[clang] [llvm] [X86] AVX10_V2_AUX Implementation (PR #206888)

[Evgenii Kudriashov via cfe-commits]

================
@@ -5039,6 +5060,307 @@ let Features = "avx10.2", Attributes = [NoThrow, Const, RequiredVectorWidth<512>
   def vgetmantbf16512_mask : X86Builtin<"_Vector<32, __bf16>(_Vector<32, __bf16>, _Constant int, _Vector<32, __bf16>, unsigned int)">;
 }
 
+// AVX10 V2 AUX - Convert instructions
+
+// Group A: PS(f32) -> i8 truncating conversions (quarter-size: output always v16i8)
----------------
e-kud wrote:

I get it that the comment is based on builtins but it is misleading saying that it is f32->i8 conversion when it is fp32->fp8 conversion.

https://github.com/llvm/llvm-project/pull/206888