Update clang-format

Author: seagater

apps/nccl/src/allreduce.cu

@@ -71,7 +71,7 @@ struct NvlsAdapter {
                           mscclpp::DeviceHandle<mscclpp::SwitchChannel>* nvlsOutChannels, size_t channelInOffset,
                           size_t channelOutOffset, size_t, int rank, int nRanksPerNode, int, size_t nelems,
                           cudaStream_t stream, uint32_t*, uint32_t*, uint32_t*, uint32_t) {
-#if defined(__CUDA_ARCH__) // Skip the __CUDA_ARCH__ < 1000 since FP8 has not been supported for NVLS
+#if defined(__CUDA_ARCH__)  // Skip the __CUDA_ARCH__ < 1000 since FP8 has not been supported for NVLS
     if constexpr (std::is_same_v<T, __fp8_e4m3> || std::is_same_v<T, __fp8_e5m2>) {
       return cudaErrorNotSupported;
     } else
@@ -95,7 +95,7 @@ struct NvlsWithCopyAdapter {
                           mscclpp::DeviceHandle<mscclpp::SwitchChannel>*, size_t, size_t, size_t scratchBufferSize,
                           int rank, int nRanksPerNode, int, size_t nelems, cudaStream_t stream, uint32_t*, uint32_t*,
                           uint32_t*, uint32_t) {
-#if defined(__CUDA_ARCH__) // Skip the __CUDA_ARCH__ < 1000 since FP8 has not been supported for NVLS
+#if defined(__CUDA_ARCH__)  // Skip the __CUDA_ARCH__ < 1000 since FP8 has not been supported for NVLS
     if constexpr (std::is_same_v<T, __fp8_e4m3> || std::is_same_v<T, __fp8_e5m2>) {
       return cudaErrorNotSupported;
     } else