nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 5 Packages Projects Releases Wiki Activity

CUDA: add fast walsh-hadamard transform (#23615)

Browse Source 
* CUDA: add fast walsh-hadamard transform

* review: add unrolls + change size_t -> int

* warp size 64

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
Aman Gupta
2026-05-25 21:12:10 +08:00
committed by GitHub
co-authored by Johannes Gäßler
parent 5a4126adc1
commit c1f1e28d29
4 changed files with 120 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+1
View File
@@ -8308,6 +8308,7 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_mul_mat_hadamard(GGML_TYPE_F32, GGML_TYPE_F32, 64, 1, 64));
test_cases.emplace_back(new test_mul_mat_hadamard(GGML_TYPE_F32, GGML_TYPE_F32, 256, 1, 256));
test_cases.emplace_back(new test_mul_mat_hadamard(GGML_TYPE_F32, GGML_TYPE_F32, 128, 32, 128));
test_cases.emplace_back(new test_mul_mat_hadamard(GGML_TYPE_F32, GGML_TYPE_F32, 128, 4, 128, {2, 3}));
#if 0
// > 4GB A matrix. Too slow to be enabled by default.