memset_zero_128( buf + vptr + 1, 13 - vptr );
buf[ 13 ] = _mm_or_si128( buf[ 13 ], _mm_set1_epi32( 0x01000000UL ) );
buf[ 14 ] = mm128_bswap_32( _mm_set1_epi32( th ) );
buf[ 15 ] = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( ctx, buf + vptr, 64 - ptr );
}
else
{
memset_zero_128( buf + (ptr>>2) + 1, (60-ptr) >> 2 );
blake32_4way( sc, buf + (ptr>>2), 64 - ptr );
sc->T0 = SPH_C32(0xFFFFFE00UL);
sc->T1 = SPH_C32(0xFFFFFFFFUL);
memset_zero_128( buf, 56>>2 );
if (out_size_w32 == 8)
buf[52>>2] = _mm_set1_epi32( 0x01000000UL );
*(buf+(56>>2)) = mm128_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( sc, buf, 64 );
memset_zero_128( buf + vptr + 1, (60-ptr) >> 2 );
blake32_4way( ctx, buf + vptr, 64 - ptr );
ctx->T0 = 0xFFFFFE00UL;
ctx->T1 = 0xFFFFFFFFUL;
memset_zero_128( buf, 56>>2 );
buf[ 13 ] = _mm_or_si128( buf[ 13 ], _mm_set1_epi32( 0x01000000UL ) );
buf[ 14 ] = mm128_bswap_32( _mm_set1_epi32( th ) );
buf[ 15 ] = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( ctx, buf, 64 );
}
out = (__m128i*)dst;
for ( k = 0; k < out_size_w32; k++ )
out[k] = mm128_bswap_32( sc->H[k] );
casti_m128i( dst, 0 ) = mm128_bswap_32( casti_m128i( ctx->H, 0 ) );
casti_m128i( dst, 1 ) = mm128_bswap_32( casti_m128i( ctx->H, 1 ) );
casti_m128i( dst, 2 ) = mm128_bswap_32( casti_m128i( ctx->H, 2 ) );
casti_m128i( dst, 3 ) = mm128_bswap_32( casti_m128i( ctx->H, 3 ) );
casti_m128i( dst, 4 ) = mm128_bswap_32( casti_m128i( ctx->H, 4 ) );
casti_m128i( dst, 5 ) = mm128_bswap_32( casti_m128i( ctx->H, 5 ) );
casti_m128i( dst, 6 ) = mm128_bswap_32( casti_m128i( ctx->H, 6 ) );
casti_m128i( dst, 7 ) = mm128_bswap_32( casti_m128i( ctx->H, 7 ) );
}
#if defined (__AVX2__)
@@ -1217,163 +928,32 @@ blake32_8way_close( blake_8way_small_context *sc, unsigned ub, unsigned n,
out[k] = mm256_bswap_32( sc->H[k] );
}
// Blake-512 4 way
static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };
static void
blake64_4way_init( blake_4way_big_context *sc, const sph_u64 *iv,
const sph_u64 *salt )
{
int i;
for ( i = 0; i < 8; i++ )
sc->H[i] = _mm256_set1_epi64x( iv[i] );
for ( i = 0; i < 4; i++ )
sc->S[i] = _mm256_set1_epi64x( salt[i] );
sc->T0 = sc->T1 = 0;
sc->ptr = 0;
}
static void
blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
{
__m256i *vdata = (__m256i*)data;
__m256i *buf;
size_t ptr;
DECL_STATE64_4WAY
const int buf_size = 128; // sizeof/8
buf = sc->buf;
ptr = sc->ptr;
if ( len < (buf_size - ptr) )
{
memcpy_256( buf + (ptr>>3), vdata, len>>3 );
ptr += len;
sc->ptr = ptr;
return;
}
READ_STATE64_4WAY(sc);
while ( len > 0 )
{
size_t clen;
clen = buf_size - ptr;
if ( clen > len )
clen = len;
memcpy_256( buf + (ptr>>3), vdata, clen>>3 );
ptr += clen;
vdata = vdata + (clen>>3);
len -= clen;
if (ptr == buf_size )
{
if ((T0 = SPH_T64(T0 + 1024)) < 1024)
T1 = SPH_T64(T1 + 1);
COMPRESS64_4WAY;
ptr = 0;
}
}
WRITE_STATE64_4WAY(sc);
sc->ptr = ptr;
}
static void
blake64_4way_close( blake_4way_big_context *sc,
unsigned ub, unsigned n, void *dst, size_t out_size_w64)
{
// union {
__m256i buf[16];
// sph_u64 dummy;
// } u;
size_t ptr, k;
unsigned bit_len;
uint64_t z, zz;
sph_u64 th, tl;
__m256i *out;
ptr = sc->ptr;
bit_len = ((unsigned)ptr << 3);
z = 0x80 >> n;
zz = ((ub & -z) | z) & 0xFF;
buf[ptr>>3] = _mm256_set_epi64x( zz, zz, zz, zz );
tl = sc->T0 + bit_len;
th = sc->T1;
if (ptr == 0 )
{
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
}
else if ( sc->T0 == 0 )
{
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL) + bit_len;
sc->T1 = SPH_T64(sc->T1 - 1);
}
else
{
sc->T0 -= 1024 - bit_len;
}
if ( ptr <= 104 )
{
memset_zero_256( buf + (ptr>>3) + 1, (104-ptr) >> 3 );
if ( out_size_w64 == 8 )
buf[(104>>3)] = _mm256_or_si256( buf[(104>>3)],
_mm256_set1_epi64x( 0x0100000000000000ULL ) );
*(buf+(112>>3)) = mm256_bswap_64(
_mm256_set_epi64x( th, th, th, th ) );
*(buf+(120>>3)) = mm256_bswap_64(
_mm256_set_epi64x( tl, tl, tl, tl ) );
blake64_4way( sc, buf + (ptr>>3), 128 - ptr );
}
else
{
memset_zero_256( buf + (ptr>>3) + 1, (120 - ptr) >> 3 );
blake64_4way( sc, buf + (ptr>>3), 128 - ptr );
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
memset_zero_256( buf, 112>>3 );
if ( out_size_w64 == 8 )
buf[104>>3] = _mm256_set1_epi64x( 0x0100000000000000ULL );
*(buf+(112>>3)) = mm256_bswap_64(
_mm256_set_epi64x( th, th, th, th ) );
*(buf+(120>>3)) = mm256_bswap_64(
_mm256_set_epi64x( tl, tl, tl, tl ) );
blake64_4way( sc, buf, 128 );
}
out = (__m256i*)dst;
for ( k = 0; k < out_size_w64; k++ )
out[k] = mm256_bswap_64( sc->H[k] );
}
#endif
// Blake-256 4 way
// default 14 rounds, backward copatibility
void
blake256_4way_init(void *cc)
blake256_4way_init(void *ctx)
{
blake32_4way_init( cc, IV256, salt_zero_4way_small, 14 );
blake32_4way_init( ctx, IV256, salt_zero_4way_small, 14 );
}
void
blake256_4way(void *cc, const void *data, size_t len)
blake256_4way(void *ctx, const void *data, size_t len)
{
blake32_4way(cc, data, len);
blake32_4way(ctx, data, len);
}
void
blake256_4way_close(void *cc, void *dst)
blake256_4way_close(void *ctx, void *dst)
{
blake32_4way_close(cc, 0, 0, dst, 8);
blake32_4way_close(ctx, 0, 0, dst, 8);
}
#if defined(__AVX2__)
// Blake-256 8way
// Blake-256 8 way
void
blake256_8way_init(void *cc)
@@ -1473,38 +1053,8 @@ blake256r8_8way_close(void *cc, void *dst)
#endif
// Blake-512 4 way
#if defined (__AVX2__)
void
blake512_4way_init(void *cc)
{
blake64_4way_init(cc, IV512, salt_zero_big);
}
void
blake512_4way(void *cc, const void *data, size_t len)
{
blake64_4way(cc, data, len);
}
void
blake512_4way_close(void *cc, void *dst)
{
blake512_4way_addbits_and_close(cc, 0, 0, dst);
}
void
blake512_4way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
blake64_4way_close(cc, ub, n, dst, 8);
}
#endif
#ifdef __cplusplus
}
#endif
#endif
//#endif

322
algo/blake/blake256-hash-4way.c.new Normal file
View File

@@ -0,0 +1,322 @@
// convert blake256 32 bit to use 64 bit with serial vectoring
//
// cut calls to GS in half
//
// combine V
// v0 = {V0,V1}
// v1 = {V2,V3}
// v2 = {V4,V5}
// v3 = {V6,V7}
// v4 = {V8,V9}
// v5 = {VA,VB}
// v6 = {VC,VD}
// v7 = {CE,VF}
//
// v6x = {VD,VC} swap(VC,VD) swap(v6)
// v7x = {VF,VE} swap(VE,VF) swap(v7)
//
// V0 = v1v0
// V1 = v3v2
// V2 = v5v4
// V3 = v7v6
// V4 = v9v8
// V5 = vbva
// V6 = vdvc
// V7 = vfve
//
// The rotate in ROUND is to effect straddle and unstraddle for the third
// and 4th iteration of GS.
// It concatenates 2 contiguous 256 bit vectors and extracts the middle
// 256 bits. After the transform they must be restored with only the
// chosen bits modified in the original 2 vectors.
// ror1x128 achieves this by putting the chosen bits in arg1, the "low"
// 256 bit vector and saves the untouched bits temporailly in arg0, the
// "high" 256 bit vector. Simply reverse the process to restore data back
// to original positions.
// Use standard 4way when AVX2 is not available use x2 mode with AVX2.
//
// Data is organised the same as 32 bit 4 way, in effect serial vectoring
// on top of parallel vectoring. Same data in the same place just taking
// two chunks at a time.
//
// Transparent to user, x2 mode used when AVX2 detected.
// Use existing 4way context but revert to scalar types.
// Same interleave function (128 bit) or x2 with 256 bit?
// User trsnaparency would have to apply to interleave as well.
//
// Use common 4way update and close
/*
typedef struct {
unsigned char buf[64<<2];
uint32_t H[8<<2];
uint32_t S[4<<2];
size_t ptr;
uint32_t T0, T1;
int rounds; // 14 for blake, 8 for blakecoin & vanilla
} blakex2_4way_small_context __attribute__ ((aligned (64)));
*/
static void
blake32x2_4way_init( blake_4way_small_context *ctx, const uint32_t *iv,
const uint32_t *salt, int rounds )
{
casti_m128i( ctx->H, 0 ) = _mm_set1_epi32( iv[0] );
casti_m128i( ctx->H, 1 ) = _mm_set1_epi32( iv[1] );
casti_m128i( ctx->H, 2 ) = _mm_set1_epi32( iv[2] );
casti_m128i( ctx->H, 3 ) = _mm_set1_epi32( iv[3] );
casti_m128i( ctx->H, 4 ) = _mm_set1_epi32( iv[4] );
casti_m128i( ctx->H, 5 ) = _mm_set1_epi32( iv[5] );
casti_m128i( ctx->H, 6 ) = _mm_set1_epi32( iv[6] );
casti_m128i( ctx->H, 7 ) = _mm_set1_epi32( iv[7] );
casti_m128i( ctx->S, 0 ) = m128_zero;
casti_m128i( ctx->S, 1 ) = m128_zero;
casti_m128i( ctx->S, 2 ) = m128_zero;
casti_m128i( ctx->S, 3 ) = m128_zero;
/*
sc->S[0] = _mm_set1_epi32( salt[0] );
sc->S[1] = _mm_set1_epi32( salt[1] );
sc->S[2] = _mm_set1_epi32( salt[2] );
sc->S[3] = _mm_set1_epi32( salt[3] );
*/
ctx->T0 = ctx->T1 = 0;
ctx->ptr = 0;
ctx->rounds = rounds;
}
static void
blake32x2( blake_4way_small_context *ctx, const void *data, size_t len )
{
__m128i *buf = (__m256i*)ctx->buf;
size_t bptr = ctx->ptr << 2;
size_t vptr = ctx->ptr >> 3;
size_t blen = len << 2;
// unsigned char *buf = ctx->buf;
// size_t ptr = ctx->ptr<<4; // repurposed
DECL_STATE32x2
// buf = sc->buf;
// ptr = sc->ptr;
// adjust len for use with ptr, clen, all absolute bytes.
// int blen = len<<2;
if ( blen < (sizeof ctx->buf) - bptr )
{
memcpy( buf + vptr, data, blen );
ptr += blen;
ctx->ptr = bptr >> 2;;
return;
}
READ_STATE32( ctx );
while ( blen > 0 )
{
size_t clen;
clen = ( sizeof sc->buf ) - ptr;
if ( clen > blen )
clen = blen;
memcpy( buf + vptr, data, clen );
bptr += clen;
vptr = bptr >> 5;
data = (const unsigned char *)data + clen;
blen -= clen;
if ( bptr == sizeof ctx->buf )
{
if ( ( T0 = T0 + 512 ) < 512 ) // not needed, will never rollover
T1 += 1;
COMPRESS32x2_4WAY( ctx->rounds );
ptr = 0;
}
}
WRITE_STATE32x2( ctx );
ctx->ptr = bptr >> 2;
}
static void
blake32x2_4way_close( blake_4way_small_context *ctx, void *dst )
{
__m256i buf[8] __attribute__ ((aligned (64)));
size_t ptr = ctx->ptr;
size_t vptr = ctx->ptr>>2;
unsigned bit_len = ( (unsigned)ptr << 3 ); // one lane
uint32_t th = ctx->T1;
uint32_t tl = ctx->T0 + bit_len;
if ( ptr == 0 )
{
ctx->T0 = 0xFFFFFE00UL;
ctx->T1 = 0xFFFFFFFFUL;
}
else if ( ctx->T0 == 0 )
{
ctx->T0 = 0xFFFFFE00UL + bit_len;
ctx->T1 -= 1;
}
else
ctx->T0 -= 512 - bit_len;
// memset doesn't do ints
buf[ vptr ] = _mm256_set_epi32( 0,0,0,0, 0x80, 0x80, 0x80, 0x80 );
if ( vptr < 5 )
{
memset_zero_256( buf + vptr + 1, 6 - vptr );
buf[ 6 ] = _mm256_or_si256( vbuf[ 6 ], _mm256_set_epi32(
0x01000000UL,0x01000000UL,0x01000000UL,0x01000000UL, 0,0,0,0 ) );
buf[ 7 ] = mm256_bswap_32( _mm256_set_epi32( tl,tl,tl,tl,
th,th,th,th ) );
blake32x2_4way( ctx, buf + vptr, 64 - ptr );
}
else
{
memset_zero_256( vbuf + vptr + 1, 7 - vptr );
blake32x2_4way( ctx, vbuf + ptr, 64 - ptr );
ctx->T0 = 0xFFFFFE00UL;
ctx->T1 = 0xFFFFFFFFUL;
buf[ 6 ] = mm256_zero;
buf[ 6 ] = _mm256_set_epi32( 0,0,0,0,
0x01000000UL,0x01000000UL,0x01000000UL,0x01000000UL );
buf[ 7 ] = mm256_bswap_32( _mm256_set_epi32( tl, tl, tl, tl,
th, th, th, th );
blake32x2_4way( ctx, buf, 64 );
}
casti_m256i( dst, 0 ) = mm256_bswap_32( casti_m256i( ctx->H, 0 ) );
casti_m256i( dst, 1 ) = mm256_bswap_32( casti_m256i( ctx->H, 1 ) );
casti_m256i( dst, 2 ) = mm256_bswap_32( casti_m256i( ctx->H, 2 ) );
casti_m256i( dst, 3 ) = mm256_bswap_32( casti_m256i( ctx->H, 3 ) );
}
#define DECL_STATE32x2_4WAY \
__m256i H0, H1, H2, H3; \
__m256i S0, S1; \
uint32_t T0, T1;
#define READ_STATE32x2_4WAY(state) do \
{ \
H0 = casti_m256i( state->H, 0 ); \
H1 = casti_m256i( state->H, 1 ); \
H2 = casti_m256i( state->H, 2 ); \
H3 = casti_m256i( state->H, 3 ); \
S0 = casti_m256i( state->S, 0 ); \
S1 = casti_m256i( state->S, 1 ); \
T0 = state->T0; \
T1 = state->T1; \
#define WRITE_STATE32x2_4WAY(state) do { \
casti_m256i( state->H, 0 ) = H0; \
casti_m256i( state->H, 1 ) = H1; \
casti_m256i( state->H, 2 ) = H2; \
casti_m256i( state->H, 3 ) = H3; \
casti_m256i( state->S, 0 ) = S0; \
casti_m256i( state->S, 1 ) = S1; \
state->T0 = T0; \
state->T1 = T1; \
} while (0)
#define GSx2_4WAY( m0m2, m1m3, c0c2, c1c3, a, b, c, d ) do \
{ \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set_epi32( c1,c3, c1,c3, c1,c3, c1,c3 ), \
_mm256_set_epi32( m0,m2, m0,m2, m0,m2, m0,m2 ) ), b ), a ); \
d = mm256_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set_epi32( c0,c2, c0,c2, c0,c2, c0,c2 ), \
_mm256_set_epi32( m1,m3, m1,m3, m1,m3, m1,m3 ) ), b ), a ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
} while (0)
#define ROUND_Sx2_4WAY(r) do \
{ \
GS2_4WAY( Mx(r, 0), Mx(r, 1), Mx(r, 2), Mx(r, 3), \
CSx(r, 0), CSx(r, 1), CSx(r, 2), CSx(r, 3), V0, V2, V4, V6 ); \
GS2_4WAY( Mx(r, 4), Mx(r, 5), Mx(r, 6), Mx(r, 7), \
CSx(r, 4), CSx(r, 5), CSx(r, 6), CSx(r, 7), V1, V3, V5, V7 ); \
mm256_ror1x128_512( V3, V2 ); \
mm256_ror1x128_512( V6, V7 ); \
GS2_4WAY( Mx(r, 8), Mx(r, 9), Mx(r, A), Mx(r, B), \
CSx(r, 8), CSx(r, 9), CSx(r, A), CSx(r, B), V0, V2, V5, V7 ); \
GS2_4WAY( Mx(r, C), Mx(r, D), Mx(r, C), Mx(r, D), \
CSx(r, C), CSx(r, D), CSx(r, C), CSx(r, D), V1, V3, V4, V6 ); \
mm256_rol1x128_512( V2, V3 ); \
mm256_rol1x128_512( V7, V6 );
#define COMPRESS32x2_4WAY( rounds ) do \
{ \
__m256i M0, M1, M2, M3, M4, M5, M6, M7; \
__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
unsigned r; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = _mm256_xor_si256( S0, _mm256_set_epi32( CS1, CS1, CS1, CS1, \
CS0, CS0, CS0, CS0 ) ); \
V5 = _mm256_xor_si256( S1, _mm256_set_epi32( CS3, CS3, CS3, CS3, \
CS2, CS2, CS2, CS2 ) ); \
V6 = _mm256_xor_si256( _mm256_set1_epi32( T0 ), \
_mm256_set_epi32( CS5, CS5, CS5, CS5, \
CS4, CS4, CS4, CS4 ) ); \
V7 = _mm256_xor_si256( _mm256_set1_epi32( T1 ), \
_mm256_set_epi32( CS7, CS7, CS7, CS7, \
CS6, CS6, CS6, CS6 ) ); \
M0 = mm256_bswap_32( buf[ 0] ); \
M1 = mm256_bswap_32( buf[ 1] ); \
M2 = mm256_bswap_32( buf[ 2] ); \
M3 = mm256_bswap_32( buf[ 3] ); \
M4 = mm256_bswap_32( buf[ 4] ); \
M5 = mm256_bswap_32( buf[ 5] ); \
M6 = mm256_bswap_32( buf[ 6] ); \
M7 = mm256_bswap_32( buf[ 7] ); \
ROUND_Sx2_4WAY(0); \
ROUND_Sx2_4WAY(1); \
ROUND_Sx2_4WAY(2); \
ROUND_Sx2_4WAY(3); \
ROUND_Sx2_4WAY(4); \
ROUND_Sx2_4WAY(5); \
ROUND_Sx2_4WAY(6); \
ROUND_Sx2_4WAY(7); \
if (rounds == 14) \
{ \
ROUND_Sx2_4WAY(8); \
ROUND_Sx2_4WAY(9); \
ROUND_Sx2_4WAY(0); \
ROUND_Sx2_4WAY(1); \
ROUND_Sx2_4WAY(2); \
ROUND_Sx2_4WAY(3); \
} \
H0 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( V8, V0 ), S0 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( V9, V1 ), S1 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( VA, V2 ), S2 ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( VB, V3 ), S3 ), H3 ); \
} while (0)

