Skip to content

Use cuBLAS for large batches and quants with block size 16 #559

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
Jun 27, 2025
Merged

Use cuBLAS for large batches and quants with block size 16 #559

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
34 changes: 18 additions & 16 deletions ggml/src/ggml-cuda/mmq.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -163,20 +163,33 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11) {
bool mmq_supported;

switch (type) {
case GGML_TYPE_Q2_K: mmq_supported = ne11 < 384; break;
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ2_S:
mmq_supported = ne11 < 1536;
break;
case GGML_TYPE_IQ2_K:
case GGML_TYPE_IQ3_K:
case GGML_TYPE_IQ4_K:
case GGML_TYPE_IQ5_K:
case GGML_TYPE_IQ6_K:
case GGML_TYPE_IQ2_K_R4:
case GGML_TYPE_IQ3_K_R4:
case GGML_TYPE_IQ4_K_R4:
case GGML_TYPE_IQ5_K_R4:
mmq_supported = ne11 < 1024;
break;
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q6_0:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
case GGML_TYPE_IQ2_S:
case GGML_TYPE_IQ3_XXS:
case GGML_TYPE_IQ3_S:
case GGML_TYPE_IQ1_S:
Expand All @@ -188,22 +201,11 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11) {
case GGML_TYPE_IQ5_KS:
case GGML_TYPE_IQ5_KS_R4:
case GGML_TYPE_IQ2_KS:
case GGML_TYPE_IQ2_K:
case GGML_TYPE_IQ3_K:
case GGML_TYPE_IQ4_K:
case GGML_TYPE_IQ5_K:
case GGML_TYPE_IQ6_K:
case GGML_TYPE_IQ2_KT:
case GGML_TYPE_IQ3_KT:
case GGML_TYPE_IQ4_KT:
mmq_supported = true;
break;
case GGML_TYPE_IQ2_K_R4:
case GGML_TYPE_IQ3_K_R4:
case GGML_TYPE_IQ4_K_R4:
case GGML_TYPE_IQ5_K_R4:
mmq_supported = ne11 < 1024;
break;
default:
mmq_supported = false;
break;
Expand Down