github-jwinterm/xmr-stak
1
0
Fork 0
mirror of https://github.com/jwinterm/xmr-stak.git synced 2026-10-17 05:41:41 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #77 from psychocrypt/topic-appveyorCudaSupport

Browse Source 
appveyor CUDA support
This commit is contained in:
fireice-uk
2017-10-30 17:21:25 +00:00
committed by GitHub
parent a7fb89bec2 34a3c701cf
commit b68e0fa5ff
7 changed files with 102 additions and 80 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.appveyor.yml
+8 -4
View File
@@ -1,5 +1,5 @@
# image
image: Visual Studio 2017
image: Visual Studio 2015
# build platform
platform: x64
@@ -11,17 +11,21 @@ install:
- mkdir c:\xmr-stak-dep
- curl -sL https://github.com/fireice-uk/xmr-stak-dep/releases/download/v1/xmr-stak-dep.zip -o xmr-stak-dep.zip
- 7z x xmr-stak-dep.zip -o"c:\xmr-stak-dep" -y > nul
- appveyor DownloadFile https://developer.nvidia.com/compute/cuda/8.0/prod/local_installers/cuda_8.0.44_windows-exe -FileName cuda_8.0.44_windows.exe
- cuda_8.0.44_windows.exe -s compiler_8.0 cudart_8.0
- set PATH=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin;%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v8.0\libnvvp;%PATH%
- nvcc -V
build_script:
- call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\Tools\VsMSBuildCmd.bat"
- call "C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7\Tools\VsMSBuildCmd.bat"
- cd c:\xmr-stak
- mkdir build
- cd build
- set CMAKE_PREFIX_PATH=C:\xmr-stak-dep\hwloc;C:\xmr-stak-dep\libmicrohttpd;C:\xmr-stak-dep\openssl;
- cmake -G "Visual Studio 15 2017 Win64" -T v140,host=x64 -DCUDA_ENABLE=OFF -DOpenCL_ENABLE=OFF ..
- cmake -G "Visual Studio 14 2015 Win64" -T v140,host=x64 .. -DWIN_UAC=OFF -DCUDA_ARCH=30
- cmake --build . --config Release --target install
test_script:
- cd c:\xmr-stak\build\bin\Release
- dir
# - xmr-stak.exe
- xmr-stak.exe --help
xmrstak/backend/cpu/hwlocMemory.cpp
+64
View File
@@ -0,0 +1,64 @@
#include "xmrstak/backend/cpu/hwlocMemory.hpp"
#ifndef CONF_NO_HWLOC
#include "xmrstak/misc/console.hpp"
#include <hwloc.h>
/** pin memory to NUMA node
*
* Set the default memory policy for the current thread to bind memory to the
* NUMA node.
*
* @param puId core id
*/
void bindMemoryToNUMANode( size_t puId )
{
int depth;
hwloc_topology_t topology;
hwloc_topology_init(&topology);
hwloc_topology_load(topology);
if(!hwloc_topology_get_support(topology)->membind->set_thisthread_membind)
{
printer::inst()->print_msg(L0, "hwloc: set_thisthread_membind not supported");
hwloc_topology_destroy(topology);
return;
}
depth = hwloc_get_type_depth(topology, HWLOC_OBJ_PU);
for( size_t i = 0;
i < hwloc_get_nbobjs_by_depth(topology, depth);
i++ )
{
hwloc_obj_t pu = hwloc_get_obj_by_depth(topology, depth, i);
if( pu->os_index == puId )
{
if( 0 > hwloc_set_membind_nodeset(
topology,
pu->nodeset,
HWLOC_MEMBIND_BIND,
HWLOC_MEMBIND_THREAD))
{
printer::inst()->print_msg(L0, "hwloc: can't bind memory");
}
else
{
printer::inst()->print_msg(L0, "hwloc: memory pinned");
break;
}
}
}
hwloc_topology_destroy(topology);
}
#else
void bindMemoryToNUMANode( size_t )
{
}
#endif
xmrstak/backend/cpu/hwlocMemory.hpp
+2 -54
View File
@@ -1,10 +1,6 @@
#pragma once
#include "xmrstak/misc/console.hpp"
#ifndef CONF_NO_HWLOC
#include <hwloc.h>
#include <cstddef>
/** pin memory to NUMA node
*
@@ -13,52 +9,4 @@
*
* @param puId core id
*/
void bindMemoryToNUMANode( size_t puId )
{
int depth;
hwloc_topology_t topology;
hwloc_topology_init(&topology);
hwloc_topology_load(topology);
if(!hwloc_topology_get_support(topology)->membind->set_thisthread_membind)
{
printer::inst()->print_msg(L0, "hwloc: set_thisthread_membind not supported");
hwloc_topology_destroy(topology);
return;
}
depth = hwloc_get_type_depth(topology, HWLOC_OBJ_PU);
for( size_t i = 0;
i < hwloc_get_nbobjs_by_depth(topology, depth);
i++ )
{
hwloc_obj_t pu = hwloc_get_obj_by_depth(topology, depth, i);
if( pu->os_index == puId )
{
if( 0 > hwloc_set_membind_nodeset(
topology,
pu->nodeset,
HWLOC_MEMBIND_BIND,
HWLOC_MEMBIND_THREAD))
{
printer::inst()->print_msg(L0, "hwloc: can't bind memory");
}
else
{
printer::inst()->print_msg(L0, "hwloc: memory pinned");
break;
}
}
}
hwloc_topology_destroy(topology);
}
#else
void bindMemoryToNUMANode( size_t )
{
}
#endif
void bindMemoryToNUMANode( size_t puId );
xmrstak/backend/nvidia/minethd.cpp
+3 -13
View File
@@ -222,7 +222,6 @@ void minethd::work_main()
}
bool mineMonero = strcmp_i(::jconf::inst()->GetCurrency(), "monero");
bool useAEON = strcmp_i(::jconf::inst()->GetCurrency(), "aeon");
while (bQuit == 0)
{
@@ -261,18 +260,9 @@ void minethd::work_main()
uint32_t foundCount;
cryptonight_extra_cpu_prepare(&ctx, iNonce);
#ifndef CONF_NO_MONERO
if(mineMonero)
{
cryptonight_core_cpu_hash<MONERO_ITER, MONERO_MASK, 19>(&ctx);
}
#endif
#ifndef CONF_NO_AEON
if(useAEON)
{
cryptonight_core_cpu_hash<MONERO_ITER, MONERO_MASK, 18>(&ctx);
}
#endif
cryptonight_core_cpu_hash(&ctx, mineMonero);
cryptonight_extra_cpu_final(&ctx, iNonce, oWork.iTarget, &foundCount, foundNonce);
for(size_t i = 0; i < foundCount; i++)
xmrstak/backend/nvidia/minethd.hpp
+1 -1
View File
@@ -4,7 +4,7 @@
#include "jconf.hpp"
#include "nvcc_code/cryptonight.hpp"
#include "xmrstak/backend/cpu/crypto/cryptonight.h"
#include "xmrstak/backend/cpu/minethd.hpp"
#include "xmrstak/backend/iBackend.hpp"
#include "xmrstak/misc/environment.hpp"
xmrstak/backend/nvidia/nvcc_code/cryptonight.hpp
+1 -3
View File
@@ -44,6 +44,4 @@ void cryptonight_extra_cpu_prepare(nvid_ctx* ctx, uint32_t startNonce);
void cryptonight_extra_cpu_final(nvid_ctx* ctx, uint32_t startNonce, uint64_t target, uint32_t* rescount, uint32_t *resnonce);
}
template<size_t ITERATIONS, size_t THREAD_SHIFT, uint32_t MASK>
void cryptonight_core_cpu_hash(nvid_ctx* ctx);
void cryptonight_core_cpu_hash(nvid_ctx* ctx, bool mineMonero);
xmrstak/backend/nvidia/nvcc_code/cuda_core.cu
+23 -5
View File
@@ -1,3 +1,5 @@
#include "xmrstak/backend/cryptonight.hpp"
#include <stdio.h>
#include <stdint.h>
#include <string.h>
@@ -92,7 +94,7 @@ __device__ __forceinline__ void storeGlobal32( T* addr, T const & val )
asm volatile( "st.global.cg.u32 [%0], %1;" : : "l"( addr ), "r"( val ) );
}
template<size_t ITERATIONS, size_t THREAD_SHIFT>
template<size_t ITERATIONS, uint32_t THREAD_SHIFT>
__global__ void cryptonight_core_gpu_phase1( int threads, int bfactor, int partidx, uint32_t * __restrict__ long_state, uint32_t * __restrict__ ctx_state, uint32_t * __restrict__ ctx_key1 )
{
__shared__ uint32_t sharedMemory[1024];
@@ -168,7 +170,7 @@ __forceinline__ __device__ uint32_t shuffle(volatile uint32_t* ptr,const uint32_
#ifdef XMR_STAK_THREADS
__launch_bounds__( XMR_STAK_THREADS * 4 )
#endif
template<size_t ITERATIONS, size_t THREAD_SHIFT, uint32_t MASK>
template<size_t ITERATIONS, uint32_t THREAD_SHIFT, uint32_t MASK>
__global__ void cryptonight_core_gpu_phase2( int threads, int bfactor, int partidx, uint32_t * d_long_state, uint32_t * d_ctx_a, uint32_t * d_ctx_b )
{
__shared__ uint32_t sharedMemory[1024];
@@ -257,7 +259,7 @@ __global__ void cryptonight_core_gpu_phase2( int threads, int bfactor, int parti
}
}
template<size_t ITERATIONS, size_t THREAD_SHIFT>
template<size_t ITERATIONS, uint32_t THREAD_SHIFT>
__global__ void cryptonight_core_gpu_phase3( int threads, int bfactor, int partidx, const uint32_t * __restrict__ long_state, uint32_t * __restrict__ d_ctx_state, uint32_t * __restrict__ d_ctx_key2 )
{
__shared__ uint32_t sharedMemory[1024];
@@ -292,8 +294,8 @@ __global__ void cryptonight_core_gpu_phase3( int threads, int bfactor, int parti
MEMCPY8( d_ctx_state + thread * 50 + sub + 16, text, 2 );
}
template<size_t ITERATIONS, size_t THREAD_SHIFT, uint32_t MASK>
void cryptonight_core_cpu_hash(nvid_ctx* ctx)
template<size_t ITERATIONS, uint32_t MASK, uint32_t THREAD_SHIFT>
void cryptonight_core_gpu_hash(nvid_ctx* ctx)
{
dim3 grid( ctx->device_blocks );
dim3 block( ctx->device_threads );
@@ -349,3 +351,19 @@ void cryptonight_core_cpu_hash(nvid_ctx* ctx)
ctx->d_ctx_state, ctx->d_ctx_key2 ));
}
}
void cryptonight_core_cpu_hash(nvid_ctx* ctx, bool mineMonero)
{
#ifndef CONF_NO_MONERO
if(mineMonero)
{
cryptonight_core_gpu_hash<MONERO_ITER, MONERO_MASK, 19u>(ctx);
}
#endif
#ifndef CONF_NO_AEON
if(!mineMonero)
{
cryptonight_core_gpu_hash<AEON_ITER, AEON_MASK, 18u>(ctx);
}
#endif
}