RegScavenger should not exclude undef uses

When computing currently-live registers, the register scavenger excludes undef
uses. As a result, undef uses are ignored when computing the restore points of
registers spilled into the emergency slots. While the register scavenger
normally excludes from consideration, when scavenging, registers used by the
current instruction, we need to not exclude undef uses. Otherwise, we might end
up requiring more emergency spill slots than we have (in the case where the
undef use *is* the currently-spilled register).

Another bug found by llvm-stress.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186067 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Hal Finkel 2013-07-11 05:55:57 +00:00
parent 2ebba647ea
commit 838a7fb1a3
2 changed files with 49 additions and 1 deletions

View File

@ -368,7 +368,7 @@ unsigned RegScavenger::scavengeRegister(const TargetRegisterClass *RC,
// Exclude all the registers being used by the instruction.
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
MachineOperand &MO = I->getOperand(i);
if (MO.isReg() && MO.getReg() != 0 &&
if (MO.isReg() && MO.getReg() != 0 && !(MO.isUse() && MO.isUndef()) &&
!TargetRegisterInfo::isVirtualRegister(MO.getReg()))
Candidates.reset(MO.getReg());
}

View File

@ -0,0 +1,48 @@
; RUN: llc -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr7 < %s
target triple = "powerpc64-unknown-linux-gnu"
define void @autogen_SD156869(i8*, i64*) {
BB:
%A3 = alloca <2 x i1>
%A2 = alloca <8 x i32>
br label %CF
CF: ; preds = %CF85, %CF, %BB
br i1 undef, label %CF, label %CF82.critedge
CF82.critedge: ; preds = %CF
store i8 -59, i8* %0
br label %CF82
CF82: ; preds = %CF82, %CF82.critedge
%L17 = load i8* %0
%E18 = extractelement <2 x i64> undef, i32 0
%PC = bitcast <2 x i1>* %A3 to i64*
br i1 undef, label %CF82, label %CF84.critedge
CF84.critedge: ; preds = %CF82
store i64 455385, i64* %PC
br label %CF84
CF84: ; preds = %CF84, %CF84.critedge
%L40 = load i64* %PC
store i64 -1, i64* %PC
%Sl46 = select i1 undef, i1 undef, i1 false
br i1 %Sl46, label %CF84, label %CF85
CF85: ; preds = %CF84
%L47 = load i64* %PC
store i64 %E18, i64* %PC
%PC52 = bitcast <8 x i32>* %A2 to ppc_fp128*
store ppc_fp128 0xM4D436562A0416DE00000000000000000, ppc_fp128* %PC52
%PC59 = bitcast i64* %1 to i8*
%Cmp61 = icmp slt i64 %L47, %L40
br i1 %Cmp61, label %CF, label %CF77
CF77: ; preds = %CF77, %CF85
br i1 undef, label %CF77, label %CF81
CF81: ; preds = %CF77
store i8 %L17, i8* %PC59
ret void
}