extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-04-05 00:30:58 +02:00
Code Issues Packages Projects Releases Wiki Activity

musa: refine compute capability (llama/12493)

Browse Source 
* musa: refine compute capability

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

* Address review comments

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

---------

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
This commit is contained in:
R0CKSTAR 2025-03-22 17:11:37 +08:00 committed by Georgi Gerganov
parent cbb88c4050
commit d487a28ae1
5 changed files with 44 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
ggml/src/ggml-cuda/common.cuh
View File

@ -48,7 +48,10 @@
#define GGML_CUDA_CC_AMPERE 800 #define GGML_CUDA_CC_AMPERE 800
#define GGML_CUDA_CC_ADA_LOVELACE 890 #define GGML_CUDA_CC_ADA_LOVELACE 890
#define GGML_CUDA_CC_OFFSET_AMD 0x1000000 #define GGML_CUDA_CC_OFFSET_AMD 0x1000000
#define GGML_CUDA_CC_OFFSET_MTHREADS 0x0100000
#define GGML_CUDA_CC_IS_NVIDIA(cc) (cc < GGML_CUDA_CC_OFFSET_MTHREADS)
// AMD
// GCN/CNDA, wave size is 64 // GCN/CNDA, wave size is 64
#define GGML_CUDA_CC_GCN4 (GGML_CUDA_CC_OFFSET_AMD + 0x803) // Tonga, Fiji, Polaris, minimum for fast fp16 #define GGML_CUDA_CC_GCN4 (GGML_CUDA_CC_OFFSET_AMD + 0x803) // Tonga, Fiji, Polaris, minimum for fast fp16
#define GGML_CUDA_CC_VEGA (GGML_CUDA_CC_OFFSET_AMD + 0x900) // Vega56/64, minimum for fp16 dual issue #define GGML_CUDA_CC_VEGA (GGML_CUDA_CC_OFFSET_AMD + 0x900) // Vega56/64, minimum for fp16 dual issue
@ -70,8 +73,17 @@
#define GGML_CUDA_CC_IS_GCN(cc) (cc > GGML_CUDA_CC_OFFSET_AMD && cc < GGML_CUDA_CC_CDNA) #define GGML_CUDA_CC_IS_GCN(cc) (cc > GGML_CUDA_CC_OFFSET_AMD && cc < GGML_CUDA_CC_CDNA)
#define GGML_CUDA_CC_IS_CDNA(cc) (cc >= GGML_CUDA_CC_CDNA && cc < GGML_CUDA_CC_RDNA1) #define GGML_CUDA_CC_IS_CDNA(cc) (cc >= GGML_CUDA_CC_CDNA && cc < GGML_CUDA_CC_RDNA1)
#define GGML_CUDA_CC_QY1 210 // Moore Threads
#define GGML_CUDA_CC_QY2 220 #define GGML_CUDA_MUSA_ARCH_IS_QY1 (__MUSA_ARCH__ <= 210)
#define GGML_CUDA_CC_QY1 (GGML_MUSA_CC_OFFSET_MTHREADS + 0x210) // MTT S80, MTT S3000
#define GGML_CUDA_CC_QY2 (GGML_MUSA_CC_OFFSET_MTHREADS + 0x220) // MTT S4000
#define GGML_CUDA_CC_NG (GGML_MUSA_CC_OFFSET_MTHREADS + 0x310) // TBD
#define GGML_CUDA_CC_IS_MTHREADS(cc) (cc >= GGML_CUDA_CC_OFFSET_MTHREADS && cc < GGML_CUDA_CC_OFFSET_AMD)
#define GGML_CUDA_CC_IS_QY1(cc) (cc >= GGML_CUDA_CC_QY1 && cc < GGML_CUDA_CC_QY2)
#define GGML_CUDA_CC_IS_QY2(cc) (cc >= GGML_CUDA_CC_QY2 && cc < GGML_CUDA_CC_NEXT)
#define GGML_CUDA_CC_IS_NG(cc) (cc >= GGML_CUDA_CC_NG)
#ifdef __CUDA_ARCH_LIST__ #ifdef __CUDA_ARCH_LIST__
constexpr bool ggml_cuda_has_arch_impl(int) { constexpr bool ggml_cuda_has_arch_impl(int) {
@ -209,21 +221,21 @@ typedef float2 dfloat2;
#define CP_ASYNC_AVAILABLE #define CP_ASYNC_AVAILABLE
#endif // !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= GGML_CUDA_CC_AMPERE #endif // !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= GGML_CUDA_CC_AMPERE
#if !defined(GGML_CUDA_NO_FA) && !(defined(GGML_USE_MUSA) && __MUSA_ARCH__ <= GGML_CUDA_CC_QY1) #if !defined(GGML_CUDA_NO_FA) && !(defined(GGML_USE_MUSA) && GGML_CUDA_MUSA_ARCH_IS_QY1)
