restirct to amd

IMbackK · IMbackK · commit 68a6761e61f3 · 2025-06-23T22:39:43.000+02:00
diff --git a/ggml/src/ggml-cuda/mmv.cu b/ggml/src/ggml-cuda/mmv.cu
@@ -456,7 +456,7 @@ bool ggml_cuda_should_use_mmv(enum ggml_type type, int cc, const int64_t * src0_
                     return ne11 <= 4;
                 }
                 return ne11 <= 3;
-            } else if (fp32_mma_hardware_available(cc)) {
+            } else if (GGML_CUDA_CC_IS_AMD(cc) && fp32_mma_hardware_available(cc)) {
                 return ne11 <= 3;
             }
             return ne11 <= 8;
@@ -470,7 +470,7 @@ bool ggml_cuda_should_use_mmv(enum ggml_type type, int cc, const int64_t * src0_
                     return src0_small && ne11 <= 3;
                 }
                 return ne11 <= 8;
-            } else if (fp16_mma_hardware_available(cc)) {
+            } else if (GGML_CUDA_CC_IS_AMD(cc) && fp16_mma_hardware_available(cc)) {
                 if (GGML_CUDA_CC_IS_RDNA3(cc) || GGML_CUDA_CC_IS_RDNA4(cc)) {
                     return ne11 <= 5;
                 }
@@ -487,7 +487,7 @@ bool ggml_cuda_should_use_mmv(enum ggml_type type, int cc, const int64_t * src0_
                     return src0_small && ne11 <= 3;
                 }
                 return ne11 <= 8;
-            } else if (bf16_mma_hardware_available(cc)) {
+            } else if (GGML_CUDA_CC_IS_AMD(cc) && bf16_mma_hardware_available(cc)) {
                 return ne11 <= 3;
             }
             return ne11 <= 8;

Original file line number	Diff line number	Diff line change
`@@ -456,7 +456,7 @@ bool ggml_cuda_should_use_mmv(enum ggml_type type, int cc, const int64_t * src0_`
`456`	`456`	`return ne11 <= 4;`
`457`	`457`	`}`
`458`	`458`	`return ne11 <= 3;`
`459`		`- } else if (fp32_mma_hardware_available(cc)) {`
	`459`	`+ } else if (GGML_CUDA_CC_IS_AMD(cc) && fp32_mma_hardware_available(cc)) {`
`460`	`460`	`return ne11 <= 3;`
`461`	`461`	`}`
`462`	`462`	`return ne11 <= 8;`
`@@ -470,7 +470,7 @@ bool ggml_cuda_should_use_mmv(enum ggml_type type, int cc, const int64_t * src0_`
`470`	`470`	`return src0_small && ne11 <= 3;`
`471`	`471`	`}`
`472`	`472`	`return ne11 <= 8;`
`473`		`- } else if (fp16_mma_hardware_available(cc)) {`
	`473`	`+ } else if (GGML_CUDA_CC_IS_AMD(cc) && fp16_mma_hardware_available(cc)) {`
`474`	`474`	`if (GGML_CUDA_CC_IS_RDNA3(cc) \|\| GGML_CUDA_CC_IS_RDNA4(cc)) {`
`475`	`475`	`return ne11 <= 5;`
`476`	`476`	`}`
`@@ -487,7 +487,7 @@ bool ggml_cuda_should_use_mmv(enum ggml_type type, int cc, const int64_t * src0_`
`487`	`487`	`return src0_small && ne11 <= 3;`
`488`	`488`	`}`
`489`	`489`	`return ne11 <= 8;`
`490`		`- } else if (bf16_mma_hardware_available(cc)) {`
	`490`	`+ } else if (GGML_CUDA_CC_IS_AMD(cc) && bf16_mma_hardware_available(cc)) {`
`491`	`491`	`return ne11 <= 3;`
`492`	`492`	`}`
`493`	`493`	`return ne11 <= 8;`