701
algo/blake/blake512-hash-4way.c Normal file
View File

@@ -0,0 +1,701 @@
/* $Id: blake.c 252 2011-06-07 17:55:14Z tp $ */
/*
* BLAKE implementation.
*
* ==========================(LICENSE BEGIN)============================
*
* Copyright (c) 2007-2010 Projet RNRT SAPHIR
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ===========================(LICENSE END)=============================
*
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#if defined (__AVX2__)
#include <stddef.h>
#include <string.h>
#include <limits.h>
#include "blake-hash-4way.h"
#ifdef __cplusplus
extern "C"{
#endif
#if SPH_SMALL_FOOTPRINT && !defined SPH_SMALL_FOOTPRINT_BLAKE
#define SPH_SMALL_FOOTPRINT_BLAKE 1
#endif
#if SPH_64 && (SPH_SMALL_FOOTPRINT_BLAKE || !SPH_64_TRUE)
#define SPH_COMPACT_BLAKE_64 1
#endif
#ifdef _MSC_VER
#pragma warning (disable: 4146)
#endif
// Blake-512
static const sph_u64 IV512[8] = {
SPH_C64(0x6A09E667F3BCC908), SPH_C64(0xBB67AE8584CAA73B),
SPH_C64(0x3C6EF372FE94F82B), SPH_C64(0xA54FF53A5F1D36F1),
SPH_C64(0x510E527FADE682D1), SPH_C64(0x9B05688C2B3E6C1F),
SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
};
#if SPH_COMPACT_BLAKE_32 || SPH_COMPACT_BLAKE_64
// Blake-256 4 & 8 way, Blake-512 4 way
static const unsigned sigma[16][16] = {
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
{ 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
{ 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
{ 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
{ 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
{ 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
{ 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
{ 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
{ 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0 },
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
{ 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
{ 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
{ 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
{ 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 }
};
#endif
#define Z00 0
#define Z01 1
#define Z02 2
#define Z03 3
#define Z04 4
#define Z05 5
#define Z06 6
#define Z07 7
#define Z08 8
#define Z09 9
#define Z0A A
#define Z0B B
#define Z0C C
#define Z0D D
#define Z0E E
#define Z0F F
#define Z10 E
#define Z11 A
#define Z12 4
#define Z13 8
#define Z14 9
#define Z15 F
#define Z16 D
#define Z17 6
#define Z18 1
#define Z19 C
#define Z1A 0
#define Z1B 2
#define Z1C B
#define Z1D 7
#define Z1E 5
#define Z1F 3
#define Z20 B
#define Z21 8
#define Z22 C
#define Z23 0
#define Z24 5
#define Z25 2
#define Z26 F
#define Z27 D
#define Z28 A
#define Z29 E
#define Z2A 3
#define Z2B 6
#define Z2C 7
#define Z2D 1
#define Z2E 9
#define Z2F 4
#define Z30 7
#define Z31 9
#define Z32 3
#define Z33 1
#define Z34 D
#define Z35 C
#define Z36 B
#define Z37 E
#define Z38 2
#define Z39 6
#define Z3A 5
#define Z3B A
#define Z3C 4
#define Z3D 0
#define Z3E F
#define Z3F 8
#define Z40 9
#define Z41 0
#define Z42 5
#define Z43 7
#define Z44 2
#define Z45 4
#define Z46 A
#define Z47 F
#define Z48 E
#define Z49 1
#define Z4A B
#define Z4B C
#define Z4C 6
#define Z4D 8
#define Z4E 3
#define Z4F D
#define Z50 2
#define Z51 C
#define Z52 6
#define Z53 A
#define Z54 0
#define Z55 B
#define Z56 8
#define Z57 3
#define Z58 4
#define Z59 D
#define Z5A 7
#define Z5B 5
#define Z5C F
#define Z5D E
#define Z5E 1
#define Z5F 9
#define Z60 C
#define Z61 5
#define Z62 1
#define Z63 F
#define Z64 E
#define Z65 D
#define Z66 4
#define Z67 A
#define Z68 0
#define Z69 7
#define Z6A 6
#define Z6B 3
#define Z6C 9
#define Z6D 2
#define Z6E 8
#define Z6F B
#define Z70 D
#define Z71 B
#define Z72 7
#define Z73 E
#define Z74 C
#define Z75 1
#define Z76 3
#define Z77 9
#define Z78 5
#define Z79 0
#define Z7A F
#define Z7B 4
#define Z7C 8
#define Z7D 6
#define Z7E 2
#define Z7F A
#define Z80 6
#define Z81 F
#define Z82 E
#define Z83 9
#define Z84 B
#define Z85 3
#define Z86 0
#define Z87 8
#define Z88 C
#define Z89 2
#define Z8A D
#define Z8B 7
#define Z8C 1
#define Z8D 4
#define Z8E A
#define Z8F 5
#define Z90 A
#define Z91 2
#define Z92 8
#define Z93 4
#define Z94 7
#define Z95 6
#define Z96 1
#define Z97 5
#define Z98 F
#define Z99 B
#define Z9A 9
#define Z9B E
#define Z9C 3
#define Z9D C
#define Z9E D
#define Z9F 0
#define Mx(r, i) Mx_(Z ## r ## i)
#define Mx_(n) Mx__(n)
#define Mx__(n) M ## n
// Blake-512 4 way
#define CBx(r, i) CBx_(Z ## r ## i)
#define CBx_(n) CBx__(n)
#define CBx__(n) CB ## n
#define CB0 SPH_C64(0x243F6A8885A308D3)
#define CB1 SPH_C64(0x13198A2E03707344)
#define CB2 SPH_C64(0xA4093822299F31D0)
#define CB3 SPH_C64(0x082EFA98EC4E6C89)
#define CB4 SPH_C64(0x452821E638D01377)
#define CB5 SPH_C64(0xBE5466CF34E90C6C)
#define CB6 SPH_C64(0xC0AC29B7C97C50DD)
#define CB7 SPH_C64(0x3F84D5B5B5470917)
#define CB8 SPH_C64(0x9216D5D98979FB1B)
#define CB9 SPH_C64(0xD1310BA698DFB5AC)
#define CBA SPH_C64(0x2FFD72DBD01ADFB7)
#define CBB SPH_C64(0xB8E1AFED6A267E96)
#define CBC SPH_C64(0xBA7C9045F12C7F99)
#define CBD SPH_C64(0x24A19947B3916CF7)
#define CBE SPH_C64(0x0801F2E2858EFC16)
#define CBF SPH_C64(0x636920D871574E69)
#if SPH_COMPACT_BLAKE_64
// not used
static const sph_u64 CB[16] = {
SPH_C64(0x243F6A8885A308D3), SPH_C64(0x13198A2E03707344),
SPH_C64(0xA4093822299F31D0), SPH_C64(0x082EFA98EC4E6C89),
SPH_C64(0x452821E638D01377), SPH_C64(0xBE5466CF34E90C6C),
SPH_C64(0xC0AC29B7C97C50DD), SPH_C64(0x3F84D5B5B5470917),
SPH_C64(0x9216D5D98979FB1B), SPH_C64(0xD1310BA698DFB5AC),
SPH_C64(0x2FFD72DBD01ADFB7), SPH_C64(0xB8E1AFED6A267E96),
SPH_C64(0xBA7C9045F12C7F99), SPH_C64(0x24A19947B3916CF7),
SPH_C64(0x0801F2E2858EFC16), SPH_C64(0x636920D871574E69)
};
#endif
// Blake-512 4 way
#define GB_4WAY(m0, m1, c0, c1, a, b, c, d) do { \
a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
_mm256_set_epi64x( c1, c1, c1, c1 ), m0 ), b ), a ); \
d = mm256_ror_64( _mm256_xor_si256( d, a ), 32 ); \
c = _mm256_add_epi64( c, d ); \
b = mm256_ror_64( _mm256_xor_si256( b, c ), 25 ); \
a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
_mm256_set_epi64x( c0, c0, c0, c0 ), m1 ), b ), a ); \
d = mm256_ror_64( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi64( c, d ); \
b = mm256_ror_64( _mm256_xor_si256( b, c ), 11 ); \
} while (0)
#if SPH_COMPACT_BLAKE_64
// not used
#define ROUND_B_4WAY(r) do { \
GB_4WAY(M[sigma[r][0x0]], M[sigma[r][0x1]], \
CB[sigma[r][0x0]], CB[sigma[r][0x1]], V0, V4, V8, VC); \
GB_4WAY(M[sigma[r][0x2]], M[sigma[r][0x3]], \
CB[sigma[r][0x2]], CB[sigma[r][0x3]], V1, V5, V9, VD); \
GB_4WAY(M[sigma[r][0x4]], M[sigma[r][0x5]], \
CB[sigma[r][0x4]], CB[sigma[r][0x5]], V2, V6, VA, VE); \
GB_4WAY(M[sigma[r][0x6]], M[sigma[r][0x7]], \
CB[sigma[r][0x6]], CB[sigma[r][0x7]], V3, V7, VB, VF); \
GB_4WAY(M[sigma[r][0x8]], M[sigma[r][0x9]], \
CB[sigma[r][0x8]], CB[sigma[r][0x9]], V0, V5, VA, VF); \
GB_4WAY(M[sigma[r][0xA]], M[sigma[r][0xB]], \
CB[sigma[r][0xA]], CB[sigma[r][0xB]], V1, V6, VB, VC); \
GB_4WAY(M[sigma[r][0xC]], M[sigma[r][0xD]], \
CB[sigma[r][0xC]], CB[sigma[r][0xD]], V2, V7, V8, VD); \
GB_4WAY(M[sigma[r][0xE]], M[sigma[r][0xF]], \
CB[sigma[r][0xE]], CB[sigma[r][0xF]], V3, V4, V9, VE); \
} while (0)
#else
//current_impl
#define ROUND_B_4WAY(r) do { \
GB_4WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
GB_4WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
GB_4WAY(Mx(r, 4), Mx(r, 5), CBx(r, 4), CBx(r, 5), V2, V6, VA, VE); \
GB_4WAY(Mx(r, 6), Mx(r, 7), CBx(r, 6), CBx(r, 7), V3, V7, VB, VF); \
GB_4WAY(Mx(r, 8), Mx(r, 9), CBx(r, 8), CBx(r, 9), V0, V5, VA, VF); \
GB_4WAY(Mx(r, A), Mx(r, B), CBx(r, A), CBx(r, B), V1, V6, VB, VC); \
GB_4WAY(Mx(r, C), Mx(r, D), CBx(r, C), CBx(r, D), V2, V7, V8, VD); \
GB_4WAY(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
} while (0)
#endif
// Blake-512 4 way
#define DECL_STATE64_4WAY \
__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
__m256i S0, S1, S2, S3; \
sph_u64 T0, T1;
#define READ_STATE64_4WAY(state) do { \
H0 = (state)->H[0]; \
H1 = (state)->H[1]; \
H2 = (state)->H[2]; \
H3 = (state)->H[3]; \
H4 = (state)->H[4]; \
H5 = (state)->H[5]; \
H6 = (state)->H[6]; \
H7 = (state)->H[7]; \
S0 = (state)->S[0]; \
S1 = (state)->S[1]; \
S2 = (state)->S[2]; \
S3 = (state)->S[3]; \
T0 = (state)->T0; \
T1 = (state)->T1; \
} while (0)
#define WRITE_STATE64_4WAY(state) do { \
(state)->H[0] = H0; \
(state)->H[1] = H1; \
(state)->H[2] = H2; \
(state)->H[3] = H3; \
(state)->H[4] = H4; \
(state)->H[5] = H5; \
(state)->H[6] = H6; \
(state)->H[7] = H7; \
(state)->S[0] = S0; \
(state)->S[1] = S1; \
(state)->S[2] = S2; \
(state)->S[3] = S3; \
(state)->T0 = T0; \
(state)->T1 = T1; \
} while (0)
#if SPH_COMPACT_BLAKE_64
// not used
#define COMPRESS64_4WAY do { \
__m256i M[16]; \
__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
__m256i V8, V9, VA, VB, VC, VD, VE, VF; \
unsigned r; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = H4; \
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm256_xor_si256( S0, _mm256_set_epi64x( CB0, CB0, CB0, CB0 ) ); \
V9 = _mm256_xor_si256( S1, _mm256_set_epi64x( CB1, CB1, CB1, CB1 ) ); \
VA = _mm256_xor_si256( S2, _mm256_set_epi64x( CB2, CB2, CB2, CB2 ) ); \
VB = _mm256_xor_si256( S3, _mm256_set_epi64x( CB3, CB3, CB3, CB3 ) ); \
VC = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB4, CB4, CB4, CB4 ) ); \
VD = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB5, CB5, CB5, CB5 ) ); \
VE = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB6, CB6, CB6, CB6 ) ); \
VF = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB7, CB7, CB7, CB7 ) ); \
M[0x0] = mm256_bswap_64( *(buf+0) ); \
M[0x1] = mm256_bswap_64( *(buf+1) ); \
M[0x2] = mm256_bswap_64( *(buf+2) ); \
M[0x3] = mm256_bswap_64( *(buf+3) ); \
M[0x4] = mm256_bswap_64( *(buf+4) ); \
M[0x5] = mm256_bswap_64( *(buf+5) ); \
M[0x6] = mm256_bswap_64( *(buf+6) ); \
M[0x7] = mm256_bswap_64( *(buf+7) ); \
M[0x8] = mm256_bswap_64( *(buf+8) ); \
M[0x9] = mm256_bswap_64( *(buf+9) ); \
M[0xA] = mm256_bswap_64( *(buf+10) ); \
M[0xB] = mm256_bswap_64( *(buf+11) ); \
M[0xC] = mm256_bswap_64( *(buf+12) ); \
M[0xD] = mm256_bswap_64( *(buf+13) ); \
M[0xE] = mm256_bswap_64( *(buf+14) ); \
M[0xF] = mm256_bswap_64( *(buf+15) ); \
for (r = 0; r < 16; r ++) \
ROUND_B_4WAY(r); \
H0 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V2 ), VA ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V3 ), VB ), H3 ); \
H4 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V4 ), VC ), H4 ); \
H5 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V5 ), VD ), H5 ); \
H6 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V6 ), VE ), H6 ); \
H7 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V7 ), VF ), H7 ); \
} while (0)
#else
//current impl
#define COMPRESS64_4WAY do { \
__m256i M0, M1, M2, M3, M4, M5, M6, M7; \
__m256i M8, M9, MA, MB, MC, MD, ME, MF; \
__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
__m256i V8, V9, VA, VB, VC, VD, VE, VF; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = H4; \
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm256_xor_si256( S0, _mm256_set_epi64x( CB0, CB0, CB0, CB0 ) ); \
V9 = _mm256_xor_si256( S1, _mm256_set_epi64x( CB1, CB1, CB1, CB1 ) ); \
VA = _mm256_xor_si256( S2, _mm256_set_epi64x( CB2, CB2, CB2, CB2 ) ); \
VB = _mm256_xor_si256( S3, _mm256_set_epi64x( CB3, CB3, CB3, CB3 ) ); \
VC = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB4, CB4, CB4, CB4 ) ); \
VD = _mm256_xor_si256( _mm256_set_epi64x( T0, T0, T0, T0 ), \
_mm256_set_epi64x( CB5, CB5, CB5, CB5 ) ); \
VE = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB6, CB6, CB6, CB6 ) ); \
VF = _mm256_xor_si256( _mm256_set_epi64x( T1, T1, T1, T1 ), \
_mm256_set_epi64x( CB7, CB7, CB7, CB7 ) ); \
M0 = mm256_bswap_64( *(buf + 0) ); \
M1 = mm256_bswap_64( *(buf + 1) ); \
M2 = mm256_bswap_64( *(buf + 2) ); \
M3 = mm256_bswap_64( *(buf + 3) ); \
M4 = mm256_bswap_64( *(buf + 4) ); \
M5 = mm256_bswap_64( *(buf + 5) ); \
M6 = mm256_bswap_64( *(buf + 6) ); \
M7 = mm256_bswap_64( *(buf + 7) ); \
M8 = mm256_bswap_64( *(buf + 8) ); \
M9 = mm256_bswap_64( *(buf + 9) ); \
MA = mm256_bswap_64( *(buf + 10) ); \
MB = mm256_bswap_64( *(buf + 11) ); \
MC = mm256_bswap_64( *(buf + 12) ); \
MD = mm256_bswap_64( *(buf + 13) ); \
ME = mm256_bswap_64( *(buf + 14) ); \
MF = mm256_bswap_64( *(buf + 15) ); \
ROUND_B_4WAY(0); \
ROUND_B_4WAY(1); \
ROUND_B_4WAY(2); \
ROUND_B_4WAY(3); \
ROUND_B_4WAY(4); \
ROUND_B_4WAY(5); \
ROUND_B_4WAY(6); \
ROUND_B_4WAY(7); \
ROUND_B_4WAY(8); \
ROUND_B_4WAY(9); \
ROUND_B_4WAY(0); \
ROUND_B_4WAY(1); \
ROUND_B_4WAY(2); \
ROUND_B_4WAY(3); \
ROUND_B_4WAY(4); \
ROUND_B_4WAY(5); \
H0 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V2 ), VA ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V3 ), VB ), H3 ); \
H4 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V4 ), VC ), H4 ); \
H5 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V5 ), VD ), H5 ); \
H6 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V6 ), VE ), H6 ); \
H7 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V7 ), VF ), H7 ); \
} while (0)
#endif
static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };
static void
blake64_4way_init( blake_4way_big_context *sc, const sph_u64 *iv,
const sph_u64 *salt )
{
int i;
for ( i = 0; i < 8; i++ )
sc->H[i] = _mm256_set1_epi64x( iv[i] );
for ( i = 0; i < 4; i++ )
sc->S[i] = _mm256_set1_epi64x( salt[i] );
sc->T0 = sc->T1 = 0;
sc->ptr = 0;
}
static void
blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
{
__m256i *vdata = (__m256i*)data;
__m256i *buf;
size_t ptr;
DECL_STATE64_4WAY
const int buf_size = 128; // sizeof/8
buf = sc->buf;
ptr = sc->ptr;
if ( len < (buf_size - ptr) )
{
memcpy_256( buf + (ptr>>3), vdata, len>>3 );
ptr += len;
sc->ptr = ptr;
return;
}
READ_STATE64_4WAY(sc);
while ( len > 0 )
{
size_t clen;
clen = buf_size - ptr;
if ( clen > len )
clen = len;
memcpy_256( buf + (ptr>>3), vdata, clen>>3 );
ptr += clen;
vdata = vdata + (clen>>3);
len -= clen;
if (ptr == buf_size )
{
if ((T0 = SPH_T64(T0 + 1024)) < 1024)
T1 = SPH_T64(T1 + 1);
COMPRESS64_4WAY;
ptr = 0;
}
}
WRITE_STATE64_4WAY(sc);
sc->ptr = ptr;
}
static void
blake64_4way_close( blake_4way_big_context *sc,
unsigned ub, unsigned n, void *dst, size_t out_size_w64)
{
// union {
__m256i buf[16];
// sph_u64 dummy;
// } u;
size_t ptr, k;
unsigned bit_len;
uint64_t z, zz;
sph_u64 th, tl;
__m256i *out;
ptr = sc->ptr;
bit_len = ((unsigned)ptr << 3);
z = 0x80 >> n;
zz = ((ub & -z) | z) & 0xFF;
buf[ptr>>3] = _mm256_set_epi64x( zz, zz, zz, zz );
tl = sc->T0 + bit_len;
th = sc->T1;
if (ptr == 0 )
{
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
}
else if ( sc->T0 == 0 )
{
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL) + bit_len;
sc->T1 = SPH_T64(sc->T1 - 1);
}
else
{
sc->T0 -= 1024 - bit_len;
}
if ( ptr <= 104 )
{
memset_zero_256( buf + (ptr>>3) + 1, (104-ptr) >> 3 );
if ( out_size_w64 == 8 )
buf[(104>>3)] = _mm256_or_si256( buf[(104>>3)],
_mm256_set1_epi64x( 0x0100000000000000ULL ) );
*(buf+(112>>3)) = mm256_bswap_64(
_mm256_set_epi64x( th, th, th, th ) );
*(buf+(120>>3)) = mm256_bswap_64(
_mm256_set_epi64x( tl, tl, tl, tl ) );
blake64_4way( sc, buf + (ptr>>3), 128 - ptr );
}
else
{
memset_zero_256( buf + (ptr>>3) + 1, (120 - ptr) >> 3 );
blake64_4way( sc, buf + (ptr>>3), 128 - ptr );
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
memset_zero_256( buf, 112>>3 );
if ( out_size_w64 == 8 )
buf[104>>3] = _mm256_set1_epi64x( 0x0100000000000000ULL );
*(buf+(112>>3)) = mm256_bswap_64(
_mm256_set_epi64x( th, th, th, th ) );
*(buf+(120>>3)) = mm256_bswap_64(
_mm256_set_epi64x( tl, tl, tl, tl ) );
blake64_4way( sc, buf, 128 );
}
out = (__m256i*)dst;
for ( k = 0; k < out_size_w64; k++ )
out[k] = mm256_bswap_64( sc->H[k] );
}
void
blake512_4way_init(void *cc)
{
blake64_4way_init(cc, IV512, salt_zero_big);
}
void
blake512_4way(void *cc, const void *data, size_t len)
{
blake64_4way(cc, data, len);
}
void
blake512_4way_close(void *cc, void *dst)
{
blake512_4way_addbits_and_close(cc, 0, 0, dst);
}
void
blake512_4way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
blake64_4way_close(cc, ub, n, dst, 8);
}
#ifdef __cplusplus
}
#endif
#endif

74
algo/cubehash/cube-hash-2way.c
View File

@@ -7,6 +7,24 @@
// 2x128
// The result of hashing 10 rounds of initial data which consists of params
// zero padded.
static const uint64_t IV256[] =
{
0xCCD6F29FEA2BD4B4, 0x35481EAE63117E71, 0xE5D94E6322512D5B, 0xF4CC12BE7E624131,
0x42AF2070C2D0B696, 0x3361DA8CD0720C35, 0x8EF8AD8328CCECA4, 0x40E5FBAB4680AC00,
0x6107FBD5D89041C3, 0xF0B266796C859D41, 0x5FA2560309392549, 0x93CB628565C892FD,
0x9E4B4E602AF2B5AE, 0x85254725774ABFDD, 0x4AB6AAD615815AEB, 0xD6032C0A9CDAF8AF
};
static const uint64_t IV512[] =
{
0x50F494D42AEA2A61, 0x4167D83E2D538B8B, 0xC701CF8C3FEE2313, 0x50AC5695CC39968E,
0xA647A8B34D42C787, 0x825B453797CF0BEF, 0xF22090C4EEF864D2, 0xA23911AED0E5CD33,
0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
};
static void transform_2way( cube_2way_context *sp )
{
int r;
@@ -45,10 +63,10 @@ static void transform_2way( cube_2way_context *sp )
x1 = _mm256_xor_si256( x1, x5 );
x2 = _mm256_xor_si256( x2, x6 );
x3 = _mm256_xor_si256( x3, x7 );
x4 = mm256_swap128_64( x4 );
x5 = mm256_swap128_64( x5 );
x6 = mm256_swap128_64( x6 );
x7 = mm256_swap128_64( x7 );
x4 = mm256_swap64_128( x4 );
x5 = mm256_swap64_128( x5 );
x6 = mm256_swap64_128( x6 );
x7 = mm256_swap64_128( x7 );
x4 = _mm256_add_epi32( x0, x4 );
x5 = _mm256_add_epi32( x1, x5 );
x6 = _mm256_add_epi32( x2, x6 );
@@ -69,10 +87,10 @@ static void transform_2way( cube_2way_context *sp )
x1 = _mm256_xor_si256( x1, x5 );
x2 = _mm256_xor_si256( x2, x6 );
x3 = _mm256_xor_si256( x3, x7 );
x4 = mm256_swap64_32( x4 );
x5 = mm256_swap64_32( x5 );
x6 = mm256_swap64_32( x6 );
x7 = mm256_swap64_32( x7 );
x4 = mm256_swap32_64( x4 );
x5 = mm256_swap32_64( x5 );
x6 = mm256_swap32_64( x6 );
x7 = mm256_swap32_64( x7 );
}
_mm256_store_si256( (__m256i*)sp->h, x0 );
@@ -86,36 +104,26 @@ static void transform_2way( cube_2way_context *sp )
}
cube_2way_context cube_2way_ctx_cache __attribute__ ((aligned (64)));
int cube_2way_reinit( cube_2way_context *sp )
{
memcpy( sp, &cube_2way_ctx_cache, sizeof(cube_2way_context) );
return 0;
}
int cube_2way_init( cube_2way_context *sp, int hashbitlen, int rounds,
int blockbytes )
int blockbytes )
{
int i;
const uint64_t* iv = hashbitlen == 512 ? IV512 : IV256;
sp->hashlen = hashbitlen/128;
sp->blocksize = blockbytes/16;
sp->rounds = rounds;
sp->pos = 0;
// all sizes of __m128i
cube_2way_ctx_cache.hashlen = hashbitlen/128;
cube_2way_ctx_cache.blocksize = blockbytes/16;
cube_2way_ctx_cache.rounds = rounds;
cube_2way_ctx_cache.pos = 0;
__m256i* h = (__m256i*)sp->h;
for ( i = 0; i < 8; ++i )
cube_2way_ctx_cache.h[i] = m256_zero;
h[0] = _mm256_set_epi64x( iv[ 1], iv[ 0], iv[ 1], iv[ 0] );
h[1] = _mm256_set_epi64x( iv[ 3], iv[ 2], iv[ 3], iv[ 2] );
h[2] = _mm256_set_epi64x( iv[ 5], iv[ 4], iv[ 5], iv[ 4] );
h[3] = _mm256_set_epi64x( iv[ 7], iv[ 6], iv[ 7], iv[ 6] );
h[4] = _mm256_set_epi64x( iv[ 9], iv[ 8], iv[ 9], iv[ 8] );
h[5] = _mm256_set_epi64x( iv[11], iv[10], iv[11], iv[10] );
h[6] = _mm256_set_epi64x( iv[13], iv[12], iv[13], iv[12] );
h[7] = _mm256_set_epi64x( iv[15], iv[14], iv[15], iv[14] );
cube_2way_ctx_cache.h[0] = _mm256_set_epi32(
0, rounds, blockbytes, hashbitlen / 8,
0, rounds, blockbytes, hashbitlen / 8 );
for ( i = 0; i < 10; ++i )
transform_2way( &cube_2way_ctx_cache );
memcpy( sp, &cube_2way_ctx_cache, sizeof(cube_2way_context) );
return 0;
}

76
algo/cubehash/cubehash_sse2.c
View File

@@ -14,6 +14,25 @@
#include <unistd.h>
#include <memory.h>
#include "avxdefs.h"
#include <stdio.h>
// The result of hashing 10 rounds of initial data which is params and
// mostly zeros.
static const uint64_t IV256[] =
{
0xCCD6F29FEA2BD4B4, 0x35481EAE63117E71, 0xE5D94E6322512D5B, 0xF4CC12BE7E624131,
0x42AF2070C2D0B696, 0x3361DA8CD0720C35, 0x8EF8AD8328CCECA4, 0x40E5FBAB4680AC00,
0x6107FBD5D89041C3, 0xF0B266796C859D41, 0x5FA2560309392549, 0x93CB628565C892FD,
0x9E4B4E602AF2B5AE, 0x85254725774ABFDD, 0x4AB6AAD615815AEB, 0xD6032C0A9CDAF8AF
};
static const uint64_t IV512[] =
{
0x50F494D42AEA2A61, 0x4167D83E2D538B8B, 0xC701CF8C3FEE2313, 0x50AC5695CC39968E,
0xA647A8B34D42C787, 0x825B453797CF0BEF, 0xF22090C4EEF864D2, 0xA23911AED0E5CD33,
0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
};
static void transform( cubehashParam *sp )
{
@@ -128,48 +147,37 @@ static void transform( cubehashParam *sp )
#endif
} // transform
// Cubehash context initializing is very expensive.
// Cache the intial value for faster reinitializing.
cubehashParam cube_ctx_cache __attribute__ ((aligned (64)));
int cubehashReinit( cubehashParam *sp )
{
memcpy( sp, &cube_ctx_cache, sizeof(cubehashParam) );
return SUCCESS;
}
// Initialize the cache then copy to sp.
int cubehashInit(cubehashParam *sp, int hashbitlen, int rounds, int blockbytes)
{
int i;
const uint64_t* iv = hashbitlen == 512 ? IV512 : IV256;
sp->hashlen = hashbitlen/128;
sp->blocksize = blockbytes/16;
sp->rounds = rounds;
sp->pos = 0;
#if defined(__AVX2__)
if ( hashbitlen < 8 ) return BAD_HASHBITLEN;
if ( hashbitlen > 512 ) return BAD_HASHBITLEN;
if ( hashbitlen != 8 * (hashbitlen / 8) ) return BAD_HASHBITLEN;
__m256i* x = (__m256i*)sp->x;
/* Sanity checks */
if ( rounds <= 0 || rounds > 32 )
rounds = CUBEHASH_ROUNDS;
if ( blockbytes <= 0 || blockbytes >= 256)
blockbytes = CUBEHASH_BLOCKBYTES;
x[0] = _mm256_set_epi64x( iv[ 3], iv[ 2], iv[ 1], iv[ 0] );
x[1] = _mm256_set_epi64x( iv[ 7], iv[ 6], iv[ 5], iv[ 4] );
x[2] = _mm256_set_epi64x( iv[11], iv[10], iv[ 9], iv[ 8] );
x[3] = _mm256_set_epi64x( iv[15], iv[14], iv[13], iv[12] );
// all sizes of __m128i
cube_ctx_cache.hashlen = hashbitlen/128;
cube_ctx_cache.blocksize = blockbytes/16;
cube_ctx_cache.rounds = rounds;
cube_ctx_cache.pos = 0;
#else
for ( i = 0; i < 8; ++i )
cube_ctx_cache.x[i] = _mm_setzero_si128();;
__m128i* x = (__m128i*)sp->x;
cube_ctx_cache.x[0] = _mm_set_epi32( 0, rounds, blockbytes,
hashbitlen / 8 );
x[0] = _mm_set_epi64x( iv[ 1], iv[ 0] );
x[1] = _mm_set_epi64x( iv[ 3], iv[ 2] );
x[2] = _mm_set_epi64x( iv[ 5], iv[ 4] );
x[3] = _mm_set_epi64x( iv[ 7], iv[ 6] );
x[4] = _mm_set_epi64x( iv[ 9], iv[ 8] );
x[5] = _mm_set_epi64x( iv[11], iv[10] );
x[6] = _mm_set_epi64x( iv[13], iv[12] );
x[7] = _mm_set_epi64x( iv[15], iv[14] );
for ( i = 0; i < 10; ++i )
transform( &cube_ctx_cache );
memcpy( sp, &cube_ctx_cache, sizeof(cubehashParam) );
#endif
return SUCCESS;
}

4
algo/hodl/aes.c
View File

@@ -3,7 +3,7 @@
#include "wolf-aes.h"
#include "miner.h"
#ifndef NO_AES_NI
#if defined(__AES__)
static inline void ExpandAESKey256_sub1(__m128i *tmp1, __m128i *tmp2)
{
@@ -151,7 +151,7 @@ void AES256CBC(__m128i** data, const __m128i** next, __m128i ExpandedKey[][16],
}
}
#else // NO AVX
#else // NO SSE4.2
static inline __m128i AES256Core(__m128i State, const __m128i *ExpandedKey)
{

38
algo/hodl/hodl-gate.c
View File

@@ -101,39 +101,6 @@ void hodl_build_block_header( struct work* g_work, uint32_t version,
g_work->data[31] = 0x00000280;
}
// hodl build_extra_header is redundant, hodl can use std_build_extra_header
// and call hodl_build_block_header.
#if 0
void hodl_build_extraheader( struct work* g_work, struct stratum_ctx *sctx )
{
uchar merkle_tree[64] = { 0 };
size_t t;
// int i;
algo_gate.gen_merkle_root( merkle_tree, sctx );
// Increment extranonce2
for ( t = 0; t < sctx->xnonce2_size && !( ++sctx->job.xnonce2[t] ); t++ );
algo_gate.build_block_header( g_work, le32dec( sctx->job.version ),
(uint32_t*) sctx->job.prevhash, (uint32_t*) merkle_tree,
le32dec( sctx->job.ntime ), le32dec( sctx->job.nbits ) );
/*
// Assemble block header
memset( g_work->data, 0, sizeof(g_work->data) );
g_work->data[0] = le32dec( sctx->job.version );
for ( i = 0; i < 8; i++ )
g_work->data[1 + i] = le32dec( (uint32_t *) sctx->job.prevhash + i );
for ( i = 0; i < 8; i++ )
g_work->data[9 + i] = be32dec( (uint32_t *) merkle_root + i );
g_work->data[ algo_gate.ntime_index ] = le32dec( sctx->job.ntime );
g_work->data[ algo_gate.nbits_index ] = le32dec( sctx->job.nbits );
g_work->data[22] = 0x80000000;
g_work->data[31] = 0x00000280;
*/
}
#endif
// called only by thread 0, saves a backup of g_work
void hodl_get_new_work( struct work* work, struct work* g_work)
{
@@ -179,7 +146,7 @@ bool hodl_do_this_thread( int thr_id )
int hodl_scanhash( int thr_id, struct work* work, uint32_t max_nonce,
uint64_t *hashes_done )
{
#ifndef NO_AES_NI
#if defined(__AES__)
GenRandomGarbage( (CacheEntry*)hodl_scratchbuf, work->data, thr_id );
pthread_barrier_wait( &hodl_barrier );
return scanhash_hodl_wolf( thr_id, work, max_nonce, hashes_done );
@@ -189,7 +156,7 @@ int hodl_scanhash( int thr_id, struct work* work, uint32_t max_nonce,
bool register_hodl_algo( algo_gate_t* gate )
{
#ifdef NO_AES_NI
#if defined(__AES__)
applog( LOG_ERR, "Only CPUs with AES are supported, use legacy version.");
return false;
#endif
@@ -207,7 +174,6 @@ bool register_hodl_algo( algo_gate_t* gate )
gate->build_stratum_request = (void*)&hodl_le_build_stratum_request;
gate->malloc_txs_request = (void*)&hodl_malloc_txs_request;
gate->build_block_header = (void*)&hodl_build_block_header;
// gate->build_extraheader = (void*)&hodl_build_extraheader;
gate->resync_threads = (void*)&hodl_resync_threads;
gate->do_this_thread = (void*)&hodl_do_this_thread;
gate->work_cmp_size = 76;

10
algo/hodl/hodl-wolf.c
View File

@@ -8,7 +8,7 @@
#include "hodl-wolf.h"
#include "miner.h"
#ifndef NO_AES_NI
#if defined(__AES__)
void GenerateGarbageCore( CacheEntry *Garbage, int ThreadID, int ThreadCount,
void *MidHash )
@@ -139,7 +139,7 @@ int scanhash_hodl_wolf( int threadNumber, struct work* work, uint32_t max_nonce,
return(0);
#else // no AVX
#else // no SSE4.2
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
@@ -160,7 +160,6 @@ int scanhash_hodl_wolf( int threadNumber, struct work* work, uint32_t max_nonce,
{
// copy data to first l2 cache
memcpy(Cache.dwords, Garbage + k, GARBAGE_SLICE_SIZE);
#ifndef NO_AES_NI
for(int j = 0; j < AES_ITERATIONS; j++)
{
CacheEntry TmpXOR;
@@ -184,7 +183,6 @@ int scanhash_hodl_wolf( int threadNumber, struct work* work, uint32_t max_nonce,
AES256CBC( Cache.dqwords, TmpXOR.dqwords, ExpKey,
TmpXOR.dqwords[ (GARBAGE_SLICE_SIZE / sizeof(__m128i))
- 1 ], 256 ); }
#endif
// use last X bits as solution
if( ( Cache.dwords[ (GARBAGE_SLICE_SIZE >> 2) - 1 ]
& (COMPARE_SIZE - 1) ) < 1000 )
@@ -206,7 +204,7 @@ int scanhash_hodl_wolf( int threadNumber, struct work* work, uint32_t max_nonce,
*hashes_done = CollisionCount;
return(0);
#endif
#endif // SSE4.2 else
}
@@ -218,5 +216,5 @@ void GenRandomGarbage(CacheEntry *Garbage, uint32_t *pdata, int thr_id)
GenerateGarbageCore(Garbage, thr_id, opt_n_threads, MidHash);
}
#endif
#endif // AES

8
algo/hodl/sha512-avx.h
View File

@@ -22,16 +22,20 @@ typedef struct
#ifdef __AVX2__
__m256i h[8];
__m256i w[80];
#else // AVX
#elif defined(__SSE4_2__)
__m128i h[8];
__m128i w[80];
#else
int dummy;
#endif
} Sha512Context;
#ifdef __AVX2__
#define SHA512_PARALLEL_N 8
#else // AVX
#elif defined(__SSE$_2__)
#define SHA512_PARALLEL_N 4
#else
#define SHA512_PARALLEL_N 1 // dummy value
#endif
//SHA-512 related functions

1
algo/hodl/sha512_avx.c
View File

@@ -1,4 +1,5 @@
#ifndef __AVX2__
#ifdef __SSE4_2__
//#ifdef __AVX__

13
algo/luffa/luffa_for_sse2.c
View File

@@ -30,6 +30,19 @@
a1 = _mm_or_si128( _mm_srli_si128(a1,4), _mm_slli_si128(b,12) ); \
} while(0)
/*
static inline __m256i mult2_avx2( a )
{
__m128 a0, a0, b;
a0 = mm128_extractlo_256( a );
a1 = mm128_extracthi_256( a );
b = _mm_xor_si128( a0, _mm_shuffle_epi32( _mm_and_si128(a1,MASK), 16 ) );
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) );
return mm256_concat_128( a1, a0 );
}
*/
#define STEP_PART(x,c,t)\
SUBCRUMB(*x,*(x+1),*(x+2),*(x+3),*t);\
SUBCRUMB(*(x+5),*(x+6),*(x+7),*(x+4),*t);\

6
algo/lyra2/allium-4way.c
View File

@@ -55,11 +55,11 @@ void allium_4way_hash( void *state, const void *input )
LYRA2RE( hash3, 32, hash3, 32, hash3, 32, 1, 8, 8 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*)hash0, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash1, (const byte*)hash1, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash2, (const byte*)hash2, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash3, (const byte*)hash3, 32 );
LYRA2RE( hash0, 32, hash0, 32, hash0, 32, 1, 8, 8 );

2
algo/lyra2/lyra2-gate.c
View File

@@ -27,7 +27,7 @@ bool register_lyra2rev3_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_lyra2rev3;
gate->hash = (void*)&lyra2rev3_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->optimizations = SSE2_OPT | SSE42_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2rev3_thread_init;
gate->set_target = (void*)&alt_set_target;
return true;

4
algo/lyra2/lyra2-gate.h
View File

@@ -17,14 +17,14 @@ bool register_lyra2rev3_algo( algo_gate_t* gate );
void lyra2rev3_4way_hash( void *state, const void *input );
int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
bool init_lyra2rev3_4way_ctx();
#else
void lyra2rev3_hash( void *state, const void *input );
int scanhash_lyra2rev3( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
bool init_lyra2rev3_ctx();
#endif

21
algo/lyra2/lyra2re.c
View File

@@ -7,8 +7,7 @@
#include "lyra2.h"
#include "algo-gate-api.h"
#include "avxdefs.h"
#ifndef NO_AES_NI
#if defined(__AES__)
#include "algo/groestl/aes_ni/hash-groestl256.h"
#endif
@@ -18,10 +17,10 @@ typedef struct {
sph_blake256_context blake;
sph_keccak256_context keccak;
sph_skein256_context skein;
#ifdef NO_AES_NI
sph_groestl256_context groestl;
#else
#if defined(__AES__)
hashState_groestl256 groestl;
#else
sph_groestl256_context groestl;
#endif
} lyra2re_ctx_holder;
@@ -33,10 +32,10 @@ void init_lyra2re_ctx()
sph_blake256_init(&lyra2re_ctx.blake);
sph_keccak256_init(&lyra2re_ctx.keccak);
sph_skein256_init(&lyra2re_ctx.skein);
#ifdef NO_AES_NI
sph_groestl256_init(&lyra2re_ctx.groestl);
#else
#if defined(__AES__)
init_groestl256( &lyra2re_ctx.groestl, 32 );
#else
sph_groestl256_init(&lyra2re_ctx.groestl);
#endif
}
@@ -72,11 +71,11 @@ void lyra2re_hash(void *state, const void *input)
sph_skein256(&ctx.skein, hashA, 32);
sph_skein256_close(&ctx.skein, hashB);
#ifdef NO_AES_NI
#if defined(__AES__)
update_and_final_groestl256( &ctx.groestl, hashA, hashB, 256 );
#else
sph_groestl256( &ctx.groestl, hashB, 32 );
sph_groestl256_close( &ctx.groestl, hashA );
#else
update_and_final_groestl256( &ctx.groestl, hashA, hashB, 256 );
#endif
memcpy(state, hashA, 32);

14
algo/lyra2/lyra2rev2-4way.c
View File

@@ -48,11 +48,11 @@ void lyra2rev2_4way_hash( void *state, const void *input )
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
LYRA2REV2( l2v2_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
@@ -65,13 +65,13 @@ void lyra2rev2_4way_hash( void *state, const void *input )
skein256_4way_close( &ctx.skein, vhash64 );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );

59
algo/lyra2/lyra2rev3-4way.c
View File

@@ -43,11 +43,11 @@ void lyra2rev3_4way_hash( void *state, const void *input )
LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 256, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
@@ -57,54 +57,67 @@ void lyra2rev3_4way_hash( void *state, const void *input )
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
bmw256_4way( &ctx.bmw, vhash, 32 );
bmw256_4way_close( &ctx.bmw, vhash );
bmw256_4way_close( &ctx.bmw, state );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t lane_hash[8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
const uint32_t Htarg = ptarget[7];
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 76; // 19*4
__m128i *noncev = (__m128i*)vdata + 19; // aligned
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
if ( opt_benchmark )
( (uint32_t*)ptarget )[7] = 0x0000ff;
swab32_array( edata, pdata, 20 );
// Need big endian data
casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
do {
be32enc( noncep, n );
be32enc( noncep+1, n+1 );
be32enc( noncep+2, n+2 );
be32enc( noncep+3, n+3 );
do
{
*noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
lyra2rev3_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
work_set_target_ratio( work, lane_hash );
if ( submit_work( mythr, work ) )
applog( LOG_NOTICE, "Share %d submitted by thread %d, lane %d.",
accepted_share_count + rejected_share_count + 1,
thr_id, lane );
else
applog( LOG_WARNING, "Failed to submit share." );
}
}
n += 4;
} while ( (num_found == 0) && (n < max_nonce-4)
&& !work_restart[thr_id].restart);
} while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
*hashes_done = n - first_nonce + 1;
return num_found;
return 0;
}
#endif

71
algo/lyra2/lyra2rev3.c
View File

@@ -8,7 +8,6 @@
typedef struct {
cubehashParam cube;
// cubehashParam cube2;
sph_blake256_context blake;
sph_bmw256_context bmw;
@@ -20,7 +19,6 @@ static __thread sph_blake256_context l2v3_blake_mid;
bool init_lyra2rev3_ctx()
{
cubehashInit( &lyra2v3_ctx.cube, 256, 16, 32 );
// cubehashInit( &lyra2v3_ctx.cube2, 256, 16, 32 );
sph_blake256_init( &lyra2v3_ctx.blake );
sph_bmw256_init( &lyra2v3_ctx.bmw );
return true;
@@ -59,44 +57,51 @@ void lyra2rev3_hash( void *state, const void *input )
memcpy( state, hash, 32 );
}
int scanhash_lyra2rev3(int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done)
int scanhash_lyra2rev3( int thr_id, 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 endiandata[20] __attribute__ ((aligned (64)));
uint32_t hash[8] __attribute__((aligned(64)));
const uint32_t first_nonce = pdata[19];
uint32_t nonce = first_nonce;
const uint32_t Htarg = ptarget[7];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t endiandata[20] __attribute__ ((aligned (64)));
uint32_t hash[8] __attribute__((aligned(64)));
const uint32_t first_nonce = pdata[19];
uint32_t nonce = first_nonce;
const uint32_t Htarg = ptarget[7];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x0000ff;
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x0000ff;
swab32_array( endiandata, pdata, 20 );
// need big endian data
casti_m128i( endiandata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( endiandata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( endiandata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( endiandata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
l2v3_blake256_midstate( endiandata );
l2v3_blake256_midstate( endiandata );
do {
be32enc(&endiandata[19], nonce);
lyra2rev3_hash(hash, endiandata);
do
{
be32enc(&endiandata[19], nonce);
lyra2rev3_hash(hash, endiandata);
if (hash[7] <= Htarg )
{
if( fulltest(hash, ptarget) )
{
pdata[19] = nonce;
work_set_target_ratio( work, hash );
*hashes_done = pdata[19] - first_nonce;
return 1;
}
}
nonce++;
if (hash[7] <= Htarg )
{
if( fulltest(hash, ptarget) )
{
pdata[19] = nonce;
work_set_target_ratio( work, hash );
*hashes_done = pdata[19] - first_nonce;
return 1;
}
}
nonce++;
} while (nonce < max_nonce && !work_restart[thr_id].restart);
} while (nonce < max_nonce && !work_restart[thr_id].restart);
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}

14
algo/lyra2/sponge.h
View File

@@ -91,7 +91,7 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
#elif defined(__SSE4_2__)
#elif defined(__SSE2__)
// process 2 columns in parallel
// returns void, all args updated
@@ -108,14 +108,14 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
#define LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \
G_2X64( s0, s2, s4, s6 ); \
G_2X64( s1, s3, s5, s7 ); \
mm128_ror256_1x64( s2, s3 ); \
mm128_swap256_128( s4, s5 ); \
mm128_rol256_1x64( s6, s7 ); \
mm128_rol1x64_256( s2, s3 ); \
mm128_swap128_256( s4, s5 ); \
mm128_rol1x64_256( s6, s7 ); \
G_2X64( s0, s2, s4, s6 ); \
G_2X64( s1, s3, s5, s7 ); \
mm128_rol256_1x64( s2, s3 ); \
mm128_swap256_128( s4, s5 ); \
mm128_ror256_1x64( s6, s7 );
mm128_rol1x64_256( s2, s3 ); \
mm128_swap128_256( s4, s5 ); \
mm128_ror1x64_256( s6, s7 );
#define LYRA_12_ROUNDS_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \
LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \

49
algo/m7m.c
View File

@@ -7,7 +7,6 @@
#include <string.h>
#include <float.h>
#include <math.h>
#include "algo/sha/sph_sha2.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/haval/sph-haval.h"
#include "algo/tiger/sph_tiger.h"
@@ -117,13 +116,8 @@ uint32_t sw2_(int nnounce)
}
typedef struct {
#ifndef USE_SPH_SHA
SHA256_CTX sha256;
SHA512_CTX sha512;
#else
sph_sha256_context sha256;
sph_sha512_context sha512;
#endif
sph_keccak512_context keccak;
sph_whirlpool_context whirlpool;
sph_haval256_5_context haval;
@@ -135,13 +129,8 @@ m7m_ctx_holder m7m_ctx;
void init_m7m_ctx()
{
#ifndef USE_SPH_SHA
SHA256_Init( &m7m_ctx.sha256 );
SHA512_Init( &m7m_ctx.sha512 );
#else
sph_sha256_init( &m7m_ctx.sha256 );
sph_sha512_init( &m7m_ctx.sha512 );
#endif
sph_keccak512_init( &m7m_ctx.keccak );
sph_whirlpool_init( &m7m_ctx.whirlpool );
sph_haval256_5_init( &m7m_ctx.haval );
@@ -176,28 +165,18 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
m7m_ctx_holder ctx1, ctx2 __attribute__ ((aligned (64)));
memcpy( &ctx1, &m7m_ctx, sizeof(m7m_ctx) );
#ifndef USE_SPH_SHA
SHA256_CTX ctxf_sha256;
#else
sph_sha256_context ctxf_sha256;
#endif
memcpy(data, pdata, 80);
#ifndef USE_SPH_SHA
SHA256_Update( &ctx1.sha256, data, M7_MIDSTATE_LEN );
SHA512_Update( &ctx1.sha512, data, M7_MIDSTATE_LEN );
#else
sph_sha256( &ctx1.sha256, data, M7_MIDSTATE_LEN );
sph_sha512( &ctx1.sha512, data, M7_MIDSTATE_LEN );
#endif
sph_keccak512( &ctx1.keccak, data, M7_MIDSTATE_LEN );
sph_whirlpool( &ctx1.whirlpool, data, M7_MIDSTATE_LEN );
sph_haval256_5( &ctx1.haval, data, M7_MIDSTATE_LEN );
sph_tiger( &ctx1.tiger, data, M7_MIDSTATE_LEN );
sph_ripemd160( &ctx1.ripemd, data, M7_MIDSTATE_LEN );
// the following calculations can be performed once and the results shared
mpz_t magipi, magisw, product, bns0, bns1;
mpf_t magifpi, magifpi0, mpt1, mpt2, mptmp, mpten;
@@ -222,22 +201,11 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
memcpy( &ctx2, &ctx1, sizeof(m7m_ctx) );
// with 4 way can a single midstate be shared among lanes?
// do sinlge round of midstate and inyerleave for final
#ifndef USE_SPH_SHA
SHA256_Update( &ctx2.sha256, data_p64, 80 - M7_MIDSTATE_LEN );
SHA256_Final( (unsigned char*) (bhash[0]), &ctx2.sha256 );
SHA512_Update( &ctx2.sha512, data_p64, 80 - M7_MIDSTATE_LEN );
SHA512_Final( (unsigned char*) (bhash[1]), &ctx2.sha512 );
#else
sph_sha256( &ctx2.sha256, data_p64, 80 - M7_MIDSTATE_LEN );
sph_sha256_close( &ctx2.sha256, (void*)(bhash[0]) );
sph_sha512( &ctx2.sha512, data_p64, 80 - M7_MIDSTATE_LEN );
sph_sha512_close( &ctx2.sha512, (void*)(bhash[1]) );
#endif
sph_keccak512( &ctx2.keccak, data_p64, 80 - M7_MIDSTATE_LEN );
sph_keccak512_close( &ctx2.keccak, (void*)(bhash[2]) );
@@ -253,7 +221,6 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
sph_ripemd160( &ctx2.ripemd, data_p64, 80 - M7_MIDSTATE_LEN );
sph_ripemd160_close( &ctx2.ripemd, (void*)(bhash[6]) );
// 4 way serial
mpz_import(bns0, a, -1, p, -1, 0, bhash[0]);
mpz_set(bns1, bns0);
mpz_set(product, bns0);
@@ -269,17 +236,10 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
bytes = mpz_sizeinbase(product, 256);
mpz_export((void *)bdata, NULL, -1, 1, 0, 0, product);
#ifndef USE_SPH_SHA
SHA256_Init( &ctxf_sha256 );
SHA256_Update( &ctxf_sha256, bdata, bytes );
SHA256_Final( (unsigned char*) hash, &ctxf_sha256 );
#else
sph_sha256_init( &ctxf_sha256 );
sph_sha256( &ctxf_sha256, bdata, bytes );
sph_sha256_close( &ctxf_sha256, (void*)(hash) );
#endif
// do once and share
digits=(int)((sqrt((double)(n/2))*(1.+EPS))/9000+75);
mp_bitcnt_t prec = (long int)(digits*BITS_PER_DIGIT+16);
mpf_set_prec_raw(magifpi, prec);
@@ -302,7 +262,6 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
mpz_set_f(magipi, magifpi);
mpz_add(magipi,magipi,magisw);
mpz_add(product,product,magipi);
// share magipi, product and do serial
mpz_import(bns0, b, -1, p, -1, 0, (void*)(hash));
mpz_add(bns1, bns1, bns0);
mpz_mul(product,product,bns1);
@@ -312,18 +271,11 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
mpzscale=bytes;
mpz_export(bdata, NULL, -1, 1, 0, 0, product);
#ifndef USE_SPH_SHA
SHA256_Init( &ctxf_sha256 );
SHA256_Update( &ctxf_sha256, bdata, bytes );
SHA256_Final( (unsigned char*) hash, &ctxf_sha256 );
#else
sph_sha256_init( &ctxf_sha256 );
sph_sha256( &ctxf_sha256, bdata, bytes );
sph_sha256_close( &ctxf_sha256, (void*)(hash) );
#endif
}
// this is the scanhash part
const unsigned char *hash_ = (const unsigned char *)hash;
const unsigned char *target_ = (const unsigned char *)ptarget;
for ( i = 31; i >= 0; i-- )
@@ -354,7 +306,6 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
pdata[19] = n;
// do this in hashm7m
out:
mpf_set_prec_raw(magifpi, prec0);
mpf_set_prec_raw(magifpi0, prec0);

334
algo/panama/sph_panama.c Normal file
View File

@@ -0,0 +1,334 @@
/* $Id: panama.c 216 2010-06-08 09:46:57Z tp $ */
/*
* PANAMA implementation.
*
* ==========================(LICENSE BEGIN)============================
*
* Copyright (c) 2007-2010 Projet RNRT SAPHIR
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ===========================(LICENSE END)=============================
*
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#include <stddef.h>
#include <string.h>
#include "sph_panama.h"
#define LVAR17(b) sph_u32 \
b ## 0, b ## 1, b ## 2, b ## 3, b ## 4, b ## 5, \
b ## 6, b ## 7, b ## 8, b ## 9, b ## 10, b ## 11, \
b ## 12, b ## 13, b ## 14, b ## 15, b ## 16;
#define LVARS \
LVAR17(a) \
LVAR17(g) \
LVAR17(p) \
LVAR17(t)
#define M17(macro) do { \
macro( 0, 1, 2, 4); \
macro( 1, 2, 3, 5); \
macro( 2, 3, 4, 6); \
macro( 3, 4, 5, 7); \
macro( 4, 5, 6, 8); \
macro( 5, 6, 7, 9); \
macro( 6, 7, 8, 10); \
macro( 7, 8, 9, 11); \
macro( 8, 9, 10, 12); \
macro( 9, 10, 11, 13); \
macro(10, 11, 12, 14); \
macro(11, 12, 13, 15); \
macro(12, 13, 14, 16); \
macro(13, 14, 15, 0); \
macro(14, 15, 16, 1); \
macro(15, 16, 0, 2); \
macro(16, 0, 1, 3); \
} while (0)
#define BUPDATE1(n0, n2) do { \
sc->buffer[ptr24][n0] ^= sc->buffer[ptr31][n2]; \
sc->buffer[ptr31][n2] ^= INW1(n2); \
} while (0)
#define BUPDATE do { \
BUPDATE1(0, 2); \
BUPDATE1(1, 3); \
BUPDATE1(2, 4); \
BUPDATE1(3, 5); \
BUPDATE1(4, 6); \
BUPDATE1(5, 7); \
BUPDATE1(6, 0); \
BUPDATE1(7, 1); \
} while (0)
#define RSTATE(n0, n1, n2, n4) (a ## n0 = sc->state[n0])
#define WSTATE(n0, n1, n2, n4) (sc->state[n0] = a ## n0)
#define GAMMA(n0, n1, n2, n4) \
(g ## n0 = a ## n0 ^ (a ## n1 | SPH_T32(~a ## n2)))
#define PI_ALL do { \
p0 = g0; \
p1 = SPH_ROTL32( g7, 1); \
p2 = SPH_ROTL32(g14, 3); \
p3 = SPH_ROTL32( g4, 6); \
p4 = SPH_ROTL32(g11, 10); \
p5 = SPH_ROTL32( g1, 15); \
p6 = SPH_ROTL32( g8, 21); \
p7 = SPH_ROTL32(g15, 28); \
p8 = SPH_ROTL32( g5, 4); \
p9 = SPH_ROTL32(g12, 13); \
p10 = SPH_ROTL32( g2, 23); \
p11 = SPH_ROTL32( g9, 2); \
p12 = SPH_ROTL32(g16, 14); \
p13 = SPH_ROTL32( g6, 27); \
p14 = SPH_ROTL32(g13, 9); \
p15 = SPH_ROTL32( g3, 24); \
p16 = SPH_ROTL32(g10, 8); \
} while (0)
#define THETA(n0, n1, n2, n4) \
(t ## n0 = p ## n0 ^ p ## n1 ^ p ## n4)
#define SIGMA_ALL do { \
a0 = t0 ^ 1; \
a1 = t1 ^ INW2(0); \
a2 = t2 ^ INW2(1); \
a3 = t3 ^ INW2(2); \
a4 = t4 ^ INW2(3); \
a5 = t5 ^ INW2(4); \
a6 = t6 ^ INW2(5); \
a7 = t7 ^ INW2(6); \
a8 = t8 ^ INW2(7); \
a9 = t9 ^ sc->buffer[ptr16][0]; \
a10 = t10 ^ sc->buffer[ptr16][1]; \
a11 = t11 ^ sc->buffer[ptr16][2]; \
a12 = t12 ^ sc->buffer[ptr16][3]; \
a13 = t13 ^ sc->buffer[ptr16][4]; \
a14 = t14 ^ sc->buffer[ptr16][5]; \
a15 = t15 ^ sc->buffer[ptr16][6]; \
a16 = t16 ^ sc->buffer[ptr16][7]; \
} while (0)
#define PANAMA_STEP do { \
unsigned ptr16, ptr24, ptr31; \
\
ptr24 = (ptr0 - 8) & 31; \
ptr31 = (ptr0 - 1) & 31; \
BUPDATE; \
M17(GAMMA); \
PI_ALL; \
M17(THETA); \
ptr16 = ptr0 ^ 16; \
SIGMA_ALL; \
ptr0 = ptr31; \
} while (0)
/*
* These macros are used to compute
*/
#define INC0 1
#define INC1 2
#define INC2 3
#define INC3 4
#define INC4 5
#define INC5 6
#define INC6 7
#define INC7 8
/*
* Push data by blocks of 32 bytes. "pbuf" must be 32-bit aligned. Each
* iteration processes 32 data bytes; "num" contains the number of
* iterations.
*/
static void
panama_push(sph_panama_context *sc, const unsigned char *pbuf, size_t num)
{
LVARS
unsigned ptr0;
#if SPH_LITTLE_FAST
#define INW1(i) sph_dec32le_aligned(pbuf + 4 * (i))
#else
sph_u32 X_var[8];
#define INW1(i) X_var[i]
#endif
#define INW2(i) INW1(i)
M17(RSTATE);
ptr0 = sc->buffer_ptr;
while (num -- > 0) {
#if !SPH_LITTLE_FAST
int i;
for (i = 0; i < 8; i ++)
X_var[i] = sph_dec32le_aligned(pbuf + 4 * (i));
#endif
PANAMA_STEP;
pbuf = (const unsigned char *)pbuf + 32;
}
M17(WSTATE);
sc->buffer_ptr = ptr0;
#undef INW1
#undef INW2
}
/*
* Perform the "pull" operation repeatedly ("num" times). The hash output
* will be extracted from the state afterwards.
*/
static void
panama_pull(sph_panama_context *sc, unsigned num)
{
LVARS
unsigned ptr0;
#define INW1(i) INW_H1(INC ## i)
#define INW_H1(i) INW_H2(i)
#define INW_H2(i) a ## i
#define INW2(i) sc->buffer[ptr4][i]
M17(RSTATE);
ptr0 = sc->buffer_ptr;
while (num -- > 0) {
unsigned ptr4;
ptr4 = (ptr0 + 4) & 31;
PANAMA_STEP;
}
M17(WSTATE);
#undef INW1
#undef INW_H1
#undef INW_H2
#undef INW2
}
/* see sph_panama.h */
void
sph_panama_init(void *cc)
{
sph_panama_context *sc;
sc = cc;
/*
* This is not completely conformant, but "it will work
* everywhere". Initial state consists of zeroes everywhere.
* Conceptually, the sph_u32 type may have padding bits which
* must not be set to 0; but such an architecture remains to
* be seen.
*/
sc->data_ptr = 0;
memset(sc->buffer, 0, sizeof sc->buffer);
sc->buffer_ptr = 0;
memset(sc->state, 0, sizeof sc->state);
}
#ifdef SPH_UPTR
static void
panama_short(void *cc, const void *data, size_t len)
#else
void
sph_panama(void *cc, const void *data, size_t len)
#endif
{
sph_panama_context *sc;
unsigned current;
sc = cc;
current = sc->data_ptr;
while (len > 0) {
unsigned clen;
clen = (sizeof sc->data) - current;
if (clen > len)
clen = len;
memcpy(sc->data + current, data, clen);
data = (const unsigned char *)data + clen;
len -= clen;
current += clen;
if (current == sizeof sc->data) {
current = 0;
panama_push(sc, sc->data, 1);
}
}
sc->data_ptr = current;
}
#ifdef SPH_UPTR
/* see sph_panama.h */
void
sph_panama(void *cc, const void *data, size_t len)
{
sph_panama_context *sc;
unsigned current;
size_t rlen;
if (len < (2 * sizeof sc->data)) {
panama_short(cc, data, len);
return;
}
sc = cc;
current = sc->data_ptr;
if (current > 0) {
unsigned t;
t = (sizeof sc->data) - current;
panama_short(sc, data, t);
data = (const unsigned char *)data + t;
len -= t;
}
#if !SPH_UNALIGNED
if (((SPH_UPTR)data & 3) != 0) {
panama_short(sc, data, len);
return;
}
#endif
panama_push(sc, data, len >> 5);
rlen = len & 31;
if (rlen > 0)
memcpy(sc->data,
(const unsigned char *)data + len - rlen, rlen);
sc->data_ptr = rlen;
}
#endif
/* see sph_panama.h */
void
sph_panama_close(void *cc, void *dst)
{
sph_panama_context *sc;
unsigned current;
int i;
sc = cc;
current = sc->data_ptr;
sc->data[current ++] = 0x01;
memset(sc->data + current, 0, (sizeof sc->data) - current);
panama_push(sc, sc->data, 1);
panama_pull(sc, 32);
for (i = 0; i < 8; i ++)
sph_enc32le((unsigned char *)dst + 4 * i, sc->state[i + 9]);
sph_panama_init(sc);
}

