Skip to content

CUDA: fix --split-mode row race condition #9413

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
Sep 11, 2024
Merged

CUDA: fix --split-mode row race condition #9413

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
6 changes: 5 additions & 1 deletion ggml/src/ggml-cuda/mmq.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,11 @@ void ggml_cuda_op_mul_mat_q(
// nrows_dst == nrows of the matrix that the kernel writes into
const int64_t nrows_dst = id == ctx.device ? ne0 : row_diff;

const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, ne11, nrows_dst};
// The stream-k decomposition is only faster for recent NVIDIA GPUs.
// Also its fixup needs to allocate a temporary buffer in the memory pool.
// There are multiple parallel CUDA streams for src1_ncols != ne11 which would introduce a race condition for this buffer.
const bool use_stream_k = compute_capability >= CC_VOLTA && compute_capability < CC_OFFSET_AMD && src1_ncols == ne11;
const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, ne11, nrows_dst, use_stream_k};

switch (src0->type) {
case GGML_TYPE_Q4_0:
Expand Down
4 changes: 2 additions & 2 deletions ggml/src/ggml-cuda/mmq.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2742,6 +2742,7 @@ struct mmq_args {
int64_t ne00; int64_t ne01; int64_t stride01;
int64_t ne10; int64_t ne11; int64_t stride11;
int64_t ne0;
bool use_stream_k;
};

template<ggml_type type>
Expand Down Expand Up @@ -2777,8 +2778,7 @@ static void launch_mul_mat_q(ggml_backend_cuda_context & ctx, const mmq_args & a
const int ntx = (args.ne11 + mmq_x - 1) / mmq_x;
const dim3 block_nums_xy_tiling(nty, ntx, 1);

const bool use_stream_k = cc >= CC_VOLTA && cc < CC_OFFSET_AMD;
if (!use_stream_k) {
if (!args.use_stream_k) {
if (args.ne01 % mmq_y == 0) {
constexpr bool need_check = false;
mul_mat_q<type, mmq_x, MMQ_NWARPS, need_check><<<block_nums_xy_tiling, block_dims, shmem, stream>>>
Expand Down
Loading