[clang] Fix syntax errors when compiling with clang

[SamGinzburg]

When compiling include/cute/arch/copy_sm100.hpp with clang instead of nvcc the following inline asm causes syntax errors.

errors look like:

cutlass/include/cute/arch/copy_sm100.hpp:335:22: error: expected ')'
  335 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^
cutlass/include/cute/arch/copy_sm100.hpp:354:22: error: expected ')'
  354 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^
cutlass/include/cute/arch/copy_sm100.hpp:372:22: error: expected ')'
  372 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^
cutlass/include/cute/arch/copy_sm100.hpp:390:22: error: expected ')'
  390 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^
cutlass/include/cute/arch/copy_sm100.hpp:410:22: error: expected ')'
  410 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^
cutlass/include/cute/arch/copy_sm100.hpp:429:22: error: expected ')'
  429 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^
cutlass/include/cute/arch/copy_sm100.hpp:448:22: error: expected ')'
  448 |     : "r"(dst_addr)  "l"(src_addr));
      |                      ^

Testing:

I built cutlass with these fixes and don't get compiler errors with clang or nvcc.

[Artem-B]

Interestingly enough, NVCC appears to accept inline asm arguments without the comma separator: https://godbolt.org/z/1MchKsT86

That explains why it went unnoticed till now.