118
algo/panama/sph_panama.h Normal file
View File

@@ -0,0 +1,118 @@
/* $Id: sph_panama.h 154 2010-04-26 17:00:24Z tp $ */
/**
* PANAMA interface.
*
* PANAMA has been published in: J. Daemen and C. Clapp, "Fast Hashing
* and Stream Encryption with PANAMA", Fast Software Encryption -
* FSE'98, LNCS 1372, Springer (1998), pp. 60--74.
*
* PANAMA is not fully defined with regards to endianness and related
* topics. This implementation follows strict little-endian conventions:
* <ul>
* <li>Each 32-byte input block is split into eight 32-bit words, the
* first (leftmost) word being numbered 0.</li>
* <li>Each such 32-bit word is decoded from memory in little-endian
* convention.</li>
* <li>The additional padding bit equal to "1" is added by considering
* the least significant bit in a byte to come first; practically, this
* means that a single byte of value 0x01 is appended to the (byte-oriented)
* message, and then 0 to 31 bytes of value 0x00.</li>
* <li>The output consists of eight 32-bit words; the word numbered 0 is
* written first (in leftmost position) and it is encoded in little-endian
* convention.
* </ul>
* With these conventions, PANAMA is sometimes known as "PANAMA-LE". The
* PANAMA reference implementation uses our conventions for input, but
* prescribes no convention for output.
*
* ==========================(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_panama.h
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#ifndef SPH_PANAMA_H__
#define SPH_PANAMA_H__
#include <stddef.h>
#include "algo/sha/sph_types.h"
/**
* Output size (in bits) for PANAMA.
*/
#define SPH_SIZE_panama 256
/**
* This structure is a context for PANAMA computations: it contains the
* intermediate values and some data from the last entered block. Once
* a PANAMA computation has been performed, the context can be reused for
* another computation.
*
* The contents of this structure are private. A running PANAMA computation
* can be cloned by copying the context (e.g. with a simple
* <code>memcpy()</code>).
*/
typedef struct {
#ifndef DOXYGEN_IGNORE
unsigned char data[32]; /* first field, for alignment */
unsigned data_ptr;
sph_u32 buffer[32][8];
unsigned buffer_ptr;
sph_u32 state[17];
#endif
} sph_panama_context;
/**
* Initialize a PANAMA context. This process performs no memory allocation.
*
* @param cc the PANAMA context (pointer to a <code>sph_panama_context</code>)
*/
void sph_panama_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the PANAMA context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_panama(void *cc, const void *data, size_t len);
/**
* Terminate the current PANAMA computation and output the result into the
* provided buffer. The destination buffer must be wide enough to
* accomodate the result (32 bytes). The context is automatically
* reinitialized.
*
* @param cc the PANAMA context
* @param dst the destination buffer
*/
void sph_panama_close(void *cc, void *dst);
#endif

48
algo/quark/anime-4way.c
View File

@@ -48,36 +48,36 @@ void anime_4way_hash( void *state, const void *input )
__m256i* vhA = (__m256i*)vhashA;
__m256i* vhB = (__m256i*)vhashB;
__m256i vh_mask;
__m256i bit3_mask; bit3_mask = _mm256_set1_epi64x( 8 );
const __m256i bit3_mask = _mm256_set1_epi64x( 8 );
int i;
anime_4way_ctx_holder ctx;
memcpy( &ctx, &anime_4way_ctx, sizeof(anime_4way_ctx) );
bmw512_4way( &ctx.bmw, vhash, 80 );
bmw512_4way( &ctx.bmw, input, 80 );
bmw512_4way_close( &ctx.bmw, vhash );
blake512_4way( &ctx.blake, input, 64 );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
m256_zero );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
update_and_final_groestl( &ctx.groestl, (char*)hash0,
(char*)hash0, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash1,
(char*)hash1, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash2,
(char*)hash2, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash3,
(char*)hash3, 512 );
mm256_interleave_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
update_and_final_groestl( &ctx.groestl, (char*)hash0,
(char*)hash0, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash1,
(char*)hash1, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash2,
(char*)hash2, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash3,
(char*)hash3, 512 );
mm256_interleave_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhashB );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhashB );
for ( i = 0; i < 8; i++ )
vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
@@ -120,13 +120,13 @@ void anime_4way_hash( void *state, const void *input )
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
m256_zero );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhashA );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhashA );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhashB );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhashB );
for ( i = 0; i < 8; i++ )
vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );

1003
algo/radiogatun/sph_radiogatun.c Normal file
View File

File diff suppressed because it is too large Load Diff

186
algo/radiogatun/sph_radiogatun.h Normal file
View File

@@ -0,0 +1,186 @@
/* $Id: sph_radiogatun.h 226 2010-06-16 17:28:08Z tp $ */
/**
* RadioGatun interface.
*
* RadioGatun has been published in: G. Bertoni, J. Daemen, M. Peeters
* and G. Van Assche, "RadioGatun, a belt-and-mill hash function",
* presented at the Second Cryptographic Hash Workshop, Santa Barbara,
* August 24-25, 2006. The main Web site, containing that article, the
* reference code and some test vectors, appears to be currently located
* at the following URL: http://radiogatun.noekeon.org/
*
* The presentation article does not specify endianness or padding. The
* reference code uses the following conventions, which we also apply
* here:
* <ul>
* <li>The input message is an integral number of sequences of three
* words. Each word is either a 32-bit of 64-bit word (depending on
* the version of RadioGatun).</li>
* <li>Input bytes are decoded into words using little-endian
* convention.</li>
* <li>Padding consists of a single bit of value 1, using little-endian
* convention within bytes (i.e. for a byte-oriented input, a single
* byte of value 0x01 is appended), then enough bits of value 0 to finish
* the current block.</li>
* <li>Output consists of 256 bits. Successive output words are encoded
* with little-endian convention.</li>
* </ul>
* These conventions are very close to those we use for PANAMA, which is
* a close ancestor or RadioGatun.
*
* RadioGatun is actually a family of functions, depending on some
* internal parameters. We implement here two functions, with a "belt
* length" of 13, a "belt width" of 3, and a "mill length" of 19. The
* RadioGatun[32] version uses 32-bit words, while the RadioGatun[64]
* variant uses 64-bit words.
*
* Strictly speaking, the name "RadioGatun" should use an acute accent
* on the "u", which we omitted here to keep strict ASCII-compatibility
* of this file.
*
* ==========================(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_radiogatun.h
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#ifndef SPH_RADIOGATUN_H__
#define SPH_RADIOGATUN_H__
#include <stddef.h>
#include "algo/sha/sph_types.h"
/**
* Output size (in bits) for RadioGatun[32].
*/
#define SPH_SIZE_radiogatun32 256
/**
* This structure is a context for RadioGatun[32] computations: it
* contains intermediate values and some data from the last entered
* block. Once a RadioGatun[32] computation has been performed, the
* context can be reused for another computation.
*
* The contents of this structure are private. A running RadioGatun[32]
* computation can be cloned by copying the context (e.g. with a
* simple <code>memcpy()</code>).
*/
typedef struct {
#ifndef DOXYGEN_IGNORE
unsigned char data[156]; /* first field, for alignment */
unsigned data_ptr;
sph_u32 a[19], b[39];
#endif
} sph_radiogatun32_context;
/**
* Initialize a RadioGatun[32] context. This process performs no
* memory allocation.
*
* @param cc the RadioGatun[32] context (pointer to a
* <code>sph_radiogatun32_context</code>)
*/
void sph_radiogatun32_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the RadioGatun[32] context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_radiogatun32(void *cc, const void *data, size_t len);
/**
* Terminate the current RadioGatun[32] computation and output the
* result into the provided buffer. The destination buffer must be wide
* enough to accomodate the result (32 bytes). The context is
* automatically reinitialized.
*
* @param cc the RadioGatun[32] context
* @param dst the destination buffer
*/
void sph_radiogatun32_close(void *cc, void *dst);
#if SPH_64
/**
* Output size (in bits) for RadioGatun[64].
*/
#define SPH_SIZE_radiogatun64 256
/**
* This structure is a context for RadioGatun[64] computations: it
* contains intermediate values and some data from the last entered
* block. Once a RadioGatun[64] computation has been performed, the
* context can be reused for another computation.
*
* The contents of this structure are private. A running RadioGatun[64]
* computation can be cloned by copying the context (e.g. with a
* simple <code>memcpy()</code>).
*/
typedef struct {
#ifndef DOXYGEN_IGNORE
unsigned char data[312]; /* first field, for alignment */
unsigned data_ptr;
sph_u64 a[19], b[39];
#endif
} sph_radiogatun64_context;
/**
* Initialize a RadioGatun[64] context. This process performs no
* memory allocation.
*
* @param cc the RadioGatun[64] context (pointer to a
* <code>sph_radiogatun64_context</code>)
*/
void sph_radiogatun64_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the RadioGatun[64] context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_radiogatun64(void *cc, const void *data, size_t len);
/**
* Terminate the current RadioGatun[64] computation and output the
* result into the provided buffer. The destination buffer must be wide
* enough to accomodate the result (32 bytes). The context is
* automatically reinitialized.
*
* @param cc the RadioGatun[64] context
* @param dst the destination buffer
*/
void sph_radiogatun64_close(void *cc, void *dst);
#endif
#endif

3
algo/ripemd/lbry-gate.h
View File

@@ -4,10 +4,13 @@
#include "algo-gate-api.h"
#include <stdint.h>
// Overide multi way on ryzen, SHA is better.
#if !defined(RYZEN_)
// need sha512 2 way AVX x2 or 1 way scalar x4 to support 4way AVX.
#if defined(__AVX2__)
#define LBRY_8WAY
#endif
#endif
#define LBRY_NTIME_INDEX 25
#define LBRY_NBITS_INDEX 26

30
algo/ripemd/lbry.c
View File

@@ -4,24 +4,17 @@
#include <string.h>
#include <stdio.h>
#include "sph_ripemd.h"
#include "algo/sha/sph_sha2.h"
#include <openssl/sha.h>
void lbry_hash(void* output, const void* input)
{
#ifndef USE_SPH_SHA
SHA256_CTX ctx_sha256 __attribute__ ((aligned (64)));
SHA512_CTX ctx_sha512 __attribute__ ((aligned (64)));
#else
sph_sha256_context ctx_sha256 __attribute__ ((aligned (64)));
sph_sha512_context ctx_sha512 __attribute__ ((aligned (64)));
#endif
sph_ripemd160_context ctx_ripemd __attribute__ ((aligned (64)));
uint32_t _ALIGN(64) hashA[16];
uint32_t _ALIGN(64) hashB[16];
uint32_t _ALIGN(64) hashC[16];
#ifndef USE_SPH_SHA
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, input, 112 );
SHA256_Final( (unsigned char*) hashA, &ctx_sha256 );
@@ -33,19 +26,6 @@ void lbry_hash(void* output, const void* input)
SHA512_Init( &ctx_sha512 );
SHA512_Update( &ctx_sha512, hashA, 32 );
SHA512_Final( (unsigned char*) hashA, &ctx_sha512 );
#else
sph_sha256_init( &ctx_sha256 );
sph_sha256 ( &ctx_sha256, input, 112 );
sph_sha256_close( &ctx_sha256, hashA );
sph_sha256_init( &ctx_sha256 );
sph_sha256 ( &ctx_sha256, hashA, 32 );
sph_sha256_close( &ctx_sha256, hashA );
sph_sha512_init( &ctx_sha512 );
sph_sha512 ( &ctx_sha512, hashA, 32 );
sph_sha512_close( &ctx_sha512, hashA );
#endif
sph_ripemd160_init( &ctx_ripemd );
sph_ripemd160 ( &ctx_ripemd, hashA, 32 );
@@ -55,7 +35,6 @@ void lbry_hash(void* output, const void* input)
sph_ripemd160 ( &ctx_ripemd, hashA+8, 32 );
sph_ripemd160_close( &ctx_ripemd, hashC );
#ifndef USE_SPH_SHA
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, hashB, 20 );
SHA256_Update( &ctx_sha256, hashC, 20 );
@@ -64,16 +43,7 @@ void lbry_hash(void* output, const void* input)
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, hashA, 32 );
SHA256_Final( (unsigned char*) hashA, &ctx_sha256 );
#else
sph_sha256_init( &ctx_sha256 );
sph_sha256 ( &ctx_sha256, hashB, 20 );
sph_sha256 ( &ctx_sha256, hashC, 20 );
sph_sha256_close( &ctx_sha256, hashA );
sph_sha256_init( &ctx_sha256 );
sph_sha256 ( &ctx_sha256, hashA, 32 );
sph_sha256_close( &ctx_sha256, hashA );
#endif
memcpy( output, hashA, 32 );
}

2
algo/scryptjane/scrypt-jane-portable-x86.h
View File

@@ -296,6 +296,7 @@ get_xgetbv(uint32_t flags) {
size_t cpu_detect_mask = (size_t)-1;
#endif
#if 0
static size_t
detect_cpu(void) {
union { uint8_t s[12]; uint32_t i[3]; } vendor_string;
@@ -354,6 +355,7 @@ detect_cpu(void) {
return cpu_flags;
}
#endif
#if defined(SCRYPT_TEST_SPEED)
static const char *

88
algo/sha/sha256t-4way.c
View File

@@ -28,8 +28,8 @@ void sha256t_8way_hash( void* output, const void* input )
}
int scanhash_sha256t_8way( int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done )
int scanhash_sha256t_8way( int thr_id, 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)));
@@ -39,9 +39,8 @@ int scanhash_sha256t_8way( int thr_id, struct work *work,
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 152; // 19*8
__m256i *noncev = (__m256i*)vdata + 19; // aligned
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
const uint64_t htmax[] = { 0,
0xF,
@@ -56,27 +55,25 @@ int scanhash_sha256t_8way( int thr_id, struct work *work,
0xFFFF0000,
0 };
for ( int k = 0; k < 20; k++ )
be32enc( &edata[k], pdata[k] );
// Need big endian data
casti_m256i( edata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
casti_m256i( edata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
mm256_interleave_8x32( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
sha256_8way_init( &sha256_ctx8 );
sha256_8way( &sha256_ctx8, vdata, 64 );
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do {
be32enc( noncep, n );
be32enc( noncep +1, n+1 );
be32enc( noncep +2, n+2 );
be32enc( noncep +3, n+3 );
be32enc( noncep +4, n+4 );
be32enc( noncep +5, n+5 );
be32enc( noncep +6, n+6 );
be32enc( noncep +7, n+7 );
pdata[19] = n;
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;
sha256t_8way_hash( hash, vdata );
@@ -91,20 +88,24 @@ int scanhash_sha256t_8way( int thr_id, struct work *work,
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
pdata[19] = n + lane;
work_set_target_ratio( work, lane_hash );
if ( submit_work( mythr, work ) )
applog( LOG_NOTICE, "Share %d submitted by thread %d, lane %d.",
accepted_share_count + rejected_share_count + 1,
thr_id, lane );
else
applog( LOG_WARNING, "Failed to submit share." );
}
}
n += 8;
} while ( (num_found == 0) && (n < max_nonce)
&& !work_restart[thr_id].restart );
} while ( (n < max_nonce-10) && !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
return 0;
}
#elif defined(SHA256T_4WAY)
@@ -130,8 +131,8 @@ void sha256t_4way_hash( void* output, const void* input )
}
int scanhash_sha256t_4way( int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done )
int scanhash_sha256t_4way( int thr_id, 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)));
@@ -143,9 +144,8 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 76; // 19*4
__m128i *noncev = (__m128i*)vdata + 19; // aligned
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
const uint64_t htmax[] = { 0,
0xF,
@@ -160,8 +160,11 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
0xFFFF0000,
0 };
for ( int k = 0; k < 19; k++ )
be32enc( &edata[k], pdata[k] );
casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
sha256_4way_init( &sha256_ctx4 );
@@ -171,11 +174,8 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
{
uint32_t mask = masks[m];
do {
be32enc( noncep, n );
be32enc( noncep +1, n+1 );
be32enc( noncep +2, n+2 );
be32enc( noncep +3, n+3 );
pdata[19] = n;
*noncev = mm128_bswap_32( _mm_set_epi32( n+3,n+2,n+1,n ) );
pdata[19] = n;
sha256t_4way_hash( hash, vdata );
@@ -186,21 +186,25 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
pdata[19] = n + lane;
work_set_target_ratio( work, lane_hash );
if ( submit_work( mythr, work ) )
applog( LOG_NOTICE, "Share %d submitted by thread %d, lane %d.",
accepted_share_count + rejected_share_count + 1,
thr_id, lane );
else
applog( LOG_WARNING, "Failed to submit share." );
}
}
n += 4;
} while ( (num_found == 0) && (n < max_nonce)
&& !work_restart[thr_id].restart );
} while ( (n < max_nonce - 4) && !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
return 0;
}
#endif

9
algo/sha/sha256t-gate.h
View File

@@ -4,12 +4,15 @@
#include <stdint.h>
#include "algo-gate-api.h"
// Override multi way on ryzen, SHA is better.
#if !defined(RYZEN_)
#if defined(__SSE4_2__)
#define SHA256T_4WAY
#endif
#if defined(__AVX2__)
#define SHA256T_8WAY
#endif
#endif
bool register_blake2s_algo( algo_gate_t* gate );
@@ -17,18 +20,18 @@ bool register_blake2s_algo( algo_gate_t* gate );
void sha256t_8way_hash( void *output, const void *input );
int scanhash_sha256t_8way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
#elif defined (SHA256T_4WAY)
void sha256t_4way_hash( void *output, const void *input );
int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
#else
void sha256t_hash( void *output, const void *input );
int scanhash_sha256t( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
#endif

3
algo/sha/sha256t.c
View File

@@ -39,7 +39,7 @@ void sha256t_hash( void* output, const void* input )
}
int scanhash_sha256t( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done)
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
@@ -52,6 +52,7 @@ int scanhash_sha256t( int thr_id, struct work *work, uint32_t max_nonce,
uint32_t hash64[8] __attribute__((aligned(32)));
#endif
uint32_t endiandata[32];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] = {
0,

32
algo/shabal/shabal-hash-4way.c
View File

@@ -248,22 +248,22 @@ do { \
*/
#define SWAP_BC \
do { \
mm128_swap256_128( B0, C0 ); \
mm128_swap256_128( B1, C1 ); \
mm128_swap256_128( B2, C2 ); \
mm128_swap256_128( B3, C3 ); \
mm128_swap256_128( B4, C4 ); \
mm128_swap256_128( B5, C5 ); \
mm128_swap256_128( B6, C6 ); \
mm128_swap256_128( B7, C7 ); \
mm128_swap256_128( B8, C8 ); \
mm128_swap256_128( B9, C9 ); \
mm128_swap256_128( BA, CA ); \
mm128_swap256_128( BB, CB ); \
mm128_swap256_128( BC, CC ); \
mm128_swap256_128( BD, CD ); \
mm128_swap256_128( BE, CE ); \
mm128_swap256_128( BF, CF ); \
mm128_swap128_256( B0, C0 ); \
mm128_swap128_256( B1, C1 ); \
mm128_swap128_256( B2, C2 ); \
mm128_swap128_256( B3, C3 ); \
mm128_swap128_256( B4, C4 ); \
mm128_swap128_256( B5, C5 ); \
mm128_swap128_256( B6, C6 ); \
mm128_swap128_256( B7, C7 ); \
mm128_swap128_256( B8, C8 ); \
mm128_swap128_256( B9, C9 ); \
mm128_swap128_256( BA, CA ); \
mm128_swap128_256( BB, CB ); \
mm128_swap128_256( BC, CC ); \
mm128_swap128_256( BD, CD ); \
mm128_swap128_256( BE, CE ); \
mm128_swap128_256( BF, CF ); \
} while (0)
#define PERM_ELT(xa0, xa1, xb0, xb1, xb2, xb3, xc, xm) \