#define FLASH_ATTN_AVAILABLE #define FLASH_ATTN_AVAILABLE
#endif // !defined(GGML_CUDA_NO_FA) && !(defined(GGML_USE_MUSA) && __MUSA_ARCH__ <= GGML_CUDA_CC_QY1) #endif // !defined(GGML_CUDA_NO_FA) && !(defined(GGML_USE_MUSA) && GGML_CUDA_MUSA_ARCH_IS_QY1)
static bool fp16_available(const int cc) { static bool fp16_available(const int cc) {
return ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_PASCAL; return ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_PASCAL;
} }
static bool fast_fp16_available(const int cc) { static bool fast_fp16_available(const int cc) {
return fp16_available(cc) && cc != 610; return (GGML_CUDA_CC_IS_NVIDIA(cc) && fp16_available(cc) && cc != 610) || GGML_CUDA_CC_IS_AMD(cc);
} }
// To be used for feature selection of external libraries, e.g. cuBLAS. // To be used for feature selection of external libraries, e.g. cuBLAS.
static bool fast_fp16_hardware_available(const int cc) { static bool fast_fp16_hardware_available(const int cc) {
return cc >= GGML_CUDA_CC_PASCAL && cc != 610; return (GGML_CUDA_CC_IS_NVIDIA(cc) && cc >= GGML_CUDA_CC_PASCAL && cc != 610) || GGML_CUDA_CC_IS_AMD(cc);
} }
// Any FP16 tensor core instructions are available for ggml code. // Any FP16 tensor core instructions are available for ggml code.
@ -231,20 +243,20 @@ static bool fp16_mma_available(const int cc) {
#if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__) && !defined(GGML_HIP_ROCWMMA_FATTN) #if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__) && !defined(GGML_HIP_ROCWMMA_FATTN)
return false; return false;
#else #else
return cc < GGML_CUDA_CC_OFFSET_AMD && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA || return GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA ||
GGML_CUDA_CC_IS_CDNA(cc) || cc >= GGML_CUDA_CC_RDNA3; GGML_CUDA_CC_IS_CDNA(cc) || GGML_CUDA_CC_IS_RDNA3(cc);
#endif // defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__) && !defined(GGML_HIP_ROCWMMA_FATTN) #endif // defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__) && !defined(GGML_HIP_ROCWMMA_FATTN)
} }
// To be used for feature selection of external libraries, e.g. cuBLAS. // To be used for feature selection of external libraries, e.g. cuBLAS.
static bool fp16_mma_hardware_available(const int cc) { static bool fp16_mma_hardware_available(const int cc) {
return cc < GGML_CUDA_CC_OFFSET_AMD && cc >= GGML_CUDA_CC_VOLTA || return GGML_CUDA_CC_IS_NVIDIA(cc) && cc >= GGML_CUDA_CC_VOLTA ||
GGML_CUDA_CC_IS_CDNA(cc) || cc >= GGML_CUDA_CC_RDNA3; GGML_CUDA_CC_IS_CDNA(cc) || GGML_CUDA_CC_IS_RDNA3(cc);
} }
// Volta technically had FP16 tensor cores but they work very differently compared to Turing and later. // Volta technically had FP16 tensor cores but they work very differently compared to Turing and later.
static bool new_mma_available(const int cc) { static bool new_mma_available(const int cc) {
return cc < GGML_CUDA_CC_OFFSET_AMD && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_TURING; return GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_TURING;
} }
static bool cp_async_available(const int cc) { static bool cp_async_available(const int cc) {

