extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-01-27 00:09:30 +01:00
Code Issues Packages Projects Releases Wiki Activity

cuda : replace remaining shfl_xor with calls to warp_reduce functions (llama/5744)

Browse Source
This commit is contained in:
Engininja2 2024-02-27 07:22:45 -06:00 committed by Georgi Gerganov
parent 1c71816eab
commit d83f371b5f
No known key found for this signature in database
GPG Key ID: BF970631944C16B7
1 changed files with 24 additions and 49 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

73
ggml-cuda.cu
View File

@ -696,18 +696,20 @@ static __device__ __forceinline__ float2 warp_reduce_sum(float2 a) {
return a; return a;
} }
//static __device__ __forceinline__ half2 warp_reduce_sum(half2 a) { #ifdef GGML_CUDA_F16
//#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL static __device__ __forceinline__ half2 warp_reduce_sum(half2 a) {
//#pragma unroll #if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL
// for (int mask = 16; mask > 0; mask >>= 1) { #pragma unroll
// a = __hadd2(a, __shfl_xor_sync(0xffffffff, a, mask, 32)); for (int mask = 16; mask > 0; mask >>= 1) {
// } a = __hadd2(a, __shfl_xor_sync(0xffffffff, a, mask, 32));
// return a; }
//#else return a;
// (void) a; #else
// NO_DEVICE_CODE; (void) a;
//#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL NO_DEVICE_CODE;
//} #endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL
}
#endif // GGML_CUDA_F16
static __device__ __forceinline__ float warp_reduce_max(float x) { static __device__ __forceinline__ float warp_reduce_max(float x) {
#pragma unroll #pragma unroll
@ -2521,10 +2523,7 @@ static __global__ void dequantize_mul_mat_vec_q2_k(const void * __restrict__ vx,
#endif #endif
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (threadIdx.x == 0) { if (threadIdx.x == 0) {
dst[row] = tmp; dst[row] = tmp;
@ -2625,10 +2624,7 @@ static __global__ void dequantize_mul_mat_vec_q3_k(const void * __restrict__ vx,
#endif #endif
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (threadIdx.x == 0) { if (threadIdx.x == 0) {
dst[row] = tmp; dst[row] = tmp;
@ -2761,10 +2757,7 @@ static __global__ void dequantize_mul_mat_vec_q4_k(const void * __restrict__ vx,
#endif #endif
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (tid == 0) { if (tid == 0) {
dst[row] = tmp; dst[row] = tmp;
@ -2877,10 +2870,7 @@ static __global__ void dequantize_mul_mat_vec_q5_k(const void * __restrict__ vx,
#endif #endif
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (threadIdx.x == 0) { if (threadIdx.x == 0) {
dst[row] = tmp; dst[row] = tmp;
@ -2987,10 +2977,7 @@ static __global__ void dequantize_mul_mat_vec_q6_k(const void * __restrict__ vx,
#endif #endif
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (tid == 0) { if (tid == 0) {
dst[row] = tmp; dst[row] = tmp;
@ -3025,11 +3012,8 @@ static __global__ void quantize_q8_1(const float * __restrict__ x, void * __rest
float amax = fabsf(xi); float amax = fabsf(xi);
float sum = xi; float sum = xi;
#pragma unroll amax = warp_reduce_max(amax);
for (int mask = 16; mask > 0; mask >>= 1) { sum = warp_reduce_sum(sum);
amax = fmaxf(amax, __shfl_xor_sync(0xffffffff, amax, mask, 32));
sum += __shfl_xor_sync(0xffffffff, sum, mask, 32);
}
const float d = amax / 127; const float d = amax / 127;
const int8_t q = amax == 0.0f ? 0 : roundf(xi / d); const int8_t q = amax == 0.0f ? 0 : roundf(xi / d);
@ -6222,10 +6206,7 @@ static __global__ void dequantize_mul_mat_vec(const void * __restrict__ vx, cons
} }
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (tid == 0) { if (tid == 0) {
#ifdef GGML_CUDA_F16 #ifdef GGML_CUDA_F16
@ -6275,10 +6256,7 @@ static __global__ void mul_mat_p021_f16_f32(
const int idst = channel*nrows_dst + row_dst; const int idst = channel*nrows_dst + row_dst;
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (threadIdx.x == 0) { if (threadIdx.x == 0) {
dst[idst] = tmp; dst[idst] = tmp;
@ -6321,10 +6299,7 @@ static __global__ void mul_mat_vec_nc_f16_f32( // nc == non-contiguous
} }
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll tmp = warp_reduce_sum(tmp);
for (int mask = 16; mask > 0; mask >>= 1) {
tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
}
if (threadIdx.x == 0) { if (threadIdx.x == 0) {
dst[idst] = tmp; dst[idst] = tmp;