406
algo/shavite/shavite-hash-2way.c Normal file
View File

@@ -0,0 +1,406 @@
#include "shavite-hash-2way.h"
#include "algo/sha/sph_types.h"
#include <stdio.h>
#if defined(__AVX2__)
static const uint32_t IV512[] =
{
0x72FCCDD8, 0x79CA4727, 0x128A077B, 0x40D55AEC,
0xD1901A06, 0x430AE307, 0xB29F5CD1, 0xDF07FBFC,
0x8E45D73D, 0x681AB538, 0xBDE86578, 0xDD577E47,
0xE275EADE, 0x502D9FCD, 0xB9357178, 0x022A4B9A
};
#define mm256_ror2x256hi_1x32( a, b ) \
_mm256_blend_epi32( mm256_ror1x32_128( a ), \
mm256_ror1x32_128( b ), 0x88 )
static void
c512_2way( shavite512_2way_context *ctx, const void *msg )
{
__m256i p0, p1, p2, p3, x;
__m256i k00, k01, k02, k03, k10, k11, k12, k13;
__m256i *m = (__m256i*)msg;
__m256i *h = (__m256i*)ctx->h;
int r;
p0 = h[0];
p1 = h[1];
p2 = h[2];
p3 = h[3];
// round
k00 = m[0];
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k00 ) );
k01 = m[1];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = m[2];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = m[3];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p0 = _mm256_xor_si256( p0, x );
k10 = m[4];
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k10 ) );
k11 = m[5];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = m[6];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = m[7];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p2 = _mm256_xor_si256( p2, x );
for ( r = 0; r < 3; r ++ )
{
// round 1, 5, 9
k00 = _mm256_xor_si256( k13, mm256_ror1x32_128(
mm256_aesenc_2x128( k00 ) ) );
if ( r == 0 )
k00 = _mm256_xor_si256( k00, _mm256_set_epi32(
~ctx->count3, ctx->count2, ctx->count1, ctx->count0,
~ctx->count3, ctx->count2, ctx->count1, ctx->count0 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p0, k00 ) );
k01 = _mm256_xor_si256( k00,
mm256_ror1x32_128( mm256_aesenc_2x128( k01 ) ) );
if ( r == 1 )
k01 = _mm256_xor_si256( k01, _mm256_set_epi32(
~ctx->count0, ctx->count1, ctx->count2, ctx->count3,
~ctx->count0, ctx->count1, ctx->count2, ctx->count3 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( k01,
mm256_ror1x32_128( mm256_aesenc_2x128( k02 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( k02,
mm256_ror1x32_128( mm256_aesenc_2x128( k03 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p3 = _mm256_xor_si256( p3, x );
k10 = _mm256_xor_si256( k03,
mm256_ror1x32_128( mm256_aesenc_2x128( k10 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p2, k10 ) );
k11 = _mm256_xor_si256( k10,
mm256_ror1x32_128( mm256_aesenc_2x128( k11 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( k11,
mm256_ror1x32_128( mm256_aesenc_2x128( k12 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( k12,
mm256_ror1x32_128( mm256_aesenc_2x128( k13 ) ) );
if ( r == 2 )
k13 = _mm256_xor_si256( k13, _mm256_set_epi32(
~ctx->count1, ctx->count0, ctx->count3, ctx->count2,
~ctx->count1, ctx->count0, ctx->count3, ctx->count2 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p1 = _mm256_xor_si256( p1, x );
// round 2, 6, 10
k00 = _mm256_xor_si256( k00, mm256_ror2x256hi_1x32( k12, k13 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k00 ) );
k01 = _mm256_xor_si256( k01, mm256_ror2x256hi_1x32( k13, k00 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( k02, mm256_ror2x256hi_1x32( k00, k01 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( k03, mm256_ror2x256hi_1x32( k01, k02 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p2 = _mm256_xor_si256( p2, x );
k10 = _mm256_xor_si256( k10, mm256_ror2x256hi_1x32( k02, k03 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k10 ) );
k11 = _mm256_xor_si256( k11, mm256_ror2x256hi_1x32( k03, k10 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( k12, mm256_ror2x256hi_1x32( k10, k11 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( k13, mm256_ror2x256hi_1x32( k11, k12 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p0 = _mm256_xor_si256( p0, x );
// round 3, 7, 11
k00 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k00 ) ), k13 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p2, k00 ) );
k01 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k01 ) ), k00 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k02 ) ), k01 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k03 ) ), k02 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p1 = _mm256_xor_si256( p1, x );
k10 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k10 ) ), k03 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p0, k10 ) );
k11 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k11 ) ), k10 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k12 ) ), k11 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k13 ) ), k12 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p3 = _mm256_xor_si256( p3, x );
// round 4, 8, 12
k00 = _mm256_xor_si256( k00, mm256_ror2x256hi_1x32( k12, k13 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k00 ) );
k01 = _mm256_xor_si256( k01, mm256_ror2x256hi_1x32( k13, k00 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( k02, mm256_ror2x256hi_1x32( k00, k01 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( k03, mm256_ror2x256hi_1x32( k01, k02 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p0 = _mm256_xor_si256( p0, x );
k10 = _mm256_xor_si256( k10, mm256_ror2x256hi_1x32( k02, k03 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k10 ) );
k11 = _mm256_xor_si256( k11, mm256_ror2x256hi_1x32( k03, k10 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( k12, mm256_ror2x256hi_1x32( k10, k11 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( k13, mm256_ror2x256hi_1x32( k11, k12 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p2 = _mm256_xor_si256( p2, x );
}
// round 13
k00 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k00 ) ), k13 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p0, k00 ) );
k01 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k01 ) ), k00 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k02 ) ), k01 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k03 ) ), k02 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p3 = _mm256_xor_si256( p3, x );
k10 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k10 ) ), k03 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p2, k10 ) );
k11 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k11 ) ), k10 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = mm256_ror1x32_128( mm256_aesenc_2x128( k12 ) );
k12 = _mm256_xor_si256( k12, _mm256_xor_si256( k11, _mm256_set_epi32(
~ctx->count2, ctx->count3, ctx->count0, ctx->count1,
~ctx->count2, ctx->count3, ctx->count0, ctx->count1 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k13 ) ), k12 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p1 = _mm256_xor_si256( p1, x );
h[0] = _mm256_xor_si256( h[0], p2 );
h[1] = _mm256_xor_si256( h[1], p3 );
h[2] = _mm256_xor_si256( h[2], p0 );
h[3] = _mm256_xor_si256( h[3], p1 );
}
void shavite512_2way_init( shavite512_2way_context *ctx )
{
casti_m256i( ctx->h, 0 ) =
_mm256_set_epi32( IV512[ 3], IV512[ 2], IV512[ 1], IV512[ 0],
IV512[ 3], IV512[ 2], IV512[ 1], IV512[ 0] );
casti_m256i( ctx->h, 1 ) =
_mm256_set_epi32( IV512[ 7], IV512[ 6], IV512[ 5], IV512[ 4],
IV512[ 7], IV512[ 6], IV512[ 5], IV512[ 4] );
casti_m256i( ctx->h, 2 ) =
_mm256_set_epi32( IV512[11], IV512[10], IV512[ 9], IV512[ 8],
IV512[11], IV512[10], IV512[ 9], IV512[ 8] );
casti_m256i( ctx->h, 3 ) =
_mm256_set_epi32( IV512[15], IV512[14], IV512[13], IV512[12],
IV512[15], IV512[14], IV512[13], IV512[12] );
ctx->ptr = 0;
ctx->count0 = 0;
ctx->count1 = 0;
ctx->count2 = 0;
ctx->count3 = 0;
}
void shavite512_2way_update( shavite512_2way_context *ctx, const void *data,
size_t len )
{
unsigned char *buf = ctx->buf;
size_t ptr = ctx->ptr;
while ( len > 0 )
{
size_t clen;
clen = (sizeof ctx->buf) - ptr;
if ( clen > len << 1 )
clen = len << 1;
memcpy( buf + ptr, data, clen );
data = (const unsigned char *)data + clen;
ptr += clen;
len -= clen >> 1;
if ( ptr == sizeof ctx->buf )
{
if ( ( ctx->count0 = ctx->count0 + 1024 ) == 0 )
{
ctx->count1 = ctx->count1 + 1;
if ( ctx->count1 == 0 )
{
ctx->count2 = ctx->count2 + 1;
if ( ctx->count2 == 0 )
ctx->count3 = ctx->count3 + 1;
}
}
c512_2way( ctx, buf );
ptr = 0;
}
}
ctx->ptr = ptr;
}
void shavite512_2way_close( shavite512_2way_context *ctx, void *dst )
{
unsigned char *buf;
union
{
uint32_t u32[4];
uint16_t u16[8];
} count;
buf = ctx->buf;
uint32_t vp = ctx->ptr>>5;
// Terminating byte then zero pad
casti_m256i( buf, vp++ ) = _mm256_set_epi32( 0,0,0,0x80, 0,0,0,0x80 );
// Zero pad full vectors up to count
for ( ; vp < 6; vp++ )
casti_m256i( buf, vp ) = m256_zero;
// Count = { 0, 16, 64, 80 }. Outsize = 16 u32 = 512 bits = 0x0200
// Count is misaligned to 16 bits and straddles a vector.
// Use u32 overlay to stage then u16 to load buf.
count.u32[0] = ctx->count0 += (ctx->ptr << 2); // ptr/2 * 8
count.u32[1] = ctx->count1;
count.u32[2] = ctx->count2;
count.u32[3] = ctx->count3;
casti_m256i( buf, 6 ) = _mm256_set_epi16( count.u16[0], 0,0,0,0,0,0,0,
count.u16[0], 0,0,0,0,0,0,0 );
casti_m256i( buf, 7 ) = _mm256_set_epi16(
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1],
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] );
c512_2way( ctx, buf);
casti_m256i( dst, 0 ) = casti_m256i( ctx->h, 0 );
casti_m256i( dst, 1 ) = casti_m256i( ctx->h, 1 );
casti_m256i( dst, 2 ) = casti_m256i( ctx->h, 2 );
casti_m256i( dst, 3 ) = casti_m256i( ctx->h, 3 );
}
void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
const void *data, size_t len )
{
unsigned char *buf = ctx->buf;
size_t ptr = ctx->ptr;
// process full blocks and load buf with remainder.
while ( len > 0 )
{
size_t clen;
clen = (sizeof ctx->buf) - ptr;
if ( clen > len << 1 )
clen = len << 1;
memcpy( buf + ptr, data, clen );
data = (const unsigned char *)data + clen;
ptr += clen;
len -= clen >> 1;
if ( ptr == sizeof ctx->buf )
{
if ( ( ctx->count0 = ctx->count0 + 1024 ) == 0 )
{
ctx->count1 = ctx->count1 + 1;
if ( ctx->count1 == 0 )
{
ctx->count2 = ctx->count2 + 1;
if ( ctx->count2 == 0 )
ctx->count3 = ctx->count3 + 1;
}
}
c512_2way( ctx, buf );
ptr = 0;
}
}
uint32_t vp = ptr>>5;
// Terminating byte then zero pad
casti_m256i( buf, vp++ ) = _mm256_set_epi32( 0,0,0,0x80, 0,0,0,0x80 );
// Zero pad full vectors up to count
for ( ; vp < 6; vp++ )
casti_m256i( buf, vp ) = m256_zero;
// Count = { 0, 16, 64, 80 }. Outsize = 16 u32 = 512 bits = 0x0200
// Count is misaligned to 16 bits and straddles a vector.
// Use u32 overlay to stage then u16 to load buf.
union
{
uint32_t u32[4];
uint16_t u16[8];
} count;
count.u32[0] = ctx->count0 += (ptr << 2); // ptr/2 * 8
count.u32[1] = ctx->count1;
count.u32[2] = ctx->count2;
count.u32[3] = ctx->count3;
casti_m256i( buf, 6 ) = _mm256_set_epi16( count.u16[0], 0,0,0,0,0,0,0,
count.u16[0], 0,0,0,0,0,0,0 );
casti_m256i( buf, 7 ) = _mm256_set_epi16(
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1],
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] );
c512_2way( ctx, buf);
casti_m256i( dst, 0 ) = casti_m256i( ctx->h, 0 );
casti_m256i( dst, 1 ) = casti_m256i( ctx->h, 1 );
casti_m256i( dst, 2 ) = casti_m256i( ctx->h, 2 );
casti_m256i( dst, 3 ) = casti_m256i( ctx->h, 3 );
}
#endif // AVX2

25
algo/shavite/shavite-hash-2way.h Normal file
View File

@@ -0,0 +1,25 @@
#ifndef SHAVITE_HASH_2WAY_H__
#define SHAVITE_HASH_2WAY_H__
#if defined(__AVX2__)
#include "avxdefs.h"
typedef struct {
unsigned char buf[128<<1];
uint32_t h[16<<1];
size_t ptr;
uint32_t count0, count1, count2, count3;
} shavite512_2way_context __attribute__ ((aligned (64)));
void shavite512_2way_init( shavite512_2way_context *ctx );
void shavite512_2way_update( shavite512_2way_context *ctx, const void *data,
size_t len );
void shavite512_2way_close( shavite512_2way_context *ctx, void *dst );
void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
const void *data, size_t len );
#endif // AVX2
#endif // SHAVITE_HASH_2WAY_H__

62
algo/shavite/sph-shavite-aesni.c
View File

@@ -102,35 +102,31 @@ c512( sph_shavite_big_context *sc, const void *msg )
k00 = m[0];
x = _mm_xor_si128( p1, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = m[1];
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = m[2];
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = m[3];
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p0 = _mm_xor_si128( p0, x );
k10 = m[4];
x = _mm_xor_si128( p3, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = m[5];
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = m[6];
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = m[7];
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p2 = _mm_xor_si128( p2, x );
for ( r = 0; r < 3; r ++ )
@@ -156,15 +152,15 @@ c512( sph_shavite_big_context *sc, const void *msg )
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm128_ror_1x32( _mm_aesenc_si128( k02, m128_zero ) );
k02 = _mm_xor_si128( k02, k01 );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = mm128_ror_1x32( _mm_aesenc_si128( k03, m128_zero ) );
k03 = _mm_xor_si128( k03, k02 );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p3 = _mm_xor_si128( p3, x );
k10 = mm128_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
k10 = _mm_xor_si128( k10, k03 );
@@ -172,12 +168,10 @@ c512( sph_shavite_big_context *sc, const void *msg )
x = _mm_aesenc_si128( x, m128_zero );
k11 = mm128_ror_1x32( _mm_aesenc_si128( k11, m128_zero ) );
k11 = _mm_xor_si128( k11, k10 );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = mm128_ror_1x32( _mm_aesenc_si128( k12, m128_zero ) );
k12 = _mm_xor_si128( k12, k11 );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
@@ -196,118 +190,103 @@ c512( sph_shavite_big_context *sc, const void *msg )
k00 = _mm_xor_si128( k00, mm128_ror256hi_1x32( k12, k13 ) );
x = _mm_xor_si128( p3, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = _mm_xor_si128( k01, mm128_ror256hi_1x32( k13, k00 ) );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = _mm_xor_si128( k02, mm128_ror256hi_1x32( k00, k01 ) );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = _mm_xor_si128( k03, mm128_ror256hi_1x32( k01, k02 ) );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p2 = _mm_xor_si128( p2, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
x = _mm_xor_si128( p1, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = _mm_xor_si128( k11, mm128_ror256hi_1x32( k03, k10 ) );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = _mm_xor_si128( k12, mm128_ror256hi_1x32( k10, k11 ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p0 = _mm_xor_si128( p0, x );
// round 3, 7, 11
k00 = mm128_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
k00 = _mm_xor_si128( k00, k13 );
x = _mm_xor_si128( p2, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = mm128_ror_1x32( _mm_aesenc_si128( k01, m128_zero ) );
k01 = _mm_xor_si128( k01, k00 );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm128_ror_1x32( _mm_aesenc_si128( k02, m128_zero ) );
k02 = _mm_xor_si128( k02, k01 );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = mm128_ror_1x32( _mm_aesenc_si128( k03, m128_zero ) );
k03 = _mm_xor_si128( k03, k02 );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p1 = _mm_xor_si128( p1, x );
k10 = mm128_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
k10 = _mm_xor_si128( k10, k03 );
x = _mm_xor_si128( p0, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = mm128_ror_1x32( _mm_aesenc_si128( k11, m128_zero ) );
k11 = _mm_xor_si128( k11, k10 );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = mm128_ror_1x32( _mm_aesenc_si128( k12, m128_zero ) );
k12 = _mm_xor_si128( k12, k11 );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p3 = _mm_xor_si128( p3, x );
// round 4, 8, 12
k00 = _mm_xor_si128( k00, mm128_ror256hi_1x32( k12, k13 ) );
x = _mm_xor_si128( p1, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = _mm_xor_si128( k01, mm128_ror256hi_1x32( k13, k00 ) );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = _mm_xor_si128( k02, mm128_ror256hi_1x32( k00, k01 ) );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = _mm_xor_si128( k03, mm128_ror256hi_1x32( k01, k02 ) );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p0 = _mm_xor_si128( p0, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
p0 = _mm_xor_si128( p0, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
x = _mm_xor_si128( p3, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = _mm_xor_si128( k11, mm128_ror256hi_1x32( k03, k10 ) );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = _mm_xor_si128( k12, mm128_ror256hi_1x32( k10, k11 ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p2 = _mm_xor_si128( p2, x );
}
@@ -315,46 +294,41 @@ c512( sph_shavite_big_context *sc, const void *msg )
k00 = mm128_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
k00 = _mm_xor_si128( k00, k13 );
x = _mm_xor_si128( p0, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = mm128_ror_1x32( _mm_aesenc_si128( k01, m128_zero ) );
k01 = _mm_xor_si128( k01, k00 );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm128_ror_1x32( _mm_aesenc_si128( k02, m128_zero ) );
k02 = _mm_xor_si128( k02, k01 );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = mm128_ror_1x32( _mm_aesenc_si128( k03, m128_zero ) );
k03 = _mm_xor_si128( k03, k02 );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p3 = _mm_xor_si128( p3, x );
k10 = mm128_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
k10 = _mm_xor_si128( k10, k03 );
x = _mm_xor_si128( p2, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = mm128_ror_1x32( _mm_aesenc_si128( k11, m128_zero ) );
k11 = _mm_xor_si128( k11, k10 );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = mm128_ror_1x32( _mm_aesenc_si128( k12, m128_zero ) );
k12 = _mm_xor_si128( k12, _mm_xor_si128( k11, _mm_set_epi32(
~sc->count2, sc->count3, sc->count0, sc->count1 ) ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p1 = _mm_xor_si128( p1, x );
h[0] = _mm_xor_si128( h[0], p2 );
@@ -427,6 +401,9 @@ shavite_big_aesni_close( sph_shavite_big_context *sc, unsigned ub, unsigned n,
count1 = sc->count1;
count2 = sc->count2;
count3 = sc->count3;
z = 0x80 >> n;
z = ((ub & -z) | z) & 0xFF;
if (ptr == 0 && n == 0) {
@@ -443,6 +420,7 @@ shavite_big_aesni_close( sph_shavite_big_context *sc, unsigned ub, unsigned n,
memset(buf, 0, 110);
sc->count0 = sc->count1 = sc->count2 = sc->count3 = 0;
}
sph_enc32le(buf + 110, count0);
sph_enc32le(buf + 114, count1);
sph_enc32le(buf + 118, count2);

1764
algo/shavite/sse2/shavite.c
View File

File diff suppressed because it is too large Load Diff

3
algo/skein/skein-gate.h
View File

@@ -3,9 +3,12 @@
#include <stdint.h>
#include "algo-gate-api.h"
// Override multi way on ryzen, SHA is better.
#if !defined(RYZEN_)
#if defined(__AVX2__)
#define SKEIN_4WAY
#endif
#endif
#if defined(SKEIN_4WAY)

11
algo/skein/skein.c
View File

@@ -3,31 +3,20 @@
#include <stdint.h>
#include "sph_skein.h"
#include <openssl/sha.h>
#include "algo/sha/sph_sha2.h"
void skeinhash(void *state, const void *input)
{
uint32_t hash[16] __attribute__ ((aligned (64)));
sph_skein512_context ctx_skein;
#ifndef USE_SPH_SHA
SHA256_CTX ctx_sha256;
#else
sph_sha256_context ctx_sha256;
#endif
sph_skein512_init( &ctx_skein );
sph_skein512( &ctx_skein, input, 80 );
sph_skein512_close( &ctx_skein, hash );
#ifndef USE_SPH_SHA
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, (unsigned char*)hash, 64 );
SHA256_Final( (unsigned char*) hash, &ctx_sha256 );
#else
sph_sha256_init( &ctx_sha256 );
sph_sha256( &ctx_sha256, hash, 64 );
sph_sha256_close( &ctx_sha256, hash );
#endif
memcpy(state, hash, 32);
}

6
algo/x12/x12-4way.c
View File

@@ -114,11 +114,11 @@ void x12_4way_hash( void *state, const void *input )
// 8 Cubehash
cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash1, (const byte*) hash1, 64 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash2, (const byte*) hash2, 64 );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash3, (const byte*) hash3, 64 );
// 9 Shavite

44
algo/x12/x12.c
View File

@@ -17,8 +17,6 @@
#include "algo/simd/sph_simd.h"
#include "algo/echo/sph_echo.h"
#include "algo/hamsi/sph_hamsi.h"
//#include "algo/fugue/sph_fugue.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
@@ -27,45 +25,42 @@
#include "algo/keccak/sse2/keccak.c"
#include "algo/skein/sse2/skein.c"
#include "algo/jh/sse2/jh_sse2_opt64.h"
#ifndef NO_AES_NI
#if defined(__AES__)
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/echo/aes_ni/hash_api.h"
#endif
typedef struct {
#ifdef NO_AES_NI
sph_groestl512_context groestl;
sph_echo512_context echo;
#else
#if defined(__AES__)
hashState_groestl groestl;
hashState_echo echo;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
#endif
hashState_luffa luffa;
cubehashParam cubehash;
sph_shavite512_context shavite;
hashState_sd simd;
sph_hamsi512_context hamsi;
// sph_fugue512_context fugue;
} x12_ctx_holder;
x12_ctx_holder x12_ctx;
void init_x12_ctx()
{
#ifdef NO_AES_NI
sph_groestl512_init(&x12_ctx.groestl);
sph_echo512_init(&x12_ctx.echo);
#else
#if defined(__AES__)
init_echo( &x12_ctx.echo, 512 );
init_groestl (&x12_ctx.groestl, 64 );
#else
sph_groestl512_init(&x12_ctx.groestl);
sph_echo512_init(&x12_ctx.echo);
#endif
init_luffa( &x12_ctx.luffa, 512 );
cubehashInit( &x12_ctx.cubehash, 512, 16, 32 );
sph_shavite512_init( &x12_ctx.shavite );
init_sd( &x12_ctx.simd, 512 );
sph_hamsi512_init( &x12_ctx.hamsi );
// sph_fugue512_init( &x13_ctx.fugue );
};
void x12hash(void *output, const void *input)
@@ -108,12 +103,12 @@ void x12hash(void *output, const void *input)
//---groetl----
#ifdef NO_AES_NI
sph_groestl512 (&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#else
#if defined(__AES__)
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512 (&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
//---skein4---
@@ -153,23 +148,18 @@ void x12hash(void *output, const void *input)
//11---echo---
#ifdef NO_AES_NI
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hashB);
#else
#if defined(__AES__)
update_final_echo ( &ctx.echo, (BitSequence *)hashB,
(const BitSequence *)hash, 512 );
#else
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hashB);
#endif
// 12 Hamsi
sph_hamsi512(&ctx.hamsi, hashB, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
/*
// 13 Fugue
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hashB);
*/
asm volatile ("emms");
memcpy(output, hashB, 32);
}

8
algo/x17/x16r-4way.c → algo/x16/x16r-4way.c
View File

@@ -183,16 +183,16 @@ void x16r_4way_hash( void* output, const void* input )
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
break;
case CUBEHASH:
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0,
(const byte*)in0, size );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1,
(const byte*)in1, size );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2,
(const byte*)in2, size );
cubehashReinit( &ctx.cube );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3,
(const byte*)in3, size );
break;

0
algo/x17/x16r-gate.c → algo/x16/x16r-gate.c
View File

0
algo/x17/x16r-gate.h → algo/x16/x16r-gate.h
View File

30
algo/x17/x16r.c → algo/x16/x16r.c
View File

@@ -25,7 +25,7 @@
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include <openssl/sha.h>
#ifndef NO_AES_NI
#if defined(__AES__)
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#endif
@@ -34,12 +34,12 @@ static __thread uint32_t s_ntime = UINT32_MAX;
static __thread char hashOrder[X16R_HASH_FUNC_COUNT + 1] = { 0 };
typedef struct {
#ifdef NO_AES_NI
sph_groestl512_context groestl;
sph_echo512_context echo;
#else
#if defined(__AES__)
hashState_echo echo;
hashState_groestl groestl;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
#endif
sph_blake512_context blake;
sph_bmw512_context bmw;
@@ -95,14 +95,14 @@ void x16r_hash( void* output, const void* input )
sph_bmw512_close(&ctx.bmw, hash);
break;
case GROESTL:
#ifdef NO_AES_NI
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in, size );
sph_groestl512_close(&ctx.groestl, hash);
#else
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)in, size<<3 );
#else
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in, size );
sph_groestl512_close(&ctx.groestl, hash);
#endif
break;
case SKEIN:
@@ -141,14 +141,14 @@ void x16r_hash( void* output, const void* input )
(const BitSequence*)in, size<<3 );
break;
case ECHO:
#ifdef NO_AES_NI
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
sph_echo512_close( &ctx.echo, hash );
#else
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
sph_echo512_close( &ctx.echo, hash );
#endif
break;
case HAMSI:

75
algo/x17/hmq1725.c
View File

@@ -16,10 +16,9 @@
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sph_sha2.h"
#include "algo/haval/sph-haval.h"
#include <openssl/sha.h>
#ifndef NO_AES_NI
#if defined(__AES__)
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/echo/aes_ni/hash_api.h"
#endif
@@ -42,18 +41,14 @@ typedef struct {
sph_fugue512_context fugue1, fugue2;
sph_shabal512_context shabal1;
sph_whirlpool_context whirlpool1, whirlpool2, whirlpool3, whirlpool4;
#ifndef USE_SPH_SHA
SHA512_CTX sha1, sha2;
#else
sph_sha512_context sha1, sha2;
#endif
sph_haval256_5_context haval1, haval2;
#ifdef NO_AES_NI
sph_groestl512_context groestl1, groestl2;
sph_echo512_context echo1, echo2;
#else
#if defined(__AES__)
hashState_echo echo1, echo2;
hashState_groestl groestl1, groestl2;
#else
sph_groestl512_context groestl1, groestl2;
sph_echo512_context echo1, echo2;
#endif
} hmq1725_ctx_holder;
@@ -101,26 +96,22 @@ void init_hmq1725_ctx()
sph_whirlpool_init(&hmq1725_ctx.whirlpool3);
sph_whirlpool_init(&hmq1725_ctx.whirlpool4);
#ifndef USE_SPH_SHA
SHA512_Init( &hmq1725_ctx.sha1 );
SHA512_Init( &hmq1725_ctx.sha2 );
#else
sph_sha512_init(&hmq1725_ctx.sha1);
sph_sha512_init(&hmq1725_ctx.sha2);
#endif
sph_haval256_5_init(&hmq1725_ctx.haval1);
sph_haval256_5_init(&hmq1725_ctx.haval2);
#ifdef NO_AES_NI
sph_groestl512_init( &hmq1725_ctx.groestl1 );
sph_groestl512_init( &hmq1725_ctx.groestl2 );
sph_echo512_init( &hmq1725_ctx.echo1 );
sph_echo512_init( &hmq1725_ctx.echo2 );
#else
#if defined(__AES__)
init_echo( &hmq1725_ctx.echo1, 512 );
init_echo( &hmq1725_ctx.echo2, 512 );
init_groestl( &hmq1725_ctx.groestl1, 64 );
init_groestl( &hmq1725_ctx.groestl2, 64 );
#else
sph_groestl512_init( &hmq1725_ctx.groestl1 );
sph_groestl512_init( &hmq1725_ctx.groestl2 );
sph_echo512_init( &hmq1725_ctx.echo1 );
sph_echo512_init( &hmq1725_ctx.echo2 );
#endif
}
@@ -151,12 +142,12 @@ extern void hmq1725hash(void *state, const void *input)
if ( hashB[0] & mask ) //1
{
#ifdef NO_AES_NI
#if defined(__AES__)
update_and_final_groestl( &h_ctx.groestl1, (char*)hashA,
(const char*)hashB, 512 );
#else
sph_groestl512 (&h_ctx.groestl1, hashB, 64); //1
sph_groestl512_close(&h_ctx.groestl1, hashA); //2
#else
update_and_final_groestl( &h_ctx.groestl1, (char*)hashA,
(const char*)hashB, 512 );
#endif
}
else
@@ -217,12 +208,12 @@ extern void hmq1725hash(void *state, const void *input)
memset(&hashB[8], 0, 32);
}
#ifdef NO_AES_NI
sph_echo512 (&h_ctx.echo1, hashB, 64); //5
sph_echo512_close(&h_ctx.echo1, hashA); //6
#else
#if defined(__AES__)
update_final_echo ( &h_ctx.echo1, (BitSequence *)hashA,
(const BitSequence *)hashB, 512 );
#else
sph_echo512 (&h_ctx.echo1, hashB, 64); //5
sph_echo512_close(&h_ctx.echo1, hashA); //6
#endif
sph_blake512 (&h_ctx.blake2, hashA, 64); //6
@@ -247,12 +238,12 @@ extern void hmq1725hash(void *state, const void *input)
if ( hashA[0] & mask ) //4
{
#ifdef NO_AES_NI
sph_echo512 (&h_ctx.echo2, hashA, 64); //
sph_echo512_close(&h_ctx.echo2, hashB); //5
#else
#if defined(__AES__)
update_final_echo ( &h_ctx.echo2, (BitSequence *)hashB,
(const BitSequence *)hashA, 512 );
#else
sph_echo512 (&h_ctx.echo2, hashA, 64); //
sph_echo512_close(&h_ctx.echo2, hashB); //5
#endif
}
else
@@ -274,30 +265,20 @@ extern void hmq1725hash(void *state, const void *input)
}
else
{
#ifndef USE_SPH_SHA
SHA512_Update( &h_ctx.sha1, hashB, 64 );
SHA512_Final( (unsigned char*) hashA, &h_ctx.sha1 );
#else
sph_sha512 (&h_ctx.sha1, hashB, 64); //7
sph_sha512_close(&h_ctx.sha1, hashA); //8
#endif
}
#ifdef NO_AES_NI
sph_groestl512 (&h_ctx.groestl2, hashA, 64); //3
sph_groestl512_close(&h_ctx.groestl2, hashB); //4
#else
#if defined(__AES__)
update_and_final_groestl( &h_ctx.groestl2, (char*)hashB,
(const char*)hashA, 512 );
#else
sph_groestl512 (&h_ctx.groestl2, hashA, 64); //3
sph_groestl512_close(&h_ctx.groestl2, hashB); //4
#endif
#ifndef USE_SPH_SHA
SHA512_Update( &h_ctx.sha2, hashB, 64 );
SHA512_Final( (unsigned char*) hashA, &h_ctx.sha2 );
#else
sph_sha512 (&h_ctx.sha2, hashB, 64); //2
sph_sha512_close(&h_ctx.sha2, hashA); //3
#endif
if ( hashA[0] & mask ) //4
{

