sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-05-15 05:24:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: batch out_prod inner loop with cublasSgemmStridedBatched (#22651)

Browse Source 
* CUDA: batch out_prod inner loop with cublasSgemmStridedBatched

* CUDA: batch out_prod inner loop with cublasSgemmStridedBatched

* CUDA: add cublasSgemmStridedBatched mapping for HIP and MUSA backends
This commit is contained in:
leonardHONG
2026-05-08 03:59:29 +08:00
committed by GitHub
parent aaf4a4d5e0
commit 05ff59cb57
4 changed files with 31 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/test-backend-ops.cpp
View File

@@ -8385,6 +8385,12 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
}
}
// ne2 sweep to cover the cublasSgemmStridedBatched path (dps2 == 1, ne2 > 1)
for (int64_t ne2 : {1, 8, 16, 32}) {
test_cases.emplace_back(new test_out_prod(GGML_TYPE_F32, GGML_TYPE_F32,
256, 16, 16, {ne2, 1}, {1, 1}));
}
// add_id
for (ggml_type type_a : {GGML_TYPE_F32}) {
for (ggml_type type_b : {GGML_TYPE_F32}) {