[llvm] [X86] Faster truncf and roundf on x86 SSE2 (PR #226513)

[Phoebe Wang via llvm-commits]

================
@@ -63,19 +74,41 @@ define i32 @isint_float_return(float %f) nounwind {
 ; X86-LABEL: isint_float_return:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-NEXT:    cvttps2dq %xmm0, %xmm1
-; X86-NEXT:    cvtdq2ps %xmm1, %xmm1
+; X86-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN,NaN,NaN]
----------------
phoebewang wrote:

Regressions in this file?

https://github.com/llvm/llvm-project/pull/226513