872
algo/x17/sonoa-4way.c Normal file
View File

@@ -0,0 +1,872 @@
#include "sonoa-gate.h"
#if defined(SONOA_4WAY)
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/blake/blake-hash-4way.h"
#include "algo/bmw/bmw-hash-4way.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/skein/skein-hash-4way.h"
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/cube-hash-2way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/shavite/shavite-hash-2way.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/shabal-hash-4way.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/haval/haval-hash-4way.h"
#include "algo/sha/sha2-hash-4way.h"
typedef struct {
blake512_4way_context blake;
bmw512_4way_context bmw;
hashState_groestl groestl;
skein512_4way_context skein;
jh512_4way_context jh;
keccak512_4way_context keccak;
luffa_2way_context luffa;
cube_2way_context cube;
shavite512_2way_context shavite;
simd_2way_context simd;
hashState_echo echo;
hamsi512_4way_context hamsi;
sph_fugue512_context fugue;
shabal512_4way_context shabal;
sph_whirlpool_context whirlpool;
sha512_4way_context sha512;
haval256_5_4way_context haval;
} sonoa_4way_ctx_holder;
sonoa_4way_ctx_holder sonoa_4way_ctx __attribute__ ((aligned (64)));
void init_sonoa_4way_ctx()
{
blake512_4way_init( &sonoa_4way_ctx.blake );
bmw512_4way_init( &sonoa_4way_ctx.bmw );
init_groestl( &sonoa_4way_ctx.groestl, 64 );
skein512_4way_init( &sonoa_4way_ctx.skein );
jh512_4way_init( &sonoa_4way_ctx.jh );
keccak512_4way_init( &sonoa_4way_ctx.keccak );
luffa_2way_init( &sonoa_4way_ctx.luffa, 512 );
cube_2way_init( &sonoa_4way_ctx.cube, 512, 16, 32 );
shavite512_2way_init( &sonoa_4way_ctx.shavite );
simd_2way_init( &sonoa_4way_ctx.simd, 512 );
init_echo( &sonoa_4way_ctx.echo, 512 );
hamsi512_4way_init( &sonoa_4way_ctx.hamsi );
sph_fugue512_init( &sonoa_4way_ctx.fugue );
shabal512_4way_init( &sonoa_4way_ctx.shabal );
sph_whirlpool_init( &sonoa_4way_ctx.whirlpool );
sha512_4way_init( &sonoa_4way_ctx.sha512 );
haval256_5_4way_init( &sonoa_4way_ctx.haval );
};
void sonoa_4way_hash( void *state, const void *input )
{
uint64_t hash0[8] __attribute__ ((aligned (64)));
uint64_t hash1[8] __attribute__ ((aligned (64)));
uint64_t hash2[8] __attribute__ ((aligned (64)));
uint64_t hash3[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*4] __attribute__ ((aligned (64)));
uint64_t vhashA[8*4] __attribute__ ((aligned (64)));
uint64_t vhashB[8*4] __attribute__ ((aligned (64)));
sonoa_4way_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &sonoa_4way_ctx, sizeof(sonoa_4way_ctx) );
// 1
blake512_4way( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
// 2
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
// 3
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
// 4
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm256_reinterleave_4x64( vhashB, vhash, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhashB, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_2x128( vhashA, hash0, hash1, 512 );
mm256_interleave_2x128( vhashB, hash2, hash3, 512 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
// 5
mm256_reinterleave_2x128_4x64( vhash, vhashA, vhashB, 512 );
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_reinterleave_4x32( vhashB, vhash, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhashB, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash1, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash2, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash2 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash3, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
// 6
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash1, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash2, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash2 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash3, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
sha512_4way( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash1, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash2, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash2 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash3, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
// 7
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash1, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash2, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash2 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash3, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
sha512_4way_init( &ctx.sha512 );
sha512_4way( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhash );
mm256_reinterleave_4x32( vhashB, vhash, 512 );
haval256_5_4way( &ctx.haval, vhashB, 64 );
haval256_5_4way_close( &ctx.haval, state );
}
int scanhash_sonoa_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t lane_hash[8];
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
const uint32_t first_nonce = pdata[19];
uint32_t *nonces = work->nonces;
int num_found = 0;
__m256i *noncev = (__m256i*)vdata + 9; // aligned
// uint32_t *noncep = vdata + 73; // 9*8 + 1
const uint32_t Htarg = ptarget[7];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] = { 0, 0xF, 0xFF,
0xFFF, 0xFFFF, 0x10000000 };
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
// Need big endian data
casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
*noncev = mm256_interleave_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0,n+2, 0,n+1, 0, n, 0 ) ),
*noncev );
sonoa_4way_hash( hash, vdata );
for ( int lane = 0; lane < 4; lane++ )
if ( ( ( hash7[ lane ] & mask ) == 0 ) )
{
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

18
algo/x17/sonoa-gate.c Normal file
View File

@@ -0,0 +1,18 @@
#include "sonoa-gate.h"
bool register_sonoa_algo( algo_gate_t* gate )
{
#if defined (SONOA_4WAY)
init_sonoa_4way_ctx();
gate->scanhash = (void*)&scanhash_sonoa_4way;
gate->hash = (void*)&sonoa_4way_hash;
#else
init_sonoa_ctx();
gate->scanhash = (void*)&scanhash_sonoa;
gate->hash = (void*)&sonoa_hash;
#endif
gate->get_max64 = (void*)&get_max64_0x1ffff;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
return true;
};

32
algo/x17/sonoa-gate.h Normal file
View File

@@ -0,0 +1,32 @@
#ifndef SONOA_GATE_H__
#define SONOA_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__AVX2__) && defined(__AES__)
#define SONOA_4WAY
#endif
bool register_sonoa_algo( algo_gate_t* gate );
#if defined(SONOA_4WAY)
void sonoa_4way_hash( void *state, const void *input );
int scanhash_sonoa_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_sonoa_4way_ctx();
#endif
void sonoa_hash( void *state, const void *input );
int scanhash_sonoa( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_sonoa_ctx();
#endif

648
algo/x17/sonoa.c Normal file
View File

@@ -0,0 +1,648 @@
#include "sonoa-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/blake/sph_blake.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/groestl/sph_groestl.h"
#include "algo/jh/sph_jh.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/haval/sph-haval.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"
#include "algo/keccak/sse2/keccak.c"
#include "algo/skein/sse2/skein.c"
#include "algo/jh/sse2/jh_sse2_opt64.h"
#include <openssl/sha.h>
#if defined(__AES__)
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#else
#include "algo/groestl/sph_groestl.h"
#include "algo/echo/sph_echo.h"
#endif
typedef struct {
sph_blake512_context blake;
sph_bmw512_context bmw;
#if defined(__AES__)
hashState_echo echo;
hashState_groestl groestl;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
#endif
sph_jh512_context jh;
sph_keccak512_context keccak;
sph_skein512_context skein;
hashState_luffa luffa;
cubehashParam cubehash;
sph_shavite512_context shavite;
hashState_sd simd;
sph_hamsi512_context hamsi;
sph_fugue512_context fugue;
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
SHA512_CTX sha512;
sph_haval256_5_context haval;
} sonoa_ctx_holder;
sonoa_ctx_holder sonoa_ctx __attribute__ ((aligned (64)));
void init_sonoa_ctx()
{
sph_blake512_init( &sonoa_ctx.blake);
sph_bmw512_init( &sonoa_ctx.bmw);
#if defined(__AES__)
init_echo( &sonoa_ctx.echo, 512 );
init_groestl( &sonoa_ctx.groestl, 64 );
#else
sph_groestl512_init(&sonoa_ctx.groestl );
sph_echo512_init( &sonoa_ctx.echo );
#endif
sph_skein512_init( &sonoa_ctx.skein);
sph_jh512_init( &sonoa_ctx.jh);
sph_keccak512_init( &sonoa_ctx.keccak );
init_luffa( &sonoa_ctx.luffa, 512 );
cubehashInit( &sonoa_ctx.cubehash, 512, 16, 32 );
sph_shavite512_init( &sonoa_ctx.shavite );
init_sd( &sonoa_ctx.simd, 512 );
sph_hamsi512_init( &sonoa_ctx.hamsi );
sph_fugue512_init( &sonoa_ctx.fugue );
sph_shabal512_init( &sonoa_ctx.shabal );
sph_whirlpool_init( &sonoa_ctx.whirlpool );
SHA512_Init( &sonoa_ctx.sha512 );
sph_haval256_5_init(&sonoa_ctx.haval);
};
void sonoa_hash( void *state, const void *input )
{
uint8_t hash[128] __attribute__ ((aligned (64)));
sonoa_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &sonoa_ctx, sizeof(sonoa_ctx) );
sph_blake512(&ctx.blake, input, 80);
sph_blake512_close(&ctx.blake, hash);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
#if defined(__AES__)
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
//
sph_bmw512_init( &ctx.bmw);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512_init(&ctx.groestl );
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512_init( &ctx.skein);
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512_init( &ctx.jh);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512_init( &ctx.keccak );
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashInit( &ctx.cubehash, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
//
sph_bmw512_init( &ctx.bmw);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512_init(&ctx.groestl );
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512_init( &ctx.skein);
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512_init( &ctx.jh);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512_init( &ctx.keccak );
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashInit( &ctx.cubehash, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hash);
//
sph_bmw512_init( &ctx.bmw);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512_init(&ctx.groestl );
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512_init( &ctx.skein);
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512_init( &ctx.jh);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512_init( &ctx.keccak );
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashInit( &ctx.cubehash, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
sph_fugue512_init( &ctx.fugue );
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hash);
sph_shabal512(&ctx.shabal, hash, 64);
sph_shabal512_close(&ctx.shabal, hash);
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
//
sph_bmw512_init( &ctx.bmw);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
sph_shabal512_init( &ctx.shabal );
sph_shabal512(&ctx.shabal, hash, 64);
sph_shabal512_close(&ctx.shabal, hash);
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512_init(&ctx.groestl );
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512_init( &ctx.skein);
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512_init( &ctx.jh);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512_init( &ctx.keccak );
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashInit( &ctx.cubehash, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
sph_fugue512_init( &ctx.fugue );
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hash);
sph_shabal512_init( &ctx.shabal );
sph_shabal512(&ctx.shabal, hash, 64);
sph_shabal512_close(&ctx.shabal, hash);
sph_whirlpool(&ctx.whirlpool, hash, 64);
sph_whirlpool_close(&ctx.whirlpool, hash);
//
sph_bmw512_init( &ctx.bmw);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512_init(&ctx.groestl );
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512_init( &ctx.skein);
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512_init( &ctx.jh);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512_init( &ctx.keccak );
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashInit( &ctx.cubehash, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
sph_fugue512_init( &ctx.fugue );
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hash);
sph_shabal512_init( &ctx.shabal );
sph_shabal512(&ctx.shabal, hash, 64);
sph_shabal512_close(&ctx.shabal, hash);
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool(&ctx.whirlpool, hash, 64);
sph_whirlpool_close(&ctx.whirlpool, hash);
SHA512_Update( &ctx.sha512, hash, 64 );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool(&ctx.whirlpool, hash, 64);
sph_whirlpool_close(&ctx.whirlpool, hash);
//
sph_bmw512_init( &ctx.bmw);
sph_bmw512(&ctx.bmw, hash, 64);
sph_bmw512_close(&ctx.bmw, hash);
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512_init(&ctx.groestl );
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512_init( &ctx.skein);
sph_skein512(&ctx.skein, hash, 64);
sph_skein512_close(&ctx.skein, hash);
sph_jh512_init( &ctx.jh);
sph_jh512(&ctx.jh, hash, 64);
sph_jh512_close(&ctx.jh, hash);
sph_keccak512_init( &ctx.keccak );
sph_keccak512(&ctx.keccak, hash, 64);
sph_keccak512_close(&ctx.keccak, hash);
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
cubehashInit( &ctx.cubehash, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hash, 64 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512(&ctx.shavite, hash, 64);
sph_shavite512_close(&ctx.shavite, hash);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512(&ctx.hamsi, hash, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
sph_fugue512_init( &ctx.fugue );
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hash);
sph_shabal512_init( &ctx.shabal );
sph_shabal512(&ctx.shabal, hash, 64);
sph_shabal512_close(&ctx.shabal, hash);
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool(&ctx.whirlpool, hash, 64);
sph_whirlpool_close(&ctx.whirlpool, hash);
SHA512_Init( &ctx.sha512 );
SHA512_Update( &ctx.sha512, hash, 64 );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
sph_haval256_5(&ctx.haval,(const void*) hash, 64);
sph_haval256_5_close(&ctx.haval, hash);
memcpy(state, hash, 32);
}
int scanhash_sonoa( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t n = pdata[19] - 1;
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] =
{
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] =
{
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
// we need bigendian data...
casti_m128i( endiandata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( endiandata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( endiandata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( endiandata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
#ifdef DEBUG_ALGO
printf("[%d] Htarg=%X\n", thr_id, Htarg);
#endif
for ( int m = 0; m < 6; m++ )
{
if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
pdata[19] = ++n;
be32enc(&endiandata[19], n);
sonoa_hash(hash32, endiandata);
#ifndef DEBUG_ALGO
if ( ( !( hash32[7] & mask ) ) && fulltest( hash32, ptarget ) )
{
work_set_target_ratio( work, hash32 );
*hashes_done = n - first_nonce + 1;
return 1;
}
#else
if (!(n % 0x1000) && !thr_id) printf(".");
if ( !(hash32[7] & mask) )
{
printf("[%d]",thr_id);
if ( fulltest( hash32, ptarget ) )
{
work_set_target_ratio( work, hash32 );
*hashes_done = n - first_nonce + 1;
return 1;
}
}
#endif
} while (n < max_nonce && !work_restart[thr_id].restart);
// see blake.c if else to understand the loop on htmax => mask
break;
}
}
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}

197
algo/x17/x17-4way.c
View File

@@ -15,6 +15,7 @@
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/cube-hash-2way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/shavite/shavite-hash-2way.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
@@ -24,7 +25,9 @@
#include "algo/haval/haval-hash-4way.h"
#include "algo/sha/sha2-hash-4way.h"
typedef struct {
//typedef struct {
union _x17_4way_context_overlay
{
blake512_4way_context blake;
bmw512_4way_context bmw;
hashState_groestl groestl;
@@ -33,7 +36,7 @@ typedef struct {
keccak512_4way_context keccak;
luffa_2way_context luffa;
cube_2way_context cube;
sph_shavite512_context shavite;
shavite512_2way_context shavite;
simd_2way_context simd;
hashState_echo echo;
hamsi512_4way_context hamsi;
@@ -42,8 +45,10 @@ typedef struct {
sph_whirlpool_context whirlpool;
sha512_4way_context sha512;
haval256_5_4way_context haval;
} x17_4way_ctx_holder;
};
typedef union _x17_4way_context_overlay x17_4way_context_overlay;
/*
x17_4way_ctx_holder x17_4way_ctx __attribute__ ((aligned (64)));
void init_x17_4way_ctx()
@@ -56,16 +61,17 @@ void init_x17_4way_ctx()
keccak512_4way_init( &x17_4way_ctx.keccak );
luffa_2way_init( &x17_4way_ctx.luffa, 512 );
cube_2way_init( &x17_4way_ctx.cube, 512, 16, 32 );
sph_shavite512_init( &x17_4way_ctx.shavite );
shavite512_2way_init( &x17_4way_ctx.shavite );
simd_2way_init( &x17_4way_ctx.simd, 512 );
init_echo( &x17_4way_ctx.echo, 512 );
hamsi512_4way_init( &x17_4way_ctx.hamsi );
sph_fugue512_init( &x17_4way_ctx.fugue );
shabal512_4way_init( &x17_4way_ctx.shabal );
sph_whirlpool_init( &x17_4way_ctx.whirlpool );
sha512_4way_init( &x17_4way_ctx.sha512 );
haval256_5_4way_init( &x17_4way_ctx.haval );
};
*/
void x17_4way_hash( void *state, const void *input )
{
uint64_t hash0[8] __attribute__ ((aligned (64)));
@@ -73,155 +79,159 @@ void x17_4way_hash( void *state, const void *input )
uint64_t hash2[8] __attribute__ ((aligned (64)));
uint64_t hash3[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*4] __attribute__ ((aligned (64)));
uint64_t vhashA[8*4] __attribute__ ((aligned (64)));
uint64_t vhashB[8*4] __attribute__ ((aligned (64)));
x17_4way_ctx_holder ctx;
memcpy( &ctx, &x17_4way_ctx, sizeof(x17_4way_ctx) );
x17_4way_context_overlay ctx;
// memcpy( &ctx, &x17_4way_ctx, sizeof(x17_4way_ctx) );
// 1 Blake 4 way 64 bit
// 1 Blake parallel 4 way 64 bit
blake512_4way_init( &ctx.blake );
blake512_4way( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
// 2 Bmw
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
// Serial
// Serialize
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 3 Groestl
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
memcpy( &ctx.groestl, &x17_4way_ctx.groestl, sizeof(hashState_groestl) );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
memcpy( &ctx.groestl, &x17_4way_ctx.groestl, sizeof(hashState_groestl) );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
memcpy( &ctx.groestl, &x17_4way_ctx.groestl, sizeof(hashState_groestl) );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
// Parallel 4way
// Parallellize
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// 4 Skein
// 4 Skein parallel 4 way 64 bit
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
// 5 JH
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
// 6 Keccak
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 7 Luffa parallel 2 way 128 bit
mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
// 7 Luffa parallel 2 way
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
mm256_interleave_2x128( vhashB, hash2, hash3, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
// 8 Cubehash parallel 2 way
cube_2way_update_close( &ctx.cube, vhash, vhash, 64 );
cube_2way_reinit( &ctx.cube );
// 8 Cubehash
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
// 9 Shavite
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
// 10 Simd
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
// 9 Shavite serial
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
memcpy( &ctx.shavite, &x17_4way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash1, 64 );
sph_shavite512_close( &ctx.shavite, hash1 );
memcpy( &ctx.shavite, &x17_4way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash2, 64 );
sph_shavite512_close( &ctx.shavite, hash2 );
memcpy( &ctx.shavite, &x17_4way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
// 10 Simd parallel 2 way 128 bit
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_interleave_2x128( vhash, hash2, hash3, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
// 11 Echo serial
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
memcpy( &ctx.echo, &x17_4way_ctx.echo, sizeof(hashState_echo) );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
memcpy( &ctx.echo, &x17_4way_ctx.echo, sizeof(hashState_echo) );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
memcpy( &ctx.echo, &x17_4way_ctx.echo, sizeof(hashState_echo) );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
// 12 Hamsi parallel 4 way 64 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 13 Fugue serial
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx.fugue, &x17_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
memcpy( &ctx.fugue, &x17_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
memcpy( &ctx.fugue, &x17_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
// 14 Shabal, parallel 4 way 32 bit SSE
// 14 Shabal, parallel 4 way 32 bit
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
// 15 Whirlpool
// 15 Whirlpool serial
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
memcpy( &ctx.whirlpool, &x17_4way_ctx.whirlpool,
sizeof(sph_whirlpool_context) );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash1, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
memcpy( &ctx.whirlpool, &x17_4way_ctx.whirlpool,
sizeof(sph_whirlpool_context) );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash2, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash2 );
memcpy( &ctx.whirlpool, &x17_4way_ctx.whirlpool,
sizeof(sph_whirlpool_context) );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash3, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
// 16 SHA512 parallel 64 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
sha512_4way_init( &ctx.sha512 );
sha512_4way( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhash );
// 17 Haval parallel 32 bit
mm256_reinterleave_4x32( vhashB, vhash, 512 );
haval256_5_4way_init( &ctx.haval );
haval256_5_4way( &ctx.haval, vhashB, 64 );
haval256_5_4way_close( &ctx.haval, state );
}
int scanhash_x17_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
@@ -234,49 +244,48 @@ int scanhash_x17_4way( int thr_id, struct work *work, uint32_t max_nonce,
const uint32_t first_nonce = pdata[19];
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 73; // 9*8 + 1
__m256i *noncev = (__m256i*)vdata + 9; // aligned
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = { 0, 0xF, 0xFF,
0xFFF, 0xFFFF, 0x10000000 };
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
// big endian encode 0..18 uint32_t, 64 bits at a time
swab32_array( endiandata, pdata, 20 );
// Need big endian data
casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
be32enc( noncep, n );
be32enc( noncep+2, n+1 );
be32enc( noncep+4, n+2 );
be32enc( noncep+6, n+3 );
for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
*noncev = mm256_interleave_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0,n+2, 0,n+1, 0, n, 0 ) ),
*noncev );
x17_4way_hash( hash, vdata );
x17_4way_hash( hash, vdata );
for ( int lane = 0; lane < 4; lane++ )
if ( ( ( hash7[ lane ] & mask ) == 0 ) )
{
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )
for ( int lane = 0; lane < 4; lane++ )
if ( ( ( hash7[ lane ] & mask ) == 0 ) )
{
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
break;
}
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;

4
algo/x17/x17-gate.c
View File

@@ -3,10 +3,12 @@
bool register_x17_algo( algo_gate_t* gate )
{
#if defined (X17_4WAY)
init_x17_4way_ctx();
printf("register x17 4way\n");
// init_x17_4way_ctx();
gate->scanhash = (void*)&scanhash_x17_4way;
gate->hash = (void*)&x17_4way_hash;
#else
printf("register x17 no 4way\n");
init_x17_ctx();
gate->scanhash = (void*)&scanhash_x17;
gate->hash = (void*)&x17_hash;

6
algo/x17/x17-gate.h
View File

@@ -15,16 +15,16 @@ bool register_x17_algo( algo_gate_t* gate );
void x17_4way_hash( void *state, const void *input );
int scanhash_x17_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_x17_4way_ctx();
//void init_x17_4way_ctx();
#endif
void x17_hash( void *state, const void *input );
int scanhash_x17( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_x17_ctx();

233
algo/x17/x17.c
View File

@@ -5,7 +5,6 @@
#include <stdio.h>
#include "algo/blake/sph_blake.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/groestl/sph_groestl.h"
#include "algo/jh/sph_jh.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
@@ -13,14 +12,11 @@
#include "algo/luffa/sph_luffa.h"
#include "algo/cubehash/sph_cubehash.h"
#include "algo/simd/sph_simd.h"
#include "algo/echo/sph_echo.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sph_sha2.h"
#include "algo/haval/sph-haval.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
@@ -30,18 +26,21 @@
#include "algo/skein/sse2/skein.c"
#include "algo/jh/sse2/jh_sse2_opt64.h"
#include <openssl/sha.h>
#ifndef NO_AES_NI
#if defined(__AES__)
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#else
#include "algo/groestl/sph_groestl.h"
#include "algo/echo/sph_echo.h"
#endif
typedef struct {
#ifdef NO_AES_NI
sph_groestl512_context groestl;
sph_echo512_context echo;
#else
hashState_echo echo;
#if defined(__AES__)
hashState_groestl groestl;
hashState_echo echo;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
#endif
hashState_luffa luffa;
cubehashParam cubehash;
@@ -51,11 +50,7 @@ typedef struct {
sph_fugue512_context fugue;
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
#ifndef USE_SPH_SHA
SHA512_CTX sha512;
#else
sph_sha512_context sha512;
#endif
sph_haval256_5_context haval;
} x17_ctx_holder;
@@ -63,12 +58,12 @@ x17_ctx_holder x17_ctx __attribute__ ((aligned (64)));
void init_x17_ctx()
{
#ifdef NO_AES_NI
#if defined(__AES__)
init_groestl( &x17_ctx.groestl, 64 );
init_echo( &x17_ctx.echo, 512 );
#else
sph_groestl512_init(&x17_ctx.groestl );
sph_echo512_init(&x17_ctx.echo);
#else
init_echo( &x17_ctx.echo, 512 );
init_groestl( &x17_ctx.groestl, 64 );
#endif
init_luffa( &x17_ctx.luffa, 512 );
cubehashInit( &x17_ctx.cubehash, 512, 16, 32 );
@@ -78,11 +73,7 @@ void init_x17_ctx()
sph_fugue512_init( &x17_ctx.fugue );
sph_shabal512_init( &x17_ctx.shabal );
sph_whirlpool_init( &x17_ctx.whirlpool );
#ifndef USE_SPH_SHA
SHA512_Init( &x17_ctx.sha512 );
#else
sph_sha512_init(&x17_ctx.sha512);
#endif
sph_haval256_5_init(&x17_ctx.haval);
};
@@ -123,12 +114,12 @@ void x17_hash(void *output, const void *input)
//---groestl----
#ifdef NO_AES_NI
sph_groestl512(&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#else
#if defined(__AES__)
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#else
sph_groestl512( &ctx.groestl, hash, 64 );
sph_groestl512_close( &ctx.groestl, hash );
#endif
//---skein4---
@@ -151,134 +142,136 @@ void x17_hash(void *output, const void *input)
KEC_C;
//--- luffa7
update_and_final_luffa( &ctx.luffa, (BitSequence*)hashB,
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)hash, 64 );
// 8 Cube
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hashB, 64 );
(const byte*)hash, 64 );
// 9 Shavite
sph_shavite512( &ctx.shavite, hash, 64);
sph_shavite512_close( &ctx.shavite, hashB);
sph_shavite512_close( &ctx.shavite, hash);
// 10 Simd
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hashB, 512 );
update_final_sd( &ctx.simd, (BitSequence*)hash,
(const BitSequence*)hash, 512 );
//11---echo---
#ifdef NO_AES_NI
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hashB);
#if defined(__AES__)
update_final_echo ( &ctx.echo, (BitSequence*)hash,
(const BitSequence*)hash, 512 );
#else
update_final_echo ( &ctx.echo, (BitSequence *)hashB,
(const BitSequence *)hash, 512 );
sph_echo512( &ctx.echo, hash, 64 );
sph_echo512_close( &ctx.echo, hash );
#endif
// X13 algos
// 12 Hamsi
sph_hamsi512(&ctx.hamsi, hashB, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
sph_hamsi512( &ctx.hamsi, hash, 64 );
sph_hamsi512_close( &ctx.hamsi, hash );
// 13 Fugue
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hashB);
sph_fugue512(&ctx.fugue, hash, 64 );
sph_fugue512_close(&ctx.fugue, hash );
// X14 Shabal
sph_shabal512(&ctx.shabal, hashB, 64);
sph_shabal512_close(&ctx.shabal, hash);
sph_shabal512(&ctx.shabal, hash, 64);
sph_shabal512_close( &ctx.shabal, hash );
// X15 Whirlpool
sph_whirlpool(&ctx.whirlpool, hash, 64);
sph_whirlpool_close(&ctx.whirlpool, hashB);
sph_whirlpool( &ctx.whirlpool, hash, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash );
#ifndef USE_SPH_SHA
SHA512_Update( &ctx.sha512, hashB, 64 );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
#else
sph_sha512(&ctx.sha512,(const void*) hashB, 64);
sph_sha512_close(&ctx.sha512,(void*) hash);
#endif
sph_haval256_5(&ctx.haval,(const void*) hash, 64);
sph_haval256_5_close(&ctx.haval,hashB);
SHA512_Update( &ctx.sha512, hash, 64 );
SHA512_Final( (unsigned char*)hash, &ctx.sha512 );
asm volatile ("emms");
memcpy(output, hashB, 32);
sph_haval256_5( &ctx.haval, (const void*)hash, 64 );
sph_haval256_5_close( &ctx.haval, output );
}
int scanhash_x17(int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done)
int scanhash_x17( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t hash64[8] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t hash64[8] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] = {
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] = {
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
uint64_t htmax[] =
{
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] =
{
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
// we need bigendian data...
swab32_array( endiandata, pdata, 20 );
// we need bigendian data...
casti_m128i( endiandata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( endiandata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( endiandata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( endiandata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
#ifdef DEBUG_ALGO
if (Htarg != 0)
printf("[%d] Htarg=%X\n", thr_id, Htarg);
if ( Htarg != 0 )
printf( "[%d] Htarg=%X\n", thr_id, Htarg );
#endif
for (int m=0; m < 6; m++) {
if (Htarg <= htmax[m]) {
uint32_t mask = masks[m];
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
x17_hash(hash64, endiandata);
for ( int m = 0; m < 6; m++ )
{
if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
pdata[19] = ++n;
be32enc( &endiandata[19], n );
x17_hash( hash64, endiandata );
#ifndef DEBUG_ALGO
if (!(hash64[7] & mask))
{
if ( fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
// else
// {
// applog(LOG_INFO, "Result does not validate on CPU!");
// }
}
if ( !( hash64[7] & mask ) )
{
if ( fulltest( hash64, ptarget ) )
{
*hashes_done = n - first_nonce + 1;
return true;
}
// else
// applog(LOG_INFO, "Result does not validate on CPU!");
}
#else
if (!(n % 0x1000) && !thr_id) printf(".");
if (!(hash64[7] & mask)) {
printf("[%d]",thr_id);
if (fulltest(hash64, ptarget)) {
work_set_target_ratio( work, hash64 );
*hashes_done = n - first_nonce + 1;
return true;
}
}
#endif
} while (n < max_nonce && !work_restart[thr_id].restart);
// see blake.c if else to understand the loop on htmax => mask
break;
if ( !( n % 0x1000 ) && !thr_id ) printf(".");
if ( !( hash64[7] & mask ) )
{
printf("[%d]",thr_id);
if ( fulltest( hash64, ptarget ) )
{
work_set_target_ratio( work, hash64 );
*hashes_done = n - first_nonce + 1;
return true;
}
}
#endif
} while (n < max_nonce && !work_restart[thr_id].restart);
// see blake.c if else to understand the loop on htmax => mask
break;
}
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}

77
algo/x17/xevan.c
View File

@@ -16,17 +16,16 @@
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sph_sha2.h"
#include "algo/haval/sph-haval.h"
#include "algo/simd/nist.h"
#include "algo/cubehash/cubehash_sse2.h"
#include <openssl/sha.h>
#ifdef NO_AES_NI
#include "algo/groestl/sph_groestl.h"
#include "algo/echo/sph_echo.h"
#else
#if defined(__AES__)
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/echo/aes_ni/hash_api.h"
#else
#include "algo/groestl/sph_groestl.h"
#include "algo/echo/sph_echo.h"
#endif
typedef struct {
@@ -43,18 +42,14 @@ typedef struct {
sph_fugue512_context fugue;
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
#ifndef USE_SPH_SHA
SHA512_CTX sha512;
#else
sph_sha512_context sha512;
#endif
sph_haval256_5_context haval;
#ifdef NO_AES_NI
sph_groestl512_context groestl;
sph_echo512_context echo;
#else
#if defined(__AES__)
hashState_echo echo;
hashState_groestl groestl;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
#endif
} xevan_ctx_holder;
@@ -77,18 +72,14 @@ void init_xevan_ctx()
sph_fugue512_init( &xevan_ctx.fugue );
sph_shabal512_init( &xevan_ctx.shabal );
sph_whirlpool_init( &xevan_ctx.whirlpool );
#ifndef USE_SPH_SHA
SHA512_Init( &xevan_ctx.sha512 );
#else
sph_sha512_init(&xevan_ctx.sha512);
#endif
sph_haval256_5_init(&xevan_ctx.haval);
#ifdef NO_AES_NI
sph_groestl512_init( &xevan_ctx.groestl );
sph_echo512_init( &xevan_ctx.echo );
#else
#if defined(__AES__)
init_groestl( &xevan_ctx.groestl, 64 );
init_echo( &xevan_ctx.echo, 512 );
#else
sph_groestl512_init( &xevan_ctx.groestl );
sph_echo512_init( &xevan_ctx.echo );
#endif
};
@@ -117,12 +108,12 @@ void xevan_hash(void *output, const void *input)
sph_bmw512(&ctx.bmw, hash, dataLen);
sph_bmw512_close(&ctx.bmw, hash);
#ifdef NO_AES_NI
#if defined(__AES__)
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, dataLen*8 );
#else
sph_groestl512(&ctx.groestl, hash, dataLen);
sph_groestl512_close(&ctx.groestl, hash);
#else
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, dataLen*8 );
#endif
sph_skein512(&ctx.skein, hash, dataLen);
@@ -146,12 +137,12 @@ void xevan_hash(void *output, const void *input)
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, dataLen*8 );
#ifdef NO_AES_NI
#if defined(__AES__)
update_final_echo( &ctx.echo, (BitSequence *) hash,
(const BitSequence *) hash, dataLen*8 );
#else
sph_echo512(&ctx.echo, hash, dataLen);
sph_echo512_close(&ctx.echo, hash);
#else
update_final_echo( &ctx.echo, (BitSequence *) hash,
(const BitSequence *) hash, dataLen*8 );
#endif
sph_hamsi512(&ctx.hamsi, hash, dataLen);
@@ -166,13 +157,9 @@ void xevan_hash(void *output, const void *input)
sph_whirlpool(&ctx.whirlpool, hash, dataLen);
sph_whirlpool_close(&ctx.whirlpool, hash);
#ifndef USE_SPH_SHA
SHA512_Update( &ctx.sha512, hash, dataLen );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
#else
sph_sha512(&ctx.sha512,(const void*) hash, dataLen);
sph_sha512_close(&ctx.sha512,(void*) hash);
#endif
sph_haval256_5(&ctx.haval,(const void*) hash, dataLen);
sph_haval256_5_close(&ctx.haval, hash);
@@ -186,12 +173,12 @@ void xevan_hash(void *output, const void *input)
sph_bmw512(&ctx.bmw, hash, dataLen);
sph_bmw512_close(&ctx.bmw, hash);
#ifdef NO_AES_NI
sph_groestl512(&ctx.groestl, hash, dataLen);
sph_groestl512_close(&ctx.groestl, hash);
#else
#if defined(__AES__)
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const BitSequence*)hash, dataLen*8 );
#else
sph_groestl512(&ctx.groestl, hash, dataLen);
sph_groestl512_close(&ctx.groestl, hash);
#endif
sph_skein512(&ctx.skein, hash, dataLen);
@@ -214,12 +201,12 @@ void xevan_hash(void *output, const void *input)
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hash, dataLen*8 );
#ifdef NO_AES_NI
sph_echo512(&ctx.echo, hash, dataLen);
sph_echo512_close(&ctx.echo, hash);
#else
#if defined(__AES__)
update_final_echo( &ctx.echo, (BitSequence *) hash,
(const BitSequence *) hash, dataLen*8 );
#else
sph_echo512(&ctx.echo, hash, dataLen);
sph_echo512_close(&ctx.echo, hash);
#endif
sph_hamsi512(&ctx.hamsi, hash, dataLen);
@@ -234,13 +221,9 @@ void xevan_hash(void *output, const void *input)
sph_whirlpool(&ctx.whirlpool, hash, dataLen);
sph_whirlpool_close(&ctx.whirlpool, hash);
#ifndef USE_SPH_SHA
SHA512_Update( &ctx.sha512, hash, dataLen );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
#else
sph_sha512(&ctx.sha512,(const void*) hash, dataLen);
sph_sha512_close(&ctx.sha512,(void*) hash);
#endif
sph_haval256_5(&ctx.haval,(const void*) hash, dataLen);
sph_haval256_5_close(&ctx.haval, hash);

34
algo/x20/x20r-gate.c Normal file
View File

@@ -0,0 +1,34 @@
#include "x20r-gate.h"
void getAlgoString( const uint8_t* prevblock, char *output )
{
char *sptr = outpuit;
for ( int j = 0; j < X20R_HASH_FUNC_COUNT; j++ )
{
char b = (19 - j) >> 1; // 16 ascii hex chars, reversed
uint8_t algoDigit = (j & 1) ? prevblock[b] & 0xF : prevblock[b] >> 4;
if (algoDigit >= 10)
sprintf(sptr, "%c", 'A' + (algoDigit - 10));
else
sprintf(sptr, "%u", (uint32_t) algoDigit);
sptr++;
}
*sptr = '\0';
}
bool register_x20r_algo( algo_gate_t* gate )
{
#if defined (X20R_4WAY)
gate->scanhash = (void*)&scanhash_x20r_4way;
gate->hash = (void*)&x20r_4way_hash;
#else
gate->scanhash = (void*)&scanhash_x20r;
gate->hash = (void*)&x20r_hash;
#endif
gate->set_target = (void*)&alt_set_target;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
x20_r_s_getAlgoString = (void*)&x20r_getAlgoString;
return true;
};

58
algo/x20/x20r-gate.h Normal file
View File

@@ -0,0 +1,58 @@
#ifndef X20R_GATE_H__
#define X20R_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
/*
#if defined(__AVX2__) && defined(__AES__)
#define X20R_4WAY
#endif
*/
enum x20r_Algo {
BLAKE = 0,
BMW,
GROESTL,
JH,
KECCAK,
SKEIN,
LUFFA,
CUBEHASH,
SHAVITE,
SIMD,
ECHO,
HAMSI,
FUGUE,
SHABAL,
WHIRLPOOL,
SHA_512,
HAVAL, // 256-bits output
GOST,
RADIOGATUN, // 256-bits output
PANAMA, // 256-bits output
X20R_HASH_FUNC_COUNT
};
void (*x20_r_s_getAlgoString) ( const uint8_t*, char* );
void x20r_getAlgoString( const uint8_t* prevblock, char *output );
bool register_xi20r_algo( algo_gate_t* gate );
#if defined(X20R_4WAY)
void x20r_4way_hash( void *state, const void *input );
int scanhash_x20r_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#endif
void x20rhash( void *state, const void *input );
int scanhash_x20r( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#endif

275
algo/x20/x20r.c Normal file
View File

@@ -0,0 +1,275 @@
#include "x20r-gate.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "algo/blake/sph_blake.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/jh/sph_jh.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/haval/sph-haval.h"
#include "algo/radiogatun/sph_radiogatun.h"
#include "algo/panama/sph_panama.h"
#include "algo/gost/sph_gost.h"
#include <openssl/sha.h>
#if defined(__AES__)
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#else
#include "algo/groestl/sph_groestl.h"
#include "algo/echo/sph_echo.h"
#endif
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
static __thread uint32_t s_ntime = UINT32_MAX;
static __thread char hashOrder[X20R_HASH_FUNC_COUNT + 1] = { 0 };
union _x20r_context_overlay
{
sph_blake512_context blake;
sph_bmw512_context bmw;
#if defined(__AES__)
hashState_groestl groestl;
hashState_echo echo;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
#endif
sph_skein512_context skein;
sph_jh512_context jh;
sph_keccak512_context keccak;
hashState_luffa luffa;
cubehashParam cube;
hashState_sd simd;
sph_shavite512_context shavite;
sph_hamsi512_context hamsi;
sph_fugue512_context fugue;
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
SHA512_CTX sha512;
sph_haval256_5_context haval;
sph_gost512_context gost;
sph_radiogatun64_context radiogatun;
sph_panama_context panama;
};
typedef union _x20r_context_overlay x20r_context_overlay;
void x20r_hash(void* output, const void* input)
{
uint32_t _ALIGN(128) hash[64/4];
x20r_context_overlay ctx;
/*
sph_blake512_context ctx_blake;
sph_bmw512_context ctx_bmw;
sph_groestl512_context ctx_groestl;
sph_skein512_context ctx_skein;
sph_jh512_context ctx_jh;
sph_keccak512_context ctx_keccak;
sph_luffa512_context ctx_luffa;
sph_cubehash512_context ctx_cubehash;
sph_shavite512_context ctx_shavite;
sph_simd512_context ctx_simd;
sph_echo512_context ctx_echo;
sph_hamsi512_context ctx_hamsi;
sph_fugue512_context ctx_fugue;
sph_shabal512_context ctx_shabal;
sph_whirlpool_context ctx_whirlpool;
sph_sha512_context ctx_sha512;
sph_haval256_5_context ctx_haval;
sph_gost512_context ctx_gost;
sph_radiogatun64_context ctx_radiogatun;
sph_panama_context ctx_panama;
*/
void *in = (void*) input;
int size = 80;
if ( s_ntime == UINT32_MAX )
{
const uint8_t* in8 = (uint8_t*) input;
x20_r_s_getAlgoString(&in8[4], hashOrder);
}
for (int i = 0; i < 20; i++)
{
const char elem = hashOrder[i];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case BLAKE:
sph_blake512_init(&ctx.blake);
sph_blake512(&ctx.blake, in, size);
sph_blake512_close(&ctx.blake, hash);
break;
case BMW:
sph_bmw512_init(&ctx.bmw);
sph_bmw512(&ctx.bmw, in, size);
sph_bmw512_close(&ctx.bmw, hash);
break;
case GROESTL:
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)in, size<<3 );
#else
sph_groestl512_init(&ctx.groestl);
sph_groestl512(&ctx.groestl, in, size);
sph_groestl512_close(&ctx.groestl, hash);
#endif
break;
case SKEIN:
sph_skein512_init(&ctx.skein);
sph_skein512(&ctx.skein, in, size);
sph_skein512_close(&ctx.skein, hash);
break;
case JH:
sph_jh512_init(&ctx.jh);
sph_jh512(&ctx.jh, in, size);
sph_jh512_close(&ctx.jh, hash);
break;
case KECCAK:
sph_keccak512_init(&ctx.keccak);
sph_keccak512(&ctx.keccak, in, size);
sph_keccak512_close(&ctx.keccak, hash);
break;
case LUFFA:
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)in, size );
break;
case CUBEHASH:
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash,
(const byte*)in, size );
break;
case SHAVITE:
sph_shavite512_init(&ctx.shavite);
sph_shavite512(&ctx.shavite, in, size);
sph_shavite512_close(&ctx.shavite, hash);
break;
case SIMD:
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)in, size<<3 );
break;
case ECHO:
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence *)in, size<<3 );
#else
sph_echo512_init(&ctx.echo);
sph_echo512(&ctx.echo, in, size);
sph_echo512_close(&ctx.echo, hash);
#endif
break;
case HAMSI:
sph_hamsi512_init(&ctx.hamsi);
sph_hamsi512(&ctx.hamsi, in, size);
sph_hamsi512_close(&ctx.hamsi, hash);
break;
case FUGUE:
sph_fugue512_init(&ctx.fugue);
sph_fugue512(&ctx.fugue, in, size);
sph_fugue512_close(&ctx.fugue, hash);
break;
case SHABAL:
sph_shabal512_init(&ctx.shabal);
sph_shabal512(&ctx.shabal, in, size);
sph_shabal512_close(&ctx.shabal, hash);
break;
case WHIRLPOOL:
sph_whirlpool_init(&ctx.whirlpool);
sph_whirlpool(&ctx.whirlpool, in, size);
sph_whirlpool_close(&ctx.whirlpool, hash);
break;
case SHA_512:
SHA512_Init( &ctx.sha512 );
SHA512_Update( &ctx.sha512, in, size );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
break;
case HAVAL:
sph_haval256_5_init(&ctx.haval);
sph_haval256_5(&ctx.haval, in, size);
sph_haval256_5_close(&ctx.haval, hash);
memset(&hash[8], 0, 32);
break;
case GOST:
sph_gost512_init(&ctx.gost);
sph_gost512(&ctx.gost, in, size);
sph_gost512_close(&ctx.gost, hash);
break;
case RADIOGATUN:
sph_radiogatun64_init(&ctx.radiogatun);
sph_radiogatun64(&ctx.radiogatun, in, size);
sph_radiogatun64_close(&ctx.radiogatun, hash);
memset(&hash[8], 0, 32);
break;
case PANAMA:
sph_panama_init(&ctx.panama);
sph_panama(&ctx.panama, in, size);
sph_panama_close(&ctx.panama, hash);
memset(&hash[8], 0, 32);
break;
}
in = (void*) hash;
size = 64;
}
memcpy(output, hash, 32);
}
int scanhash_x20r( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t nonce = first_nonce;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
for (int k=0; k < 19; k++)
be32enc( &endiandata[k], pdata[k] );
if ( s_ntime != pdata[17] )
{
uint32_t ntime = swab32(pdata[17]);
x20_r_s_getAlgoString( (const char*) (&endiandata[1]), hashOrder );
s_ntime = ntime;
if (opt_debug && !thr_id) applog(LOG_DEBUG, "hash order %s (%08x)", hashOrder, ntime);
}
if ( opt_benchmark )
ptarget[7] = 0x0cff;
do {
be32enc( &endiandata[19], nonce );
x20r_hash( hash32, endiandata );
if ( hash32[7] <= Htarg && fulltest( hash32, ptarget ) )
{
work_set_target_ratio( work, hash32 );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce;
return 1;
}
nonce++;
} while (nonce < max_nonce && !(*restart));
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}

37
algo/yescrypt/sha256_Y.c
View File

@@ -299,48 +299,26 @@ HMAC_SHA256_Init(HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen)
/* If Klen > 64, the key is really SHA256(K). */
if (Klen > 64) {
#ifndef USE_SPH_SHA
SHA256_Init(&ctx->ictx);
SHA256_Update(&ctx->ictx, K, Klen);
SHA256_Final(khash, &ctx->ictx);
#else
SHA256_Init_Y(&ctx->ictx);
SHA256_Update_Y(&ctx->ictx, K, Klen);
SHA256_Final_Y(khash, &ctx->ictx);
#endif
K = khash;
Klen = 32;
}
/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
#ifndef USE_SPH_SHA
SHA256_Init(&ctx->ictx);
#else
SHA256_Init_Y(&ctx->ictx);
#endif
memset(pad, 0x36, 64);
for (i = 0; i < Klen; i++)
pad[i] ^= K[i];
#ifndef USE_SPH_SHA
SHA256_Update(&ctx->ictx, pad, 64);
#else
SHA256_Update_Y(&ctx->ictx, pad, 64);
#endif
/* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
#ifndef USE_SPH_SHA
SHA256_Init(&ctx->octx);
#else
SHA256_Init_Y(&ctx->octx);
#endif
memset(pad, 0x5c, 64);
for (i = 0; i < Klen; i++)
pad[i] ^= K[i];
#ifndef USE_SPH_SHA
SHA256_Update(&ctx->octx, pad, 64);
#else
SHA256_Update_Y(&ctx->octx, pad, 64);
#endif
/* Clean the stack. */
//memset(khash, 0, 32);
@@ -352,11 +330,7 @@ HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void *in, size_t len)
{
/* Feed data to the inner SHA256 operation. */
#ifndef USE_SPH_SHA
SHA256_Update(&ctx->ictx, in, len);
#else
SHA256_Update_Y(&ctx->ictx, in, len);
#endif
}
/* Finish an HMAC-SHA256 operation. */
@@ -365,7 +339,6 @@ HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx)
{
unsigned char ihash[32];
#ifndef USE_SPH_SHA
/* Finish the inner SHA256 operation. */
SHA256_Final(ihash, &ctx->ictx);
@@ -374,16 +347,6 @@ HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx)
/* Finish the outer SHA256 operation. */
SHA256_Final(digest, &ctx->octx);
#else
/* Finish the inner SHA256 operation. */
SHA256_Final_Y(ihash, &ctx->ictx);
/* Feed the inner hash to the outer SHA256 operation. */
SHA256_Update_Y(&ctx->octx, ihash, 32);
/* Finish the outer SHA256 operation. */
SHA256_Final_Y(digest, &ctx->octx);
#endif
/* Clean the stack. */
//memset(ihash, 0, 32);

5
algo/yescrypt/sha256_Y.h
View File

@@ -47,13 +47,8 @@ typedef struct HMAC_SHA256Context {
*/
typedef struct HMAC_SHA256Context {
#ifndef USE_SPH_SHA
SHA256_CTX ictx;
SHA256_CTX octx;
#else
SHA256_CTX_Y ictx;
SHA256_CTX_Y octx;
#endif
} HMAC_SHA256_CTX;
void SHA256_Init_Y(SHA256_CTX_Y *);

14
algo/yescrypt/yescrypt-simd.c
View File

@@ -1303,17 +1303,10 @@ yescrypt_kdf(const yescrypt_shared_t * shared, yescrypt_local_t * local,
S = (uint8_t *)XY + XY_size;
if (t || flags) {
#ifndef USE_SPH_SHA
SHA256_CTX ctx;
SHA256_Init(&ctx);
SHA256_Update(&ctx, passwd, passwdlen);
SHA256_Final(sha256, &ctx);
#else
SHA256_CTX_Y ctx;
SHA256_Init_Y(&ctx);
SHA256_Update_Y(&ctx, passwd, passwdlen);
SHA256_Final_Y(sha256, &ctx);
#endif
passwd = sha256;
passwdlen = sizeof(sha256);
}
@@ -1372,17 +1365,10 @@ yescrypt_kdf(const yescrypt_shared_t * shared, yescrypt_local_t * local,
}
/* Compute StoredKey */
{
#ifndef USE_SPH_SHA
SHA256_CTX ctx;
SHA256_Init(&ctx);
SHA256_Update(&ctx, sha256, sizeof(sha256));
SHA256_Final(buf, &ctx);
#else
SHA256_CTX_Y ctx;
SHA256_Init_Y(&ctx);
SHA256_Update_Y(&ctx, sha256, sizeof(sha256));
SHA256_Final_Y(buf, &ctx);
#endif
}
}

3
algo/yespower/yespower-opt.c
View File

@@ -49,6 +49,7 @@
* no slowdown from the prefixes is generally observed on AMD CPUs supporting
* XOP, some slowdown is sometimes observed on Intel CPUs with AVX.
*/
/*
#ifdef __XOP__
#warning "Note: XOP is enabled. That's great."
#elif defined(__AVX__)
@@ -60,7 +61,7 @@
#else
#warning "Note: building generic code for non-x86. That's OK."
#endif
*/
/*
* The SSE4 code version has fewer instructions than the generic SSE2 version,
* but all of the instructions are SIMD, thereby wasting the scalar execution

