RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-12-17 00:50:58 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

InstCombine: Fold X-C1 <u 2 -> (X & -2) == C1

Browse Source 
Back in r179493 we determined that two transforms collided with each
other.  The fix back then was to reorder the transforms so that the
preferred transform would give it a try and then we would try the
secondary transform.  However, it was noted that the best approach would
canonicalize one transform into the other, removing the collision and
allowing us to optimize IR given to us in that form.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185808 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
David Majnemer 2013-07-08 11:53:08 +00:00
parent 4010110ccf
commit 53fc39992d
2 changed files with 18 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
lib/Transforms/InstCombine/InstCombineCompares.cpp
View File

@ -1539,6 +1539,14 @@ Instruction *InstCombiner::visitICmpInstWithInstAndIntCst(ICmpInst &ICI,
Builder->getInt(CR.getLower()));
}
}
// X-C1 <u 2 -> (X & -2) == C1
// iff C1 & 1 == 0
if (ICI.getPredicate() == ICmpInst::ICMP_ULT && LHSI->hasOneUse() &&
LHSV[0] == 0 && RHSV == 2)
return new ICmpInst(ICmpInst::ICMP_EQ,
Builder->CreateAnd(LHSI->getOperand(0), -RHSV),
ConstantExpr::getNeg(LHSC));
}
break;
}

10
test/Transforms/InstCombine/icmp.ll
View File

@ -1113,3 +1113,13 @@ define i1 @or_icmp_eq_B_0_icmp_ult_A_B(i64 %a, i64 %b) {
%3 = or i1 %1, %2
ret i1 %3
}
; CHECK: @icmp_add_ult_2
; CHECK-NEXT: [[AND:%[a-z0-9]+]] = and i32 %X, -2
; CHECK-NEXT: [[CMP:%[a-z0-9]+]] = icmp eq i32 [[AND]], 14
; CHECK-NEXT: ret i1 [[CMP]]
define i1 @icmp_add_ult_2(i32 %X) {
%add = add i32 %X, -14
%cmp = icmp ult i32 %add, 2
ret i1 %cmp
}