Skip to content

cuda: unblock mmq for MoE on sm_60 #26264

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

Open
dfriehs wants to merge 1 commit into
base: master
Choose a base branch
from
+16 −0
Open

cuda: unblock mmq for MoE on sm_60 #26264

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
16 changes: 16 additions & 0 deletions ggml/src/ggml-cuda/mmq.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -314,6 +314,22 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11, int64_t
}

if (ggml_cuda_highest_compiled_arch(cc) < GGML_CUDA_CC_DP4A) {
// for MoE, mmq can be faster even without native dp4a
// TODO: check if cards older than pascal might benefit from this as well
if (cc == GGML_CUDA_CC_PASCAL && n_experts > 0) {
// TODO: figure out why these are so slow on sm_60
// ref: https://github.com/ggml-org/llama.cpp/pull/26264#issuecomment-5141087554
switch (type) {
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
return false;
default:
return true;
}
}

return false;
}

Expand Down