From: Johannes Gäßler <redacted>
Date: Sat, 11 Apr 2026 16:52:11 +0000 (+0200)
Subject: CUDA: skip compilation of superfluous FA kernels (#21768)
X-Git-Tag: upstream/0.0.10438~1675
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=ff5ef8278615a2462b79b50abdf3cc95cfb31c6f;p=pkg%2Fggml%2Fsources%2Fllama.cpp

CUDA: skip compilation of superfluous FA kernels (#21768)
---

diff --git a/ggml/src/ggml-cuda/fattn.cu b/ggml/src/ggml-cuda/fattn.cu
index addf93205..ea6607cd3 100644
--- a/ggml/src/ggml-cuda/fattn.cu
+++ b/ggml/src/ggml-cuda/fattn.cu
@@ -75,13 +75,17 @@ static void ggml_cuda_flash_attn_ext_mma_f16_switch_ncols2(ggml_backend_cuda_con
             return;
         }
 
-        if (use_gqa_opt && gqa_ratio % 2 == 0) {
-            ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 2>(ctx, dst);
+        if constexpr (DKQ <= 256) {
+            if (use_gqa_opt && gqa_ratio % 2 == 0) {
+                ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 2>(ctx, dst);
+                return;
+            }
+
+            ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 1>(ctx, dst);
             return;
+        } else {
+            GGML_ABORT("fatal error");
         }
-
-        ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 1>(ctx, dst);
-        return;
     }
 
     if (use_gqa_opt && gqa_ratio > 4) {
@@ -94,12 +98,16 @@ static void ggml_cuda_flash_attn_ext_mma_f16_switch_ncols2(ggml_backend_cuda_con
         return;
     }
 
-    if (use_gqa_opt && gqa_ratio > 1) {
-        ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 2>(ctx, dst);
-        return;
-    }
+    if constexpr (DKQ <= 256) {
+        if (use_gqa_opt && gqa_ratio > 1) {
+            ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 2>(ctx, dst);
+            return;
+        }
 
-    ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 1>(ctx, dst);
+        ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<DKQ, DV, 1>(ctx, dst);
+    } else {
+        GGML_ABORT("fatal error");
+    }
 }
 
 static void ggml_cuda_flash_attn_ext_mma_f16(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {