cleanup unused comments

Author: Huaishun Hu

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

@@ -263,9 +263,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
                       device_vmm ? "yes" : "no", prop.warpSize);
 #elif defined(GGML_USE_MUSA)
         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
-        /** FIXME: MUSA arch should match CUDA 11.4 */
         info.devices[id].cc = 100*prop.major + 10*prop.minor;
-        // info.devices[id].cc = __MUSA_CC__;
 #else
         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
         info.devices[id].cc = 100*prop.major + 10*prop.minor;

ggml/src/ggml-cuda/vendors/musa.h

@@ -135,37 +135,3 @@
 #define cudaStreamEndCapture musaStreamEndCapture
 
 typedef mt_bfloat16 nv_bfloat16;
-
-/** FIXME: MUSA arch should match CUDA 11.4 */
-// #define GGML_CUDA_CC_PASCAL     600
-// #define GGML_CUDA_CC_DP4A       610
-// #define GGML_CUDA_CC_VOLTA      700
-// #define GGML_CUDA_CC_TURING     750
-// #define GGML_CUDA_CC_AMPERE     800
-
-#define __MUSA_CC__   610
-// #define __CUDA_ARCH__ __MUSA_CC__
-
-
-/** TODO: following apis not supported yet by musa sdk: ***********/
-
-// __device__ __half hexp(const __half a) {
-//     float f_a = __half2float(a);
-//     float f_result = expf(f_a);
-//     return __float2half(f_result);
-// }
-
-// __host__ __device__ __half2 h2exp(const __half2 a) {
-//     // Extract lower and upper halves
-//     __half lower = __low2half(a);
-//     __half upper = __high2half(a);
-
-//     // Compute exp for each half
-//     __half exp_lower = hexp(lower);
-//     __half exp_upper = hexp(upper);
-
-//     // Combine back into __half2
-//     return __halves2half2(exp_lower, exp_upper);
-// }
-
-/******************************************************************/