vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)

Browse source 
* mtgpu: add mp_21 support

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

* mtgpu: disable flash attention on qy1 (MTT S80); disable q3_k and mul_mat_batched_cublas

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

* mtgpu: enable unified memory

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

* mtgpu: map cublasOperation_t to mublasOperation_t (sync code to latest)

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

---------

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
This commit is contained in:
R0CKSTAR 2024-09-22 22:55:49 +08:00 • committed by GitHub
parent 912c331d3d
commit c35e586ea5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 31 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml/src/ggml-cuda/common.cuh
View file

@ -50,6 +50,8 @@
#define CC_RDNA1 (CC_OFFSET_AMD + 1010)
#define CC_RDNA2 (CC_OFFSET_AMD + 1030)
#define CC_RDNA3 (CC_OFFSET_AMD + 1100)
#define CC_QY1 210
#define CC_QY2 220
#define MATRIX_ROW_PADDING 512 // last row of quant. matrices is a multiple of this to avoid out-of-bounds memory accesses
@ -134,6 +136,10 @@ typedef float2 dfloat2;
#define INT8_MMA_AVAILABLE
#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_TURING
#if !(defined(GGML_USE_MUSA) && __MUSA_ARCH__ <= CC_QY1)
#define FLASH_ATTN_AVAILABLE
#endif // !(defined(GGML_USE_MUSA) && __MUSA_ARCH__ <= CC_QY1)
static constexpr bool fast_fp16_available(const int cc) {
return cc >= CC_PASCAL && cc != 610;
}

6
ggml/src/ggml-cuda/fattn-tile-f32.cu
View file

@ -44,13 +44,17 @@ static __global__ void flash_attn_tile_ext_f32(
const int ne1,
const int ne2,
const int ne3) {
#ifndef FLASH_ATTN_AVAILABLE
NO_DEVICE_CODE;
return;
#endif // FLASH_ATTN_AVAILABLE
// Skip unused kernel variants for faster compilation:
if (use_logit_softcap && !(D == 128 || D == 256)) {
NO_DEVICE_CODE;
return;
}
//In this kernel Q, K, V are matrices while i, j, k are matrix indices.
// In this kernel Q, K, V are matrices while i, j, k are matrix indices.
const int ic0 = (blockIdx.x / parallel_blocks) * ncols; // Index of the Q/QKV column to work on.
const int ip = blockIdx.x % parallel_blocks; // Index in group of blocks running for the same column in parallel.

2
ggml/src/ggml-cuda/vendors/musa.h vendored
View file

@ -26,6 +26,7 @@
#define cublasSetStream mublasSetStream
#define cublasSgemm mublasSgemm
#define cublasStatus_t mublasStatus_t
#define cublasOperation_t mublasOperation_t
#define cublasGetStatusString mublasStatus_to_string
#define cudaDataType_t musaDataType_t
#define cudaDeviceCanAccessPeer musaDeviceCanAccessPeer
@ -56,6 +57,7 @@
#define cudaLaunchHostFunc musaLaunchHostFunc
#define cudaMalloc musaMalloc
#define cudaMallocHost musaMallocHost
#define cudaMallocManaged musaMallocManaged
#define cudaMemcpy musaMemcpy
#define cudaMemcpyAsync musaMemcpyAsync
#define cudaMemcpyPeerAsync musaMemcpyPeerAsync