Skip to content

vulkan: Add missing bounds checking to scalar/coopmat1 mul_mat_id #15334

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 16, 2025
Merged

vulkan: Add missing bounds checking to scalar/coopmat1 mul_mat_id #15334

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 7 additions & 3 deletions ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -801,7 +801,7 @@ void main() {
}
#else
const uint row_i = ic * BN + loadc_b + l;
if (row_i < _ne1) {
if (row_i < _ne1 && block + loadr_b < end_k) {
Copy link
Collaborator Author

@jeffbolznv jeffbolznv Aug 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@0cc4m I couldn't reproduce a failure without this check, but it seems like it should be necessary. Am I missing something?

Copy link
Collaborator

@0cc4m 0cc4m Aug 16, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure. It does seem like it should be there, yeah.

const u16vec2 row_idx = row_ids[row_i];
buf_b[(loadc_b + l) * SHMEM_STRIDE + loadr_b] = TO_FLOAT_TYPE(data_b[pos_b + row_idx.y * p.batch_stride_b + (row_idx.x % p.ne11) * p.stride_b + loadr_b]);
} else {
Expand Down Expand Up @@ -875,7 +875,9 @@ void main() {

const u16vec2 row_idx = row_ids[row_i];

data_d[row_idx.y * p.batch_stride_d + row_idx.x * p.stride_d + dr + cm_row * TM + store_r] = D_TYPE(coopmat_stage[warp_i * TM * TN + (col + store_c) * TM + store_r]);
if (dr + cm_row * TM + store_r < p.M) {
data_d[row_idx.y * p.batch_stride_d + row_idx.x * p.stride_d + dr + cm_row * TM + store_r] = D_TYPE(coopmat_stage[warp_i * TM * TN + (col + store_c) * TM + store_r]);
}
}
}
}
Expand Down Expand Up @@ -925,7 +927,9 @@ void main() {
#endif // MUL_MAT_ID
[[unroll]] for (uint cr = 0; cr < TM; cr++) {
#ifdef MUL_MAT_ID
data_d[row_idx.y * p.batch_stride_d + row_idx.x * p.stride_d + dr_warp + cr] = D_TYPE(sums[(wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr]);
if (dr_warp + cr < p.M) {
data_d[row_idx.y * p.batch_stride_d + row_idx.x * p.stride_d + dr_warp + cr] = D_TYPE(sums[(wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr]);
}
#else
if (dr_warp + cr < p.M && dc_warp + cc < p.N) {
data_d[offsets + (dc_warp + cc) * p.stride_d + dr_warp + cr] = D_TYPE(sums[(wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr]);
Expand Down
2 changes: 2 additions & 0 deletions tests/test-backend-ops.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5823,6 +5823,8 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_mul_mat_id(GGML_TYPE_F16, GGML_TYPE_F32, 16, 16, b, 32, 1024, 16));
}

test_cases.emplace_back(new test_mul_mat_id(GGML_TYPE_F16, GGML_TYPE_F32, 1, 1, false, 8, 16, 1));

for (ggml_type type_a : base_types) {
for (ggml_type type_b : {GGML_TYPE_F32 /*, GGML_TYPE_F16 */}) {
for (int n_mats : {4, 8}) {
Expand Down
Loading