From 505aede139e4959693f1910ad5e103ec869bd330 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Sun, 16 Feb 2025 23:28:18 +0800
Subject: [PATCH] try old wmma

---
 ggml/src/ggml-cuda/fattn.cu | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/ggml/src/ggml-cuda/fattn.cu b/ggml/src/ggml-cuda/fattn.cu
index b0cf152f52cf1..6f2c7509c3f94 100644
--- a/ggml/src/ggml-cuda/fattn.cu
+++ b/ggml/src/ggml-cuda/fattn.cu
@@ -263,10 +263,11 @@ void ggml_cuda_flash_attn_ext(ggml_backend_cuda_context & ctx, ggml_tensor * dst
     }
 
     // The MMA implementation needs Turing or newer, use the old WMMA code for Volta:
-    if (cc == GGML_CUDA_CC_VOLTA) {
-        ggml_cuda_flash_attn_ext_wmma_f16(ctx, dst);
-        return;
-    }
+    ggml_cuda_flash_attn_ext_wmma_f16(ctx, dst);
+    // if (cc == GGML_CUDA_CC_VOLTA) {
+    //     ggml_cuda_flash_attn_ext_wmma_f16(ctx, dst);
+    //     return;
+    // }
 
-    ggml_cuda_flash_attn_ext_mma_f16(ctx, dst);
+    // ggml_cuda_flash_attn_ext_mma_f16(ctx, dst);
 }