Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CUDA: reduce cn-v8 shared mem footprint #2078

Merged
merged 1 commit into from
Nov 20, 2018
Merged

CUDA: reduce cn-v8 shared mem footprint #2078

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions xmrstak/backend/nvidia/nvcc_code/cuda_aes.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -303,3 +303,9 @@ __device__ __forceinline__ static void cn_aes_gpu_init(uint32_t *sharedMemory)
for(int i = threadIdx.x; i < 1024; i += blockDim.x)
sharedMemory[i] = d_t_fn[i];
}

__device__ __forceinline__ static void cn_aes_gpu_init_half(uint32_t *sharedMemory)
{
for(int i = threadIdx.x; i < 512; i += blockDim.x)
sharedMemory[i] = d_t_fn[i];
}
17 changes: 11 additions & 6 deletions xmrstak/backend/nvidia/nvcc_code/cuda_core.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,6 +121,11 @@ __device__ __forceinline__ void storeGlobal64( T* addr, T const & val )
#endif
}

__device__ __forceinline__ uint32_t rotate16( const uint32_t n )
{
return (n >> 16u) | (n << 16u);
}

template<size_t ITERATIONS, uint32_t MEMORY>
__global__ void cryptonight_core_gpu_phase1( int threads, int bfactor, int partidx, uint32_t * __restrict__ long_state, uint32_t * __restrict__ ctx_state2, uint32_t * __restrict__ ctx_key1 )
{
Expand Down Expand Up @@ -267,9 +272,9 @@ __launch_bounds__( XMR_STAK_THREADS * 2 )
__global__ void cryptonight_core_gpu_phase2_double( int threads, int bfactor, int partidx, uint32_t * d_long_state, uint32_t * d_ctx_a, uint32_t * d_ctx_b, uint32_t * d_ctx_state,
uint32_t startNonce, uint32_t * __restrict__ d_input )
{
__shared__ uint32_t sharedMemory[1024];
__shared__ uint32_t sharedMemory[512];

cn_aes_gpu_init( sharedMemory );
cn_aes_gpu_init_half( sharedMemory );

#if( __CUDA_ARCH__ < 300 )
extern __shared__ uint64_t externShared[];
Expand Down Expand Up @@ -340,8 +345,8 @@ __global__ void cryptonight_core_gpu_phase2_double( int threads, int bfactor, in
const u64 cx2 = myChunks[ idx1 + ((sub + 1) & 1) ];

u64 cx_aes = ax0 ^ u64(
t_fn0( cx.x & 0xff ) ^ t_fn1( (cx.y >> 8) & 0xff ) ^ t_fn2( (cx2.x >> 16) & 0xff ) ^ t_fn3( (cx2.y >> 24 ) ),
t_fn0( cx.y & 0xff ) ^ t_fn1( (cx2.x >> 8) & 0xff ) ^ t_fn2( (cx2.y >> 16) & 0xff ) ^ t_fn3( (cx.x >> 24 ) )
t_fn0( cx.x & 0xff ) ^ t_fn1( (cx.y >> 8) & 0xff ) ^ rotate16(t_fn0( (cx2.x >> 16) & 0xff ) ^ t_fn1( (cx2.y >> 24 ) )),
t_fn0( cx.y & 0xff ) ^ t_fn1( (cx2.x >> 8) & 0xff ) ^ rotate16(t_fn0( (cx2.y >> 16) & 0xff ) ^ t_fn1( (cx.x >> 24 ) ))
);

if(ALGO == cryptonight_monero_v8)
Expand Down Expand Up @@ -840,9 +845,9 @@ void cryptonight_core_gpu_hash(nvid_ctx* ctx, uint32_t nonce)
void cryptonight_core_cpu_hash(nvid_ctx* ctx, xmrstak_algo miner_algo, uint32_t startNonce)
{
typedef void (*cuda_hash_fn)(nvid_ctx* ctx, uint32_t nonce);

if(miner_algo == invalid_algo) return;

static const cuda_hash_fn func_table[] = {
cryptonight_core_gpu_hash<CRYPTONIGHT_ITER, CRYPTONIGHT_MASK, CRYPTONIGHT_MEMORY/4, cryptonight, 0>,
cryptonight_core_gpu_hash<CRYPTONIGHT_ITER, CRYPTONIGHT_MASK, CRYPTONIGHT_MEMORY/4, cryptonight, 1>,
Expand Down