2
ggml/src/ggml-cuda/fattn.cu
View File

@ -253,7 +253,7 @@ void ggml_cuda_flash_attn_ext(ggml_backend_cuda_context & ctx, ggml_tensor * dst
const int warp_size = ggml_cuda_info().devices[ggml_cuda_get_device()].warp_size; const int warp_size = ggml_cuda_info().devices[ggml_cuda_get_device()].warp_size;
const enum ggml_prec prec = ggml_flash_attn_ext_get_prec(KQV); const enum ggml_prec prec = ggml_flash_attn_ext_get_prec(KQV);
if (cc >= GGML_CUDA_CC_OFFSET_AMD) { if (GGML_CUDA_CC_IS_AMD(cc)) {
#if defined(GGML_HIP_ROCWMMA_FATTN) #if defined(GGML_HIP_ROCWMMA_FATTN)
if (fp16_mma_available(cc)) { if (fp16_mma_available(cc)) {
ggml_cuda_flash_attn_ext_wmma_f16(ctx, dst); ggml_cuda_flash_attn_ext_wmma_f16(ctx, dst);

10
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@ -264,9 +264,9 @@ static ggml_cuda_device_info ggml_cuda_init() {
#elif defined(GGML_USE_MUSA) #elif defined(GGML_USE_MUSA)
// FIXME: Ensure compatibility with varying warp sizes across different MUSA archs. // FIXME: Ensure compatibility with varying warp sizes across different MUSA archs.
info.devices[id].warp_size = 32; info.devices[id].warp_size = 32;
// TODO: refine the .cc to reflect MUSA's actual CC capabilities
info.devices[id].smpbo = prop.sharedMemPerBlockOptin; info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
info.devices[id].cc = 100*prop.major + 10*prop.minor; info.devices[id].cc = GGML_CUDA_CC_OFFSET_MTHREADS + prop.major * 0x100;
info.devices[id].cc += prop.minor * 0x10;
GGML_LOG_INFO(" Device %d: %s, compute capability %d.%d, VMM: %s\n", GGML_LOG_INFO(" Device %d: %s, compute capability %d.%d, VMM: %s\n",
id, prop.name, prop.major, prop.minor, device_vmm ? "yes" : "no"); id, prop.name, prop.major, prop.minor, device_vmm ? "yes" : "no");
#else #else
@ -1188,11 +1188,11 @@ static void ggml_cuda_op_mul_mat_cublas(
// ldc == nrows of the matrix that cuBLAS writes into // ldc == nrows of the matrix that cuBLAS writes into
int64_t ldc = id == ctx.device ? ne0 : row_diff; int64_t ldc = id == ctx.device ? ne0 : row_diff;
const int compute_capability = ggml_cuda_info().devices[id].cc; const int cc = ggml_cuda_info().devices[id].cc;
const bool use_fp16 = (src0->type == GGML_TYPE_F16 || ggml_is_quantized(src0->type)) && ggml_is_contiguous(src0) && row_diff == src0->ne[1] && dst->op_params[0] == GGML_PREC_DEFAULT; const bool use_fp16 = (src0->type == GGML_TYPE_F16 || ggml_is_quantized(src0->type)) && ggml_is_contiguous(src0) && row_diff == src0->ne[1] && dst->op_params[0] == GGML_PREC_DEFAULT;
if (compute_capability >= GGML_CUDA_CC_VOLTA && use_fp16) { if (((cc >= GGML_CUDA_CC_VOLTA && GGML_CUDA_CC_IS_NVIDIA(cc)) || GGML_CUDA_CC_IS_AMD(cc)) && use_fp16) {
// convert src0 and src1 to fp16, multiply as fp16, convert dst to fp32 // convert src0 and src1 to fp16, multiply as fp16, convert dst to fp32
ggml_cuda_pool_alloc<half> src0_as_f16(ctx.pool(id)); ggml_cuda_pool_alloc<half> src0_as_f16(ctx.pool(id));
if (src0->type != GGML_TYPE_F16) { if (src0->type != GGML_TYPE_F16) {
@ -1216,7 +1216,7 @@ static void ggml_cuda_op_mul_mat_cublas(
CUBLAS_CHECK(cublasSetStream(ctx.cublas_handle(id), stream)); CUBLAS_CHECK(cublasSetStream(ctx.cublas_handle(id), stream));
if (GGML_CUDA_CC_IS_CDNA(compute_capability)) { if (GGML_CUDA_CC_IS_CDNA(cc)) {
const float alpha = 1.0f; const float alpha = 1.0f;
const float beta = 0.0f; const float beta = 0.0f;
CUBLAS_CHECK( CUBLAS_CHECK(

4
ggml/src/ggml-cuda/mmq.cu
View File

@ -28,7 +28,7 @@ void ggml_cuda_op_mul_mat_q(
// Also its fixup needs to allocate a temporary buffer in the memory pool. // Also its fixup needs to allocate a temporary buffer in the memory pool.
// There are multiple parallel CUDA streams for src1_ncols != ne11 which would introduce a race condition for this buffer. // There are multiple parallel CUDA streams for src1_ncols != ne11 which would introduce a race condition for this buffer.
const bool use_stream_k = ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA && const bool use_stream_k = ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA &&
cc < GGML_CUDA_CC_OFFSET_AMD && src1_ncols == ne11; GGML_CUDA_CC_IS_NVIDIA(cc) && src1_ncols == ne11;
const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, ne11, nrows_dst, use_stream_k}; const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, ne11, nrows_dst, use_stream_k};
switch (src0->type) { switch (src0->type) {
@ -145,7 +145,7 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11) {
return true; return true;
#endif //GGML_CUDA_FORCE_MMQ #endif //GGML_CUDA_FORCE_MMQ
if (cc < GGML_CUDA_CC_OFFSET_AMD) { if (GGML_CUDA_CC_IS_NVIDIA(cc)) {
return !fp16_mma_hardware_available(cc) || ne11 < MMQ_DP4A_MAX_BATCH_SIZE; return !fp16_mma_hardware_available(cc) || ne11 < MMQ_DP4A_MAX_BATCH_SIZE;
} }

