reanimator/bee
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

Split bee-bench into perf and power workflows

Browse Source
This commit is contained in:
Mikhail Chusavitin
2026-04-14 17:33:13 +03:00
parent 54338dbae5
commit 95124d228f
17 changed files with 718 additions and 259 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
iso/builder/bee-gpu-stress.c
View File

@@ -796,6 +796,9 @@ static cudaDataType_t matmul_scale_type(const struct profile_desc *desc) {
if (desc->compute_type == CUBLAS_COMPUTE_32I) {
return CUDA_R_32I;
}
if (desc->compute_type == CUBLAS_COMPUTE_64F) {
return CUDA_R_64F;
}
return CUDA_R_32F;
}
@@ -1120,6 +1123,8 @@ static int run_cublas_profile(cublasLtHandle_t handle,
struct prepared_profile *profile) {
int32_t alpha_i32 = 1;
int32_t beta_i32 = 0;
double alpha_f64 = 1.0;
double beta_f64 = 0.0;
float alpha = 1.0f;
float beta = 0.0f;
const void *alpha_ptr = α
@@ -1127,6 +1132,9 @@ static int run_cublas_profile(cublasLtHandle_t handle,
if (profile->desc.compute_type == CUBLAS_COMPUTE_32I) {
alpha_ptr = &alpha_i32;
beta_ptr = &beta_i32;
} else if (profile->desc.compute_type == CUBLAS_COMPUTE_64F) {
alpha_ptr = &alpha_f64;
beta_ptr = &beta_f64;
}
return check_cublas(profile->desc.name,
cublas->cublasLtMatmul(handle,