clean up after rebase: remove duplicated change, revert cmake files

Author: jiachengjason

ggml/CMakeLists.txt

@@ -210,7 +210,6 @@ option(GGML_HIP_GRAPHS                      "ggml: use HIP graph, experimental,
 option(GGML_HIP_NO_VMM                      "ggml: do not try to use HIP VMM"                 ON)
 option(GGML_HIP_ROCWMMA_FATTN               "ggml: enable rocWMMA for FlashAttention"         OFF)
 option(GGML_HIP_MMQ_MFMA                    "ggml: enable MFMA MMA for CDNA in MMQ"           ON)
-option(GGML_HIP_MMQ_WMMA                    "ggml: enable WMMA MMA for RDNA4 in MMQ"           ON)
 option(GGML_HIP_EXPORT_METRICS              "ggml: enable kernel perf metrics output"         OFF)
 option(GGML_MUSA_GRAPHS                     "ggml: use MUSA graph, experimental, unstable"    OFF)
 option(GGML_MUSA_MUDNN_COPY                 "ggml: enable muDNN for accelerated copy"         OFF)

ggml/src/ggml-cuda/common.cuh

@@ -232,9 +232,6 @@ static const char * cu_get_error_str(CUresult err) {
 #if !defined(GGML_USE_HIP) && __CUDA_ARCH__ == GGML_CUDA_CC_VOLTA
 #define VOLTA_MMA_AVAILABLE
 #endif // !defined(GGML_USE_HIP) && __CUDA_ARCH__ == GGML_CUDA_CC_VOLTA
-#if defined(GGML_USE_HIP) && defined(RDNA4)
-#define AMD_WMMA_AVAILABLE
-#endif // defined(GGML_USE_HIP) && defined(RDNA4) && !defined(GGML_HIP_NO_MMQ_WMMA)
 
 #if !defined(GGML_USE_HIP) && __CUDA_ARCH__ >= GGML_CUDA_CC_TURING
 #define TURING_MMA_AVAILABLE
@@ -298,10 +295,6 @@ static bool volta_mma_available(const int cc) {
     return GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) == GGML_CUDA_CC_VOLTA;
 }
 
-static bool amd_wmma_available(const int cc) {
-    return GGML_CUDA_CC_IS_RDNA4(cc);
-}
-
 static bool turing_mma_available(const int cc) {
     return GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_TURING;
 }

ggml/src/ggml-cuda/mma.cuh

@@ -437,7 +437,20 @@ namespace ggml_cuda_mma {
             xi[0] = xs[0];
         }
 #elif defined(AMD_WMMA_AVAILABLE)
-        ggml_cuda_memcpy_1<sizeof(t.x)>(t.x, xs0 + t.get_i(0) * stride + t.get_j(0));
+        if constexpr (I == 16 && J == 4) {
+            int64_t * xi = (int64_t *) t.x;
+            const int64_t * xs = (int64_t *) ((const int *) xs0 + (threadIdx.x % t.I) * stride + 2 * (threadIdx.x / t.I));
+            xi[0] = xs[0];
+        }else if constexpr (I == 16 && J == 8) {
+            int64_t * xi = (int64_t *) t.x;
+            const int64_t * xs = (int64_t *) ((const int *) xs0 + (threadIdx.x % t.I) * stride + 4 * (threadIdx.x / t.I));
+            xi[0] = xs[0];
+
+            const int64_t * xs1 = (int64_t *) ((const int *) xs0 + (threadIdx.x % t.I) * stride + 4 * (threadIdx.x / t.I) + 2);
+            xi[1] = xs1[0];
+        }else{
+            NO_DEVICE_CODE;
+        }
 #else
 #pragma unroll
         for (int l = 0; l < t.ne; ++l) {