RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 18:54:02 +01:00
Code Issues Projects Releases Wiki Activity

[TargetLowering] optimizeSetCCToComparisonWithZero(): add extra sanity checks (PR43769)

Browse Source 
We should do the fold only if both constants are plain,
non-opaque constants, at least that is the DAG.FoldConstantArithmetic()
requirement.
And if the constant we are comparing with is zero - we shouldn't be
trying to do this fold in the first place.

Fixes https://bugs.llvm.org/show_bug.cgi?id=43769
This commit is contained in:
Roman Lebedev 2019-10-23 11:58:50 +03:00
parent 97ade712c2
commit c3b9f03b45
2 changed files with 60 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
lib/CodeGen/SelectionDAG/TargetLowering.cpp
View File

@ -3052,6 +3052,10 @@ SDValue TargetLowering::optimizeSetCCToComparisonWithZero(
assert((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
"Only for equality-comparisons.");
// The constant we are comparing with must be a non-zero, non-opaque constant.
if (N1C->isNullValue() || N1C->isOpaque())
return SDValue();
// LHS should not be used elsewhere, to avoid creating an extra node.
if (!N0.hasOneUse())
return SDValue();
@ -3072,9 +3076,9 @@ SDValue TargetLowering::optimizeSetCCToComparisonWithZero(
break;
}
// Second operand must be a constant.
// Second operand must be a non-opaque constant.
ConstantSDNode *N01C = isConstOrConstSplat(N0.getOperand(1));
if (!N01C)
if (!N01C || N01C->isOpaque())
return SDValue();
// And let's be even more specific for now, it must be a zero constant.

54
test/CodeGen/X86/pr43769.ll Normal file
View File

@ -0,0 +1,54 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=i686-w64-windows-gnu %s -o - | FileCheck %s
; Reduced from https://bugs.llvm.org/show_bug.cgi?id=43769
define i32 @b(i32 %a, i32* %c, i32 %d) {
; CHECK-LABEL: b:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: cmpl $0, {{[0-9]+}}(%esp)
; CHECK-NEXT: je LBB0_4
; CHECK-NEXT: # %bb.1: # %for.body.lr.ph
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-NEXT: movl %eax, %ecx
; CHECK-NEXT: sarl $31, %ecx
; CHECK-NEXT: addl $-2147483647, %eax # imm = 0x80000001
; CHECK-NEXT: adcl $0, %ecx
; CHECK-NEXT: jne LBB0_4
; CHECK-NEXT: # %bb.2: # %for.body.lr.ph.peel.newph
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %ecx
; CHECK-NEXT: movl $-2147483647, %edx # imm = 0x80000001
; CHECK-NEXT: movl %ecx, %eax
; CHECK-NEXT: sarl $31, %eax
; CHECK-NEXT: addl %ecx, %edx
; CHECK-NEXT: adcl $0, %eax
; CHECK-NEXT: .p2align 4, 0x90
; CHECK-NEXT: LBB0_3: # %for.body
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: testl %eax, %eax
; CHECK-NEXT: je LBB0_3
; CHECK-NEXT: LBB0_4: # %for.end
; CHECK-NEXT: retl
entry:
%tobool3 = icmp eq i32 %a, 0
br i1 %tobool3, label %for.end, label %for.body.lr.ph
for.body.lr.ph: ; preds = %entry
%0 = ptrtoint i32* %c to i32
%conv.peel = sext i32 %d to i64
%add.peel = add nsw i64 %conv.peel, 2147483649
%tobool1.peel = icmp ugt i64 %add.peel, 4294967295
br i1 %tobool1.peel, label %for.end, label %for.body.lr.ph.peel.newph
for.body.lr.ph.peel.newph: ; preds = %for.body.lr.ph
%conv = sext i32 %0 to i64
%add = add nsw i64 %conv, 2147483649
%tobool1 = icmp ugt i64 %add, 4294967295
br label %for.body
for.body: ; preds = %for.body, %for.body.lr.ph.peel.newph
br i1 %tobool1, label %for.end, label %for.body
for.end: ; preds = %for.body.lr.ph, %for.body, %entry
ret i32 undef
}