12
ggml/src/ggml-cuda/mmq.cuh
View File

@ -90,7 +90,7 @@ struct tile_x_sizes {
static int get_mmq_x_max_host(const int cc) { static int get_mmq_x_max_host(const int cc) {
return new_mma_available(cc) ? 128 : return new_mma_available(cc) ? 128 :
ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA && cc < GGML_CUDA_CC_OFFSET_AMD ? ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA && GGML_CUDA_CC_IS_NVIDIA(cc) ?
#ifdef GGML_CUDA_FORCE_MMQ #ifdef GGML_CUDA_FORCE_MMQ
128 : 64; 128 : 64;
#else #else
@ -123,8 +123,8 @@ static constexpr __device__ int get_mmq_x_max_device() {
} }
static int get_mmq_y_host(const int cc) { static int get_mmq_y_host(const int cc) {
return cc >= GGML_CUDA_CC_OFFSET_AMD ? (GGML_CUDA_CC_IS_RDNA1(cc) ? 64 : 128) : return GGML_CUDA_CC_IS_AMD(cc) ? (GGML_CUDA_CC_IS_RDNA1(cc) ? 64 : 128) :
(ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA ? 128 : 64); ((ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA && GGML_CUDA_CC_IS_NVIDIA(cc)) ? 128 : 64);
} }
static constexpr __device__ int get_mmq_y_device() { static constexpr __device__ int get_mmq_y_device() {
@ -2772,14 +2772,14 @@ static void launch_mul_mat_q(ggml_backend_cuda_context & ctx, const mmq_args & a
const int shmem = mmq_get_shmem<type>(mmq_x, mmq_y, cc); const int shmem = mmq_get_shmem<type>(mmq_x, mmq_y, cc);
#if !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)) #if !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)) && !defined(GGML_USE_MUSA)
static bool shmem_limit_raised[GGML_CUDA_MAX_DEVICES] = {false}; static bool shmem_limit_raised[GGML_CUDA_MAX_DEVICES] = {false};
if (!shmem_limit_raised[id]) { if (!shmem_limit_raised[id]) {
CUDA_CHECK(cudaFuncSetAttribute(mul_mat_q<type, mmq_x, MMQ_NWARPS, false>, cudaFuncAttributeMaxDynamicSharedMemorySize, shmem)); CUDA_CHECK(cudaFuncSetAttribute(mul_mat_q<type, mmq_x, MMQ_NWARPS, false>, cudaFuncAttributeMaxDynamicSharedMemorySize, shmem));
CUDA_CHECK(cudaFuncSetAttribute(mul_mat_q<type, mmq_x, MMQ_NWARPS, true>, cudaFuncAttributeMaxDynamicSharedMemorySize, shmem)); CUDA_CHECK(cudaFuncSetAttribute(mul_mat_q<type, mmq_x, MMQ_NWARPS, true>, cudaFuncAttributeMaxDynamicSharedMemorySize, shmem));
shmem_limit_raised[id] = true; shmem_limit_raised[id] = true;
} }
#endif // !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)) #endif // !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)) && !defined(GGML_USE_MUSA)
const int nty = (args.ne01 + mmq_y - 1) / mmq_y; const int nty = (args.ne01 + mmq_y - 1) / mmq_y;
const int ntx = (args.ne11 + mmq_x - 1) / mmq_x; const int ntx = (args.ne11 + mmq_x - 1) / mmq_x;
@ -2832,7 +2832,7 @@ void mul_mat_q_case(ggml_backend_cuda_context & ctx, const mmq_args & args, cuda
const int mmq_x_max = get_mmq_x_max_host(cc); const int mmq_x_max = get_mmq_x_max_host(cc);
const int mmq_y = get_mmq_y_host(cc); const int mmq_y = get_mmq_y_host(cc);
const int block_num_y = (args.ne01 + mmq_y - 1) / mmq_y; const int block_num_y = (args.ne01 + mmq_y - 1) / mmq_y;
const bool use_stream_k = ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA && cc < GGML_CUDA_CC_OFFSET_AMD; const bool use_stream_k = ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_VOLTA && GGML_CUDA_CC_IS_NVIDIA(cc);
int mmq_x_best = 0; int mmq_x_best = 0;
int nparts_best = INT_MAX; int nparts_best = INT_MAX;