Performance tune for gemma4-26b-a4b flash attention shape. (llama/28450)

This commit is contained in:
David M. Rogers
2026-09-23 20:46:47 +03:00
committed by Georgi Gerganov
parent dd67678efa
commit 0409ed2628
+4
View File
@@ -1173,6 +1173,10 @@ static void launch_fattn_tile_switch_ncols2(ggml_backend_sycl_context & ctx, ggm
launch_fattn_tile_switch_ncols1<DKQ, DV, 16, use_logit_softcap>(ctx, dst);
return;
}
if (use_gqa_opt && gqa_ratio % 8 == 0) {
launch_fattn_tile_switch_ncols1<DKQ, DV, 8, use_logit_softcap>(ctx, dst);
return;
}
if (use_gqa_opt && gqa_ratio % 4 == 0) {
launch_fattn_tile_switch_ncols1<DKQ, DV, 4, use_logit_softcap>(ctx, dst);
return;