Fix a serious typo in InstCombine's optimization of comparisons.

An unsigned value converted to floating-point will always be greater than
a negative constant.  Unfortunately InstCombine reversed the check so that
unsigned values were being optimized to always be greater than all positive
floating-point constants.  <rdar://problem/12029145>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@161452 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Bob Wilson 2012-08-07 22:35:16 +00:00
parent b64dd5f2b5
commit f12c95a8af
2 changed files with 10 additions and 1 deletions

View File

@ -2824,7 +2824,7 @@ Instruction *InstCombiner::FoldFCmp_IntToFP_Cst(FCmpInst &I,
case ICmpInst::ICMP_UGE:
// (float)int >= -4.4 --> true
// (float)int >= 4.4 --> int > 4
if (!RHS.isNegative())
if (RHS.isNegative())
return ReplaceInstUsesWith(I, ConstantInt::getTrue(I.getContext()));
Pred = ICmpInst::ICMP_UGT;
break;

View File

@ -45,3 +45,12 @@ define i1 @test6(i32 %val) {
ret i1 %2
; CHECK: ret i1 false
}
; Check that optimizing unsigned >= comparisons correctly distinguishes
; positive and negative constants. <rdar://problem/12029145>
define i1 @test7(i32 %val) {
%1 = uitofp i32 %val to double
%2 = fcmp oge double %1, 3.200000e+00
ret i1 %2
; CHECK: icmp ugt i32 %val, 3
}