MUSA: support ARM64 and enable __dp4a .etc

ggml-org · Feb 20, 2025 · 80a3000 · 80a3000
1 parent d04e716
commit 80a3000
Show file tree

Hide file tree

Showing 7 changed files with 18 additions and 11 deletions.
diff --git a/Makefile b/Makefile
@@ -847,7 +847,7 @@ ifdef GGML_MUSA
 	CXX := $(MUSA_PATH)/bin/clang++
 	MCC := $(CCACHE) $(MUSA_PATH)/bin/mcc
 
-	MUSAFLAGS  = -x musa -mtgpu
+	MUSAFLAGS  = -fsigned-char -x musa -mtgpu
 	MUSAFLAGS += $(foreach arch,$(subst ;, ,$(MUSA_ARCHITECTURES)),--cuda-gpu-arch=mp_$(arch))
 
 ifdef GGML_CUDA_FORCE_MMQ

diff --git a/docs/build.md b/docs/build.md
@@ -205,6 +205,13 @@ This provides GPU acceleration using the MUSA cores of your Moore Threads MTT GP
   cmake -B build -DGGML_MUSA=ON
   cmake --build build --config Release
   ```
+- For static build:
+
+  ```bash
+  cmake -B build -DGGML_MUSA=ON \
+    -DBUILD_SHARED_LIBS=OFF -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+  cmake --build build --config Release
+  ```
 
 The environment variable [`MUSA_VISIBLE_DEVICES`](https://docs.mthreads.com/musa-sdk/musa-sdk-doc-online/programming_guide/Z%E9%99%84%E5%BD%95/) can be used to specify which GPU(s) will be used.
 

diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh
@@ -411,13 +411,13 @@ static __device__ __forceinline__ int ggml_cuda_dp4a(const int a, const int b, i
 
 #else // defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
 
-#if __CUDA_ARCH__ >= GGML_CUDA_CC_DP4A
+#if __CUDA_ARCH__ >= GGML_CUDA_CC_DP4A || defined(GGML_USE_MUSA)
     return __dp4a(a, b, c);
-#else // __CUDA_ARCH__ >= GGML_CUDA_CC_DP4A
+#else // __CUDA_ARCH__ >= GGML_CUDA_CC_DP4A || defined(GGML_USE_MUSA)
     const int8_t * a8 = (const int8_t *) &a;
     const int8_t * b8 = (const int8_t *) &b;
     return c + a8[0]*b8[0] + a8[1]*b8[1] + a8[2]*b8[2] + a8[3]*b8[3];
-#endif // __CUDA_ARCH__ >= GGML_CUDA_CC_DP4A
+#endif // __CUDA_ARCH__ >= GGML_CUDA_CC_DP4A || defined(GGML_USE_MUSA)
 
 #endif // defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
 }

diff --git a/ggml/src/ggml-cuda/cross-entropy-loss.cu b/ggml/src/ggml-cuda/cross-entropy-loss.cu
@@ -126,7 +126,7 @@ void ggml_cuda_cross_entropy_loss(ggml_backend_cuda_context & ctx, ggml_tensor *
 #if !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__))
         static bool shared_memory_limit_raised[GGML_CUDA_MAX_DEVICES] = {false};
         if (!shared_memory_limit_raised[id]) {
-            CUDA_CHECK(cudaFuncSetAttribute(cross_entropy_loss_back_f32<true>, cudaFuncAttributeMaxDynamicSharedMemorySize, smpbo));
+            CUDA_CHECK(cudaFuncSetAttribute(cross_entropy_loss_f32<true>, cudaFuncAttributeMaxDynamicSharedMemorySize, smpbo));
             shared_memory_limit_raised[id] = true;
         }
 #endif // !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__))

diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -261,6 +261,10 @@ static ggml_cuda_device_info ggml_cuda_init() {
         GGML_LOG_INFO("  Device %d: %s, %s (0x%x), VMM: %s, Wave Size: %d\n",
                       id, prop.name, prop.gcnArchName, info.devices[id].cc & 0xffff,
                       device_vmm ? "yes" : "no", prop.warpSize);
+#elif defined(GGML_USE_MUSA)
+        // NOTE: MUSA will reserve some shared mem, and 24B should be enough
+        info.devices[id].smpbo = prop.sharedMemPerBlockOptin - 24;
+        info.devices[id].cc = 100*prop.major + 10*prop.minor;
 #else
         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
         info.devices[id].cc = 100*prop.major + 10*prop.minor;
@@ -1782,9 +1786,6 @@ static void ggml_cuda_mul_mat_batched_cublas(ggml_backend_cuda_context & ctx, co
         }
     }
 #else
-#ifdef GGML_USE_MUSA
-    GGML_ASSERT(false);
-#else // !GGML_USE_MUSA
     if (r2 == 1 && r3 == 1 && ggml_is_contiguous_2(src0) && ggml_is_contiguous_2(src1)) {
         // there is no broadcast and src0, src1 are contiguous across dims 2, 3
         // use cublasGemmStridedBatchedEx
@@ -1827,7 +1828,6 @@ static void ggml_cuda_mul_mat_batched_cublas(ggml_backend_cuda_context & ctx, co
                 cu_compute_type,
                 CUBLAS_GEMM_DEFAULT_TENSOR_OP));
     }
-#endif // GGML_USE_MUSA
 #endif
 
     if (dst->op_params[0] == GGML_PREC_DEFAULT) {

diff --git a/ggml/src/ggml-impl.h b/ggml/src/ggml-impl.h
@@ -16,7 +16,7 @@
 #include <arm_sve.h>
 #endif // __ARM_FEATURE_SVE
 
-#if defined(__ARM_NEON) && !defined(__CUDACC__)
+#if defined(__ARM_NEON) && !defined(__CUDACC__) && !defined(__MUSACC__)
 // if YCM cannot find <arm_neon.h>, make a symbolic link to it, for example:
 //
 //   $ ln -sfn /Library/Developer/CommandLineTools/usr/lib/clang/13.1.6/include/arm_neon.h ./src/

diff --git a/ggml/src/ggml-musa/CMakeLists.txt b/ggml/src/ggml-musa/CMakeLists.txt
@@ -49,7 +49,7 @@ if (MUSAToolkit_FOUND)
 
     set_source_files_properties(${GGML_SOURCES_MUSA} PROPERTIES LANGUAGE CXX)
     foreach(SOURCE ${GGML_SOURCES_MUSA})
-        set(COMPILE_FLAGS "-x musa -mtgpu")
+        set(COMPILE_FLAGS "-fsigned-char -x musa -mtgpu")
         foreach(ARCH ${MUSA_ARCHITECTURES})
             set(COMPILE_FLAGS "${COMPILE_FLAGS} --cuda-gpu-arch=mp_${ARCH}")
         endforeach()