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

[Phoebe Wang via llvm-commits]

================
@@ -85,10 +85,22 @@ define void @isint_branch(double %d) nounwind {
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:fr64 = COPY $xmm0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vr128 = COPY [[COPY]]
-  ; CHECK-NEXT:   [[CVTTPD2DQrr:%[0-9]+]]:vr128 = nofpexcept CVTTPD2DQrr killed [[COPY1]], implicit $mxcsr
-  ; CHECK-NEXT:   [[CVTDQ2PDrr:%[0-9]+]]:vr128 = CVTDQ2PDrr killed [[CVTTPD2DQrr]]
-  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:fr64 = COPY killed [[CVTDQ2PDrr]]
-  ; CHECK-NEXT:   nofpexcept UCOMISDrr [[COPY]], killed [[COPY2]], implicit-def $eflags, implicit $mxcsr
+  ; CHECK-NEXT:   [[MOVAPDrm:%[0-9]+]]:vr128 = MOVAPDrm $rip, 1, $noreg, %const.0, $noreg :: (load (s128) from constant-pool)
----------------
phoebewang wrote:

Regression?

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