[llvm] [X86] Use valign instead of vperm for float domain shuffles (PR #201624)

[via llvm-commits]

================
@@ -989,3 +999,8 @@ bb:
   ret void
 }
 
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; AVX512BW: {{.*}}
+; AVX512F: {{.*}}
+; FAST: {{.*}}
+; SLOW: {{.*}}
----------------
ganeshgit wrote:

Done!

https://github.com/llvm/llvm-project/pull/201624