30
api.c
View File

@@ -96,9 +96,9 @@ extern char *opt_api_allow;
extern int opt_api_listen; /* port */
extern int opt_api_remote;
extern double global_hashrate;
extern uint32_t accepted_count;
extern uint32_t rejected_count;
extern uint32_t solved_count;
//extern uint32_t accepted_count;
//extern uint32_t rejected_count;
//extern uint32_t solved_count;
#define cpu_threads opt_n_threads
@@ -136,7 +136,7 @@ static char *getsummary( char *params )
char algo[64]; *algo = '\0';
time_t ts = time(NULL);
double uptime = difftime(ts, startup);
double accps = (60.0 * accepted_count) / (uptime ? uptime : 1.0);
double accps = (60.0 * accepted_share_count) / (uptime ? uptime : 1.0);
double diff = net_diff > 0. ? net_diff : stratum_diff;
char diff_str[16];
double hrate = (double)global_hashrate;
@@ -157,16 +157,18 @@ static char *getsummary( char *params )
sprintf( diff_str, "%.6f", diff);
*buffer = '\0';
sprintf( buffer, "NAME=%s;VER=%s;API=%s;"
"ALGO=%s;CPUS=%d;URL=%s;"
"HS=%.2f;KHS=%.2f;ACC=%d;REJ=%d;SOL=%d;"
"ACCMN=%.3f;DIFF=%s;TEMP=%.1f;FAN=%d;FREQ=%d;"
"UPTIME=%.0f;TS=%u|",
PACKAGE_NAME, PACKAGE_VERSION, APIVERSION,
algo, opt_n_threads, short_url, hrate, hrate/1000.0,
accepted_count, rejected_count, solved_count,
accps, diff_str, cpu.cpu_temp, cpu.cpu_fan, cpu.cpu_clock,
uptime, (uint32_t) ts);
sprintf( buffer,
"NAME=%s;VER=%s;API=%s;"
"ALGO=%s;CPUS=%d;URL=%s;"
"HS=%.2f;KHS=%.2f;ACC=%d;REJ=%d;SOL=%d;"
"ACCMN=%.3f;DIFF=%s;TEMP=%.1f;FAN=%d;FREQ=%d;"
"UPTIME=%.0f;TS=%u|",
PACKAGE_NAME, PACKAGE_VERSION, APIVERSION,
algo, opt_n_threads, short_url,
hrate, hrate/1000.0, accepted_share_count, rejected_share_count,
solved_block_count,
accps, diff_str, cpu.cpu_temp, cpu.cpu_fan, cpu.cpu_clock,
uptime, (uint32_t) ts);
return buffer;
}

653
avxdefs.h
View File

File diff suppressed because it is too large Load Diff

26
build-allarch.sh
View File

@@ -1,10 +1,14 @@
#!/bin/bash
#
# This script is not intended for users, it is only used for compile testing
# during develpment. Howver the information contained my provide cimpilation
# tips to users.
make distclean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=core-avx2 -msha -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-avx2-sha.exe
strip -s cpuminer
@@ -14,7 +18,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=skylake-avx512 -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-avx512.exe
strip -s cpuminer
@@ -24,7 +28,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=core-avx2 -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-avx2.exe
strip -s cpuminer
@@ -34,7 +38,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=corei7-avx -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-aes-avx.exe
strip -s cpuminer
@@ -44,7 +48,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -maes -msse4.2 -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-aes-sse42.exe
strip -s cpuminer
@@ -54,7 +58,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=corei7 -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-sse42.exe
strip -s cpuminer
@@ -64,7 +68,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=core2 -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-ssse3.exe
strip -s cpuminer
@@ -74,7 +78,7 @@ make clean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -msse2 -Wall" ./configure --with-curl
make -j 4
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-sse2.exe
strip -s cpuminer
@@ -83,8 +87,8 @@ mv cpuminer cpuminer-sse2
make clean || echo done
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=znver1 -Wall" ./configure --with-curl
make -j 8
CFLAGS="-O3 -march=znver1 -DRYZEN_ -Wall" ./configure --with-curl
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-zen.exe
strip -s cpuminer
@@ -94,7 +98,7 @@ make clean || echo done
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=native -Wall" ./configure --with-curl
make -j 8
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe cpuminer-native.exe
strip -s cpuminer

20
configure vendored
View File

@@ -1,6 +1,6 @@
#! /bin/sh
# Guess values for system-dependent variables and create Makefiles.
# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.9.0.1.
# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.9.1.
#
#
# Copyright (C) 1992-1996, 1998-2012 Free Software Foundation, Inc.
@@ -577,8 +577,8 @@ MAKEFLAGS=
# Identity of this package.
PACKAGE_NAME='cpuminer-opt'
PACKAGE_TARNAME='cpuminer-opt'
PACKAGE_VERSION='3.9.0.1'
PACKAGE_STRING='cpuminer-opt 3.9.0.1'
PACKAGE_VERSION='3.9.1'
PACKAGE_STRING='cpuminer-opt 3.9.1'
PACKAGE_BUGREPORT=''
PACKAGE_URL=''
@@ -1332,7 +1332,7 @@ if test "$ac_init_help" = "long"; then
# Omit some internal or obsolete options to make the list less imposing.
# This message is too long to be a string in the A/UX 3.1 sh.
cat <<_ACEOF
\`configure' configures cpuminer-opt 3.9.0.1 to adapt to many kinds of systems.
\`configure' configures cpuminer-opt 3.9.1 to adapt to many kinds of systems.
Usage: $0 [OPTION]... [VAR=VALUE]...
@@ -1404,7 +1404,7 @@ fi
if test -n "$ac_init_help"; then
case $ac_init_help in
short | recursive ) echo "Configuration of cpuminer-opt 3.9.0.1:";;
short | recursive ) echo "Configuration of cpuminer-opt 3.9.1:";;
esac
cat <<\_ACEOF
@@ -1509,7 +1509,7 @@ fi
test -n "$ac_init_help" && exit $ac_status
if $ac_init_version; then
cat <<\_ACEOF
cpuminer-opt configure 3.9.0.1
cpuminer-opt configure 3.9.1
generated by GNU Autoconf 2.69
Copyright (C) 2012 Free Software Foundation, Inc.
@@ -2012,7 +2012,7 @@ cat >config.log <<_ACEOF
This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.
It was created by cpuminer-opt $as_me 3.9.0.1, which was
It was created by cpuminer-opt $as_me 3.9.1, which was
generated by GNU Autoconf 2.69. Invocation command line was
$ $0 $@
@@ -2993,7 +2993,7 @@ fi
# Define the identity of the package.
PACKAGE='cpuminer-opt'
VERSION='3.9.0.1'
VERSION='3.9.1'
cat >>confdefs.h <<_ACEOF
@@ -6690,7 +6690,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
# report actual input values of CONFIG_FILES etc. instead of their
# values after options handling.
ac_log="
This file was extended by cpuminer-opt $as_me 3.9.0.1, which was
This file was extended by cpuminer-opt $as_me 3.9.1, which was
generated by GNU Autoconf 2.69. Invocation command line was
CONFIG_FILES = $CONFIG_FILES
@@ -6756,7 +6756,7 @@ _ACEOF
cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
ac_cs_version="\\
cpuminer-opt config.status 3.9.0.1
cpuminer-opt config.status 3.9.1
configured by $0, generated by GNU Autoconf 2.69,
with options \\"\$ac_cs_config\\"

2
configure.ac
View File

@@ -1,4 +1,4 @@
AC_INIT([cpuminer-opt], [3.9.0.1])
AC_INIT([cpuminer-opt], [3.9.1])
AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM

58
cpu-miner.c
View File

@@ -141,9 +141,9 @@ double opt_diff_factor = 1.0;
uint32_t zr5_pok = 0;
bool opt_stratum_stats = false;
uint32_t accepted_count = 0L;
uint32_t rejected_count = 0L;
uint32_t solved_count = 0L;
uint32_t accepted_share_count = 0ULL;
uint32_t rejected_share_count = 0ULL;
uint32_t solved_block_count = 0ULL;
double *thr_hashrates;
double *thr_hashcount;
double global_hashcount = 0;
@@ -857,24 +857,24 @@ static int share_result( int result, struct work *work, const char *reason )
hashcount += thr_hashcount[i];
hashrate += thr_hashrates[i];
}
result ? accepted_count++ : rejected_count++;
result ? accepted_share_count++ : rejected_share_count++;
if ( solved )
{
solved_count++;
solved_block_count++;
if ( use_colors )
sprintf( sol, CL_GRN " Solved" CL_WHT " %d", solved_count );
sprintf( sol, CL_GRN " Solved" CL_WHT " %d", solved_block_count );
else
sprintf( sol, " Solved %d", solved_count );
sprintf( sol, " Solved %d", solved_block_count );
}
pthread_mutex_unlock(&stats_lock);
global_hashcount = hashcount;
global_hashrate = hashrate;
total_submits = accepted_count + rejected_count;
total_submits = accepted_share_count + rejected_share_count;
rate = ( result ? ( 100. * accepted_count / total_submits )
: ( 100. * rejected_count / total_submits ) );
rate = ( result ? ( 100. * accepted_share_count / total_submits )
: ( 100. * rejected_share_count / total_submits ) );
if (use_colors)
{
@@ -889,7 +889,7 @@ static int share_result( int result, struct work *work, const char *reason )
// Rates > 99% and < 100% (rejects>0) display 99.9%.
if ( result )
{
rate = 100. * accepted_count / total_submits;
rate = 100. * accepted_share_count / total_submits;
if ( rate == 100.0 )
sprintf( rate_s, "%.0f", rate );
else
@@ -897,7 +897,7 @@ static int share_result( int result, struct work *work, const char *reason )
}
else
{
rate = 100. * rejected_count / total_submits;
rate = 100. * rejected_share_count / total_submits;
if ( rate < 0.1 )
sprintf( rate_s, "%.1f", 0.10 );
else
@@ -926,26 +926,26 @@ static int share_result( int result, struct work *work, const char *reason )
{
#if ((defined(_WIN64) || defined(__WINDOWS__)))
applog( LOG_NOTICE, "%s %lu/%lu (%s%%), %s %sH, %s %sH/s",
sres, ( result ? accepted_count : rejected_count ),
total_submits, rate_s, hc, hc_units, hr, hr_units );
sres, ( result ? accepted_share_count : rejected_share_count ),
total_submits, rate_s, hc, hc_units, hr, hr_units );
#else
applog( LOG_NOTICE, "%s %lu/%lu (%s%%), %s %sH, %s %sH/s, %dC",
sres, ( result ? accepted_count : rejected_count ),
total_submits, rate_s, hc, hc_units, hr, hr_units,
(uint32_t)cpu_temp(0) );
sres, ( result ? accepted_share_count : rejected_share_count ),
total_submits, rate_s, hc, hc_units, hr, hr_units,
(uint32_t)cpu_temp(0) );
#endif
}
else
{
#if ((defined(_WIN64) || defined(__WINDOWS__)))
applog( LOG_NOTICE, "%s %lu/%lu (%s%%), diff %.3g%s, %s %sH/s",
sres, ( result ? accepted_count : rejected_count ),
total_submits, rate_s, sharediff, sol, hr, hr_units );
sres, ( result ? accepted_share_count : rejected_share_count ),
total_submits, rate_s, sharediff, sol, hr, hr_units );
#else
applog( LOG_NOTICE, "%s %lu/%lu (%s%%), diff %.3g%s, %s %sH/s, %dC",
sres, ( result ? accepted_count : rejected_count ),
total_submits, rate_s, sharediff, sol, hr, hr_units,
(uint32_t)cpu_temp(0) );
sres, ( result ? accepted_share_count : rejected_share_count ),
total_submits, rate_s, sharediff, sol, hr, hr_units,
(uint32_t)cpu_temp(0) );
#endif
}
@@ -1544,7 +1544,7 @@ static bool get_work(struct thr_info *thr, struct work *work)
return true;
}
static bool submit_work(struct thr_info *thr, const struct work *work_in)
bool submit_work(struct thr_info *thr, const struct work *work_in)
{
struct workio_cmd *wc;
/* fill out work request message */
@@ -1969,7 +1969,7 @@ static void *miner_thread( void *userdata )
// Scan for nonce
nonce_found = algo_gate.scanhash( thr_id, &work, max_nonce,
&hashes_done );
&hashes_done, mythr );
// record scanhash elapsed time
gettimeofday( &tv_end, NULL );
@@ -1998,7 +1998,7 @@ static void *miner_thread( void *userdata )
break;
}
}
else
else
{ // only 1 nonce, in work ready to submit.
if ( !submit_work( mythr, &work ) )
@@ -2043,7 +2043,7 @@ static void *miner_thread( void *userdata )
}
// Display benchmark total
// Update hashrate for API if no shares accepted yet.
if ( ( opt_benchmark || !accepted_count )
if ( ( opt_benchmark || !accepted_share_count )
&& thr_id == opt_n_threads - 1 )
{
double hashrate = 0.;
@@ -2482,15 +2482,15 @@ static void *stratum_thread(void *userdata )
if ( last_bloc_height != stratum.bloc_height )
{
last_bloc_height = stratum.bloc_height;
if ( !opt_quiet )
{
// if ( !opt_quiet )
// {
if (net_diff > 0.)
applog(LOG_BLUE, "%s block %d, network diff %.3f",
algo_names[opt_algo], stratum.bloc_height, net_diff);
else
applog(LOG_BLUE, "%s %s block %d", short_url,
algo_names[opt_algo], stratum.bloc_height);
}
// }
}
restart_threads();
}

109
interleave.h
View File

@@ -70,6 +70,13 @@
_mm_set_epi32( ((const uint32_t*)(s))[i3], ((const uint32_t*)(s))[i2], \
((const uint32_t*)(s))[i1], ((const uint32_t*)(s))[i0] )
// blend 2 vectors while interleaving: { hi[n], lo[n-1], ... hi[1], lo[0] }
#define mm128_interleave_blend_64( hi, lo ) \
_mm256_blend_epi16( hi, lo, 0x0f )
#define mm128_interleave_blend_32( hi, lo ) \
_mm6_blend_epi16( hi, lo, 0x33 )
// 1 sse2 block, 16 bytes * 4 lanes
static inline void mm128_interleave_4x32x128( void *d, const void *s0,
const void *s1, const void *s2, const void *s3 )
@@ -313,6 +320,18 @@ static inline void mm128_deinterleave_4x32x( void *dst0, void *dst1, void *dst2,
((const uint32_t*)(s))[i3], ((const uint32_t*)(s))[i2], \
((const uint32_t*)(s))[i1], ((const uint32_t*)(s))[i0] )
// Blend 2 vectors alternating hi & lo: { hi[n], lo[n-1], ... hi[1], lo[0] }
#define mm256_interleave_blend_128( hi, lo ) \
_mm256_blend_epi32( hi, lo, 0x0f )
#define mm256_interleave_blend_64( hi, lo ) \
_mm256_blend_epi32( hi, lo, 0x33 )
#define mm256_interleave_blend_32( hi, lo ) \
_mm256_blend_epi32( hi, lo, 0x55 )
// Used for AVX2 interleaving
static inline void mm256_interleave_8x32x256( void *d, const void *s00,
@@ -751,6 +770,78 @@ static inline void mm256_reinterleave_4x32( void *dst, void *src, int bit_len )
// bit_len == 1024
}
static inline void mm256_reinterleave_4x64_2x128( void *dst0, void *dst1,
const void *src, int bit_len )
{
__m256i* d0 = (__m256i*)dst0;
__m256i* d1 = (__m256i*)dst1;
uint64_t *s = (uint64_t*)src;
d0[0] = _mm256_set_epi64x( s[ 5], s[ 1], s[ 4], s[ 0] );
d1[0] = _mm256_set_epi64x( s[ 7], s[ 3], s[ 6], s[ 2] );
d0[1] = _mm256_set_epi64x( s[13], s[ 9], s[12], s[ 8] );
d1[1] = _mm256_set_epi64x( s[15], s[11], s[14], s[10] );
if ( bit_len <= 256 ) return;
d0[2] = _mm256_set_epi64x( s[21], s[17], s[20], s[16] );
d1[2] = _mm256_set_epi64x( s[23], s[19], s[22], s[18] );
d0[3] = _mm256_set_epi64x( s[29], s[25], s[28], s[24] );
d1[3] = _mm256_set_epi64x( s[31], s[27], s[30], s[26] );
if ( bit_len <= 512 ) return;
d0[4] = _mm256_set_epi64x( s[37], s[33], s[36], s[32] );
d1[4] = _mm256_set_epi64x( s[39], s[35], s[38], s[34] );
d0[5] = _mm256_set_epi64x( s[45], s[41], s[44], s[40] );
d1[5] = _mm256_set_epi64x( s[47], s[43], s[46], s[42] );
d0[6] = _mm256_set_epi64x( s[53], s[49], s[52], s[48] );
d1[6] = _mm256_set_epi64x( s[55], s[51], s[54], s[50] );
d0[7] = _mm256_set_epi64x( s[61], s[57], s[60], s[56] );
d1[7] = _mm256_set_epi64x( s[63], s[59], s[62], s[58] );
}
static inline void mm256_reinterleave_2x128_4x64( void *dst, const void *src0,
const void *src1, int bit_len )
{
__m256i* d = (__m256i*)dst;
uint64_t *s0 = (uint64_t*)src0;
uint64_t *s1 = (uint64_t*)src1;
d[ 0] = _mm256_set_epi64x( s1[2], s1[0], s0[2], s0[0] );
d[ 1] = _mm256_set_epi64x( s1[3], s1[1], s0[3], s0[1] );
d[ 2] = _mm256_set_epi64x( s1[6], s1[4], s0[6], s0[4] );
d[ 3] = _mm256_set_epi64x( s1[7], s1[5], s0[7], s0[5] );
if ( bit_len <= 256 ) return;
d[ 4] = _mm256_set_epi64x( s1[10], s1[ 8], s0[10], s0[ 8] );
d[ 5] = _mm256_set_epi64x( s1[11], s1[ 9], s0[11], s0[ 9] );
d[ 6] = _mm256_set_epi64x( s1[14], s1[12], s0[14], s0[12] );
d[ 7] = _mm256_set_epi64x( s1[15], s1[13], s0[15], s0[13] );
if ( bit_len <= 512 ) return;
d[ 8] = _mm256_set_epi64x( s1[18], s1[16], s0[18], s0[16] );
d[ 9] = _mm256_set_epi64x( s1[19], s1[17], s0[19], s0[17] );
d[10] = _mm256_set_epi64x( s1[22], s1[20], s0[22], s0[20] );
d[11] = _mm256_set_epi64x( s1[23], s1[21], s0[23], s0[21] );
d[12] = _mm256_set_epi64x( s1[26], s1[24], s0[26], s0[24] );
d[13] = _mm256_set_epi64x( s1[27], s1[25], s0[27], s0[25] );
d[14] = _mm256_set_epi64x( s1[30], s1[28], s0[30], s0[28] );
d[15] = _mm256_set_epi64x( s1[31], s1[29], s0[31], s0[29] );
}
/*
// not used
static inline void mm_reinterleave_4x32( void *dst, void *src, int bit_len )
@@ -850,6 +941,24 @@ static inline void mm256_deinterleave_2x128( void *d0, void *d1, void *s,
((const uint32_t*)(s))[i03], ((const uint32_t*)(s))[i02], \
((const uint32_t*)(s))[i01], ((const uint32_t*)(s))[i00] )
// AVX512 has no blend, can be done with permute2xvar but at what cost?
// Can also be done with shifting and mask-or'ing for 3 instructins with
// 1 dependency. Finally it can be done with 1 _mm512_set but with 8 64 bit
// array index calculations and 8 pointer reads.
// Blend 2 vectors alternating hi & lo: { hi[n], lo[n-1], ... hi[1]. lo[0] }
#define mm512_interleave_blend_128( hi, lo ) \
_mm256_permute2xvar_epi64( hi, lo, _mm512_set_epi64( \
0x7, 0x6, 0x5, 0x4, 0xb, 0xa, 0x9, 0x8 )
#define mm512_interleave_blend_64( hi, lo ) \
_mm256_permute2xvar_epi64( hi, lo, _mm512_set_epi64( \
0x7, 0x6, 0xd, 0xc, 0x3, 0x2, 0x9, 0x8 )
#define mm512_interleave_blend_32( hi, lo ) \
_mm256_permute2xvar_epi32( hi, lo, _mm512_set_epi32( \
0x0f, 0x1e, 0x0d, 0x1c, 0x0b, 0x1a, 0x09, 0x18, \
0x07, 0x16, 0x05, 0x14, 0x03, 0x12, 0x01, 0x10 )
//
static inline void mm512_interleave_16x32x512( void *d, const void *s00,

15
miner.h
View File

@@ -164,6 +164,8 @@ static inline void be32enc(void *pp, uint32_t x)
}
#endif
// Deprecated in favour of mm64_bswap_32
//
// This is a poorman's SIMD instruction, use 64 bit instruction to encode 2
// uint32_t. This function flips endian on two adjacent 32 bit quantities
// aligned to 64 bits. If source is LE output is BE, and vice versa.
@@ -540,7 +542,8 @@ enum algos {
ALGO_SKEIN,
ALGO_SKEIN2,
ALGO_SKUNK,
ALGO_TIMETRAVEL,
ALGO_SONOA,
ALGO_TIMETRAVEL,
ALGO_TIMETRAVEL10,
ALGO_TRIBUS,
ALGO_VANILLA,
@@ -626,7 +629,8 @@ static const char* const algo_names[] = {
"skein",
"skein2",
"skunk",
"timetravel",
"sonoa",
"timetravel",
"timetravel10",
"tribus",
"vanilla",
@@ -705,7 +709,9 @@ extern bool opt_stratum_stats;
extern int num_cpus;
extern int num_cpugroups;
extern int opt_priority;
extern uint32_t accepted_share_count;
extern uint32_t rejected_share_count;
extern uint32_t solved_block_count;
extern pthread_mutex_t rpc2_job_lock;
extern pthread_mutex_t rpc2_login_lock;
extern pthread_mutex_t applog_lock;
@@ -772,7 +778,8 @@ Options:\n\
skein Skein+Sha (Skeincoin)\n\
skein2 Double Skein (Woodcoin)\n\
skunk Signatum (SIGT)\n\
timetravel timeravel8, Machinecoin (MAC)\n\
sonoa Sono\n\
timetravel timeravel8, Machinecoin (MAC)\n\
timetravel10 Bitcore (BTX)\n\
tribus Denarius (DNR)\n\
vanilla blake256r8vnl (VCash)\n\

40
winbuild-cross.sh
View File

@@ -1,13 +1,27 @@
#!/bin/bash
#
# Script for building Windows binaries release package using mingw.
# Requires a custom mingw environment, not intended for users.
#
# Compiles Windows EXE files for selected CPU architectures, copies them
# as well as some DLLs that aren't available in most Windows environments
# into a release folder ready to be zipped and uploaded.
LOCAL_LIB="$HOME/usr/lib"
# define some local variables
export LOCAL_LIB="$HOME/usr/lib"
export LDFLAGS="-L$LOCAL_LIB/curl/lib/.libs -L$LOCAL_LIB/gmp/.libs -L$LOCAL_LIB/openssl"
F="--with-curl=$LOCAL_LIB/curl --with-crypto=$LOCAL_LIB/openssl --host=x86_64-w64-mingw32"
export CONFIGURE_ARGS="--with-curl=$LOCAL_LIB/curl --with-crypto=$LOCAL_LIB/openssl --host=x86_64-w64-mingw32"
# make link to local gmp header file.
ln -s $LOCAL_LIB/gmp/gmp.h ./gmp.h
# edit configure to fix pthread lib name for Windows.
sed -i 's/"-lpthread"/"-lpthreadGC2"/g' configure.ac
# make release directory and copy selected DLLs.
mkdir release
cp README.txt release/
cp /usr/x86_64-w64-mingw32/lib/zlib1.dll release/
@@ -20,27 +34,27 @@ cp $LOCAL_LIB/curl/lib/.libs/libcurl-4.dll release/
make distclean || echo clean
rm -f config.status
./autogen.sh || echo done
CFLAGS="-O3 -march=core-avx2 -msha -Wall" ./configure $F
CFLAGS="-O3 -march=znver1 -DRYZEN_ -Wall" ./configure $CONFIGURE_ARGS
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe release/cpuminer-avx2-sha.exe
mv cpuminer.exe release/cpuminer-zen.exe
#make clean || echo clean
#CFLAGS="-O3 -march=corei7-avx -msha -Wall" ./configure $F
#CFLAGS="-O3 -march=corei7-avx -msha -Wall" ./configure $CONFIGURE_ARGS
#make
#strip -s cpuminer.exe
#mv cpuminer.exe release/cpuminer-avx-sha.exe
make clean || echo clean
rm -f config.status
CFLAGS="-O3 -march=core-avx2 -Wall" ./configure $F
CFLAGS="-O3 -march=core-avx2 -Wall" ./configure $CONFIGURE_ARGS
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe release/cpuminer-avx2.exe
#make clean || echo clean
#rm -f config.status
#CFLAGS="-O3 -march=znver1 -Wall" ./configure $F
#CFLAGS="-O3 -march=znver1 -Wall" ./configure $CONFIGURE_ARGS
#make -j
#strip -s cpuminer.exe
#mv cpuminer.exe release/cpuminer-aes-sha.exe
@@ -48,7 +62,7 @@ mv cpuminer.exe release/cpuminer-avx2.exe
make clean || echo clean
rm -f config.status
CFLAGS="-O3 -march=corei7-avx -Wall" ./configure $F
CFLAGS="-O3 -march=corei7-avx -Wall" ./configure $CONFIGURE_ARGS
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe release/cpuminer-avx.exe
@@ -56,22 +70,22 @@ mv cpuminer.exe release/cpuminer-avx.exe
# -march=westmere is supported in gcc5
make clean || echo clean
rm -f config.status
CFLAGS="-O3 -march=westmere -Wall" ./configure $F
#CFLAGS="-O3 -maes -msse4.2 -Wall" ./configure $F
CFLAGS="-O3 -march=westmere -Wall" ./configure $CONFIGURE_ARGS
#CFLAGS="-O3 -maes -msse4.2 -Wall" ./configure $CONFIGURE_ARGS
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe release/cpuminer-aes-sse42.exe
#make clean || echo clean
#rm -f config.status
#CFLAGS="-O3 -march=corei7 -Wall" ./configure $F
#CFLAGS="-O3 -march=corei7 -Wall" ./configure $CONFIGURE_ARGS
#make
#strip -s cpuminer.exe
#mv cpuminer.exe release/cpuminer-sse42.exe
#make clean || echo clean
#rm -f config.status
#CFLAGS="-O3 -march=core2 -Wall" ./configure $F
#CFLAGS="-O3 -march=core2 -Wall" ./configure $CONFIGURE_ARGS
#make
#strip -s cpuminer.exe
#mv cpuminer.exe release/cpuminer-ssse3.exe
@@ -79,7 +93,7 @@ mv cpuminer.exe release/cpuminer-aes-sse42.exe
make clean || echo clean
rm -f config.status
CFLAGS="-O3 -msse2 -Wall" ./configure $F
CFLAGS="-O3 -msse2 -Wall" ./configure $CONFIGURE_ARGS
make -j 16
strip -s cpuminer.exe
mv cpuminer.exe release/cpuminer-sse2.exe