Merge branch 'bodhi/smoe+musa-ups' of sh-code.mthreads.com:sw/llama_cpp into bodhi/smoe+musa-ups

Author: Huaishun Hu

ggml/src/ggml-cuda/ggml-cuda.cu

@@ -179,11 +179,11 @@ static ggml_cuda_device_info ggml_cuda_init() {
 #if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
         info.devices[id].smpbo = prop.sharedMemPerBlock;
         info.devices[id].cc = 100*prop.major + 10*prop.minor + GGML_CUDA_CC_OFFSET_AMD;
-#elif defined(GGML_USE_MUSA)
-        /** TODO: MUSA arch should match CUDA 11.4 */
-        info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
-        // info.devices[id].cc = 100*prop.major + 10*prop.minor + CC_OFFSET_MT;
-        info.devices[id].cc = __MUSA_CC__;
+// #elif defined(GGML_USE_MUSA)
+//         /** TODO: MUSA arch should match CUDA 11.4 */
+//         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
+//         // info.devices[id].cc = 100*prop.major + 10*prop.minor + CC_OFFSET_MT;
+//         info.devices[id].cc = __MUSA_CC__;
 #else
         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
         info.devices[id].cc = 100*prop.major + 10*prop.minor;