CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

Improve cuBLAS performance by using a memory pool (#1094)

Browse Source 
* Improve cuBLAS performance by using a memory pool

* Move cuda specific definitions to ggml-cuda.h/cu

* Add CXX flags to nvcc

* Change memory pool synchronization mechanism to a spin lock
General code cleanup
This commit is contained in:
slaren
2023-04-21 21:59:17 +02:00
committed by GitHub
parent 25d7abbd1f
commit 50cb666b8a
4 changed files with 170 additions and 109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
ggml-cuda.h
View File

@@ -1,7 +1,36 @@
#include <cublas_v2.h>
#include <cuda_runtime.h>
#ifdef __cplusplus
extern "C" {
#endif
#define CUDA_CHECK(err) \
do { \
cudaError_t err_ = (err); \
if (err_ != cudaSuccess) { \
fprintf(stderr, "CUDA error %d at %s:%d: %s\n", err_, __FILE__, __LINE__, \
cudaGetErrorString(err_)); \
exit(1); \
} \
} while (0)
#define CUBLAS_CHECK(err) \
do { \
cublasStatus_t err_ = (err); \
if (err_ != CUBLAS_STATUS_SUCCESS) { \
fprintf(stderr, "cuBLAS error %d at %s:%d\n", err_, __FILE__, __LINE__); \
exit(1); \
} \
} while (0)
extern cublasHandle_t g_cublasH;
extern cudaStream_t g_cudaStream;
void ggml_init_cublas(void);
void * ggml_cuda_pool_malloc(size_t size, size_t * actual_size);
void ggml_cuda_pool_free(void * ptr, size_t size);
void dequantize_row_q4_0_cuda(const void * vx, float * y, int k, cudaStream_t stream);
void dequantize_row_q4_1_cuda(const void * vx, float * y, int k, cudaStream_t stream);
void dequantize_row_q4_2_cuda(const void * vx, float * y, int k, cudaStream_t stream);