From: Johannes Gäßler <redacted>
Date: Tue, 27 Jan 2026 13:28:56 +0000 (+0100)
Subject: CUDA: tune GLM 4.7 Flash FA kernel selection logic (llama/19097)
X-Git-Tag: v0.9.6~24
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=0dff0d5a9cd6ffca1b54ad17a0194e21e1e676a0;p=pkg%2Fggml%2Fsources%2Fggml

CUDA: tune GLM 4.7 Flash FA kernel selection logic (llama/19097)
---

diff --git a/src/ggml-cuda/fattn.cu b/src/ggml-cuda/fattn.cu
index 2f5dbd13..b061fdf9 100644
--- a/src/ggml-cuda/fattn.cu
+++ b/src/ggml-cuda/fattn.cu
@@ -148,6 +148,10 @@ static void ggml_cuda_flash_attn_ext_mma_f16(ggml_backend_cuda_context & ctx, gg
             const int gqa_ratio = Q->ne[2] / K->ne[2];
             if (gqa_ratio == 20) { // GLM 4.7 Flash
                 if (cc >= GGML_CUDA_CC_BLACKWELL) {
+                    if (Q->ne[1] <= 4 && K->ne[1] >= 65536) {
+                        ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 16>(ctx, dst);
+                        break;
+                    }
                     ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 4>(ctx, dst);
                     break;
                 }
@@ -161,6 +165,10 @@ static void ggml_cuda_flash_attn_ext_mma_f16(ggml_backend_cuda_context & ctx, gg
                 }
                 if (cc >= GGML_CUDA_CC_TURING) {
                     if (Q->ne[1] <= 4) {
+                        if (K->ne[1] <= 16384) {
+                            ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 16>(ctx, dst);
+                            break;
+                        }
                         ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 32>(ctx, dst);
                         break;
                     }