[PATCH] D50222: [CodeGen] [SelectionDAG] More efficient code for X % C == 0 (UREM case)

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added inline comments.


================
Comment at: test/CodeGen/X86/jump_sign.ll:239-243
+; CHECK-NEXT:    imull $-13107, %eax, %eax
+; CHECK-NEXT:    rorw $1, %ax
+; CHECK-NEXT:    movzwl	%ax, %eax
+; CHECK-NEXT:    cmpl	$13108, %eax
+; CHECK-NEXT:    jae	.LBB12_5
----------------
What do you mean by breaks? Is this a miscompilation?


================
Comment at: test/CodeGen/X86/urem-seteq.ll:4
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s --check-prefixes=CHECK,X64,NOBMI2
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+bmi2 < %s | FileCheck %s --check-prefixes=CHECK,X64,BMI2
+
----------------
lebedev.ri wrote:
> Actually, looking at the check-lines, i'm not sure we want to check the bmi2 version.
> I'm not seeing anything there that would benefit from anything above baseline, i think.
I've meant to drop this check-line, but apparently forgot.
I'm not 100% sure if we want/don't want it.


Repository:
  rL LLVM

https://reviews.llvm.org/D50222