[clang] [Clang] Support constexpr for AVX512 compress intrinsics (PR #187656)

[Zihao Wang via cfe-commits]

================
@@ -3550,6 +3550,38 @@ static bool interp__builtin_ia32_vpconflict(InterpState &S, CodePtr OpPC,
   return true;
 }
 
+static bool interp__builtin_ia32_compress(InterpState &S, CodePtr OpPC,
----------------
hhy3 wrote:

I think it may not be trivial to convert compress to shuffle_generic. Compress is a data-dependent compaction (output indices depend on the popcount of prior mask bits), not a stateless per-element permutation. What's your opinion?

https://github.com/llvm/llvm-project/pull/187656