vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

vulkan : fix Qantized Mat-Vec Mul on AMD GPUs for ncols < 64 (#8855)

Browse source 
* Fix Vulkan mul mat vec invalid results when ncols < warp size

* Only run backend ops mul mat vec block size test if block size not already covered
This commit is contained in:
0cc4m 2024-08-05 07:52:55 +02:00 • committed by GitHub
parent 5587e57a76
commit 064cdc265f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 14 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

7
tests/test-backend-ops.cpp
View file

@ -2271,9 +2271,10 @@ static bool test_backend(ggml_backend_t backend, test_mode mode, const char * op
for (ggml_type type_a : other_types) {
for (ggml_type type_b : {GGML_TYPE_F32}) {
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 1, ggml_blck_size(type_a), { 1, 1}, {1, 1}));
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 1, 256, { 1, 1}, {1, 1}));
if (ggml_blck_size(type_a) != 256) {
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 1, ggml_blck_size(type_a), {1, 1}, {1, 1}));
}
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 1, 256, {1, 1}, {1, 1}));
}
}