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

Disable precision fallback and pin cuBLAS 13.1

Browse Source
This commit is contained in:
Mikhail Chusavitin
2026-04-14 10:17:44 +03:00
parent 81e7c921f8
commit 82fe1f6d26
3 changed files with 33 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@@ -2,3 +2,4 @@
.DS_Store
dist/
iso/out/
build-cache/

2
iso/builder/VERSIONS
View File

@@ -6,7 +6,7 @@ NCCL_CUDA_VERSION=13.0
NCCL_SHA256=2e6faafd2c19cffc7738d9283976a3200ea9db9895907f337f0c7e5a25563186
NCCL_TESTS_VERSION=2.13.10
NVCC_VERSION=12.8
CUBLAS_VERSION=13.0.2.14-1
CUBLAS_VERSION=13.1.1.3-1
CUDA_USERSPACE_VERSION=13.0.96-1
DCGM_VERSION=4.5.3-1
JOHN_JUMBO_COMMIT=67fcf9fe5a

38
iso/builder/bee-gpu-stress.c
View File

@@ -1160,6 +1160,7 @@ static int run_cublaslt_stress(struct cuda_api *cuda,
return 0;
}
/* Count profiles matching the filter (for deciding what to run). */
for (size_t i = 0; i < sizeof(k_profiles) / sizeof(k_profiles[0]); i++) {
if (k_profiles[i].enabled && cc >= k_profiles[i].min_cc &&
(precision_filter == NULL || strcmp(k_profiles[i].block_label, precision_filter) == 0)) {
@@ -1173,18 +1174,31 @@ static int run_cublaslt_stress(struct cuda_api *cuda,
return 0;
}
/* Count all profiles active on this GPU regardless of filter.
* Used as the budget divisor so matrix sizes stay consistent whether
* running all precisions together or a single-precision phase. */
int planned_total = 0;
for (size_t i = 0; i < sizeof(k_profiles) / sizeof(k_profiles[0]); i++) {
if (k_profiles[i].enabled && cc >= k_profiles[i].min_cc) {
planned_total++;
}
}
if (planned_total < planned) {
planned_total = planned;
}
requested_budget = (size_t)size_mb * 1024u * 1024u;
if (requested_budget < (size_t)planned * MIN_PROFILE_BUDGET_BYTES) {
requested_budget = (size_t)planned * MIN_PROFILE_BUDGET_BYTES;
if (requested_budget < (size_t)planned_total * MIN_PROFILE_BUDGET_BYTES) {
requested_budget = (size_t)planned_total * MIN_PROFILE_BUDGET_BYTES;
}
total_budget = clamp_budget_to_free_memory(cuda, requested_budget);
if (total_budget < (size_t)planned * MIN_PROFILE_BUDGET_BYTES) {
total_budget = (size_t)planned * MIN_PROFILE_BUDGET_BYTES;
if (total_budget < (size_t)planned_total * MIN_PROFILE_BUDGET_BYTES) {
total_budget = (size_t)planned_total * MIN_PROFILE_BUDGET_BYTES;
}
if (query_multiprocessor_count(cuda, dev, &mp_count) &&
cuda->cuStreamCreate &&
cuda->cuStreamDestroy) {
stream_count = choose_stream_count(mp_count, planned, total_budget, 1);
stream_count = choose_stream_count(mp_count, planned_total, total_budget, 1);
}
if (stream_count > 1) {
int created = 0;
@@ -1197,7 +1211,7 @@ static int run_cublaslt_stress(struct cuda_api *cuda,
}
}
report->stream_count = stream_count;
per_profile_budget = total_budget / ((size_t)planned * (size_t)stream_count);
per_profile_budget = total_budget / ((size_t)planned_total * (size_t)stream_count);
if (per_profile_budget < MIN_PROFILE_BUDGET_BYTES) {
per_profile_budget = MIN_PROFILE_BUDGET_BYTES;
}
@@ -1425,7 +1439,17 @@ int main(int argc, char **argv) {
ok = run_cublaslt_stress(&cuda, dev, name, cc_major, cc_minor, seconds, size_mb, precision_filter, &report);
#endif
if (!ok) {
if (!run_ptx_fallback(&cuda, dev, name, cc_major, cc_minor, seconds, size_mb, &report)) {
if (precision_filter != NULL) {
fprintf(stderr,
"requested precision path unavailable: precision=%s device=%s cc=%d.%d\n",
precision_filter,
name,
cc_major,
cc_minor);
return 1;
}
int ptx_mb = size_mb;
if (!run_ptx_fallback(&cuda, dev, name, cc_major, cc_minor, seconds, ptx_mb, &report)) {
return 1;
}
}