[llvm] X86: Support FCANONICALIZE on f64 for i686 with SSE2 or AVX (PR #123917)

[Simon Pilgrim via llvm-commits]

================
@@ -0,0 +1,143 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --default-march x86_64-unknown-linux-gnu --version 5
+; RUN: llc -mattr=+sse2 -mtriple=i686-- < %s | FileCheck %s -check-prefixes=SSE2
+; RUN: llc -mattr=+avx -mtriple=i686-- < %s | FileCheck %s -check-prefixes=AVX
+
+define double @canonicalize_fp64(double %a, double %b) unnamed_addr #0 {
----------------
RKSimon wrote:

drop `unnamed_addr #0` and just use `nounwind` to remove the cfi noise - similar for other tests

https://github.com/llvm/llvm-project/pull/123917