1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 12:41:49 +01:00

[SystemZ] Make copyPhysReg() add impl-use operands of super reg.

When a 128 bit COPY is lowered into two instructions, an impl-use operand of
the super-reg should be added to each new instruction in case one of the
sub-regs is undefined.

Review: Ulrich Weigand
llvm-svn: 302146
This commit is contained in:
Jonas Paulsson 2017-05-04 13:33:30 +00:00
parent 924c731df4
commit 466bbd4878
2 changed files with 75 additions and 1 deletions

View File

@ -850,12 +850,18 @@ void SystemZInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MBBI,
const DebugLoc &DL, unsigned DestReg,
unsigned SrcReg, bool KillSrc) const {
// Split 128-bit GPR moves into two 64-bit moves. This handles ADDR128 too.
// Split 128-bit GPR moves into two 64-bit moves. Add implicit uses of the
// super register in case one of the subregs is undefined.
// This handles ADDR128 too.
if (SystemZ::GR128BitRegClass.contains(DestReg, SrcReg)) {
copyPhysReg(MBB, MBBI, DL, RI.getSubReg(DestReg, SystemZ::subreg_h64),
RI.getSubReg(SrcReg, SystemZ::subreg_h64), KillSrc);
MachineInstrBuilder(*MBB.getParent(), std::prev(MBBI))
.addReg(SrcReg, RegState::Implicit);
copyPhysReg(MBB, MBBI, DL, RI.getSubReg(DestReg, SystemZ::subreg_l64),
RI.getSubReg(SrcReg, SystemZ::subreg_l64), KillSrc);
MachineInstrBuilder(*MBB.getParent(), std::prev(MBBI))
.addReg(SrcReg, (getKillRegState(KillSrc) | RegState::Implicit));
return;
}

View File

@ -0,0 +1,68 @@
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 -join-liveintervals=false -verify-machineinstrs | FileCheck %s
;
; Check that copyPhysReg() properly adds impl-use operands of the super
; register while lowering a COPY of a GR128 bit reg.
define void @autogen_SD5585(i32*, i64) {
; CHECK: .text
BB:
%L5 = load i1, i1* undef
%I8 = insertelement <8 x i64> undef, i64 %1, i32 3
%I21 = insertelement <8 x i64> zeroinitializer, i64 475435, i32 5
br label %CF290
CF290: ; preds = %CF290, %BB
%B29 = urem <8 x i64> %I8, %I21
%Cmp31 = icmp sge i1 undef, undef
br i1 %Cmp31, label %CF290, label %CF296
CF296: ; preds = %CF290
%FC36 = sitofp <8 x i64> %B29 to <8 x double>
br label %CF302
CF302: ; preds = %CF307, %CF296
%Shuff49 = shufflevector <8 x i64> undef, <8 x i64> zeroinitializer, <8 x i32> <i32 undef, i32 9, i32 11, i32 undef, i32 15, i32 1, i32 3, i32 5>
%L69 = load i16, i16* undef
br label %CF307
CF307: ; preds = %CF302
%Cmp84 = icmp ne i16 undef, %L69
br i1 %Cmp84, label %CF302, label %CF301
CF301: ; preds = %CF307
%B126 = or i32 514315, undef
br label %CF280
CF280: ; preds = %CF280, %CF301
%I139 = insertelement <8 x i64> %Shuff49, i64 undef, i32 2
%B155 = udiv <8 x i64> %I8, %I139
%Cmp157 = icmp ne i64 -1, undef
br i1 %Cmp157, label %CF280, label %CF281
CF281: ; preds = %CF280
%Cmp164 = icmp slt i1 %L5, %Cmp84
br label %CF282
CF282: ; preds = %CF304, %CF281
br label %CF289
CF289: ; preds = %CF289, %CF282
store i32 %B126, i32* %0
%Cmp219 = icmp slt i64 undef, undef
br i1 %Cmp219, label %CF289, label %CF304
CF304: ; preds = %CF289
%Cmp234 = icmp ult i64 0, undef
br i1 %Cmp234, label %CF282, label %CF283
CF283: ; preds = %CF308, %CF283, %CF304
%E251 = extractelement <8 x i64> %B155, i32 0
br i1 undef, label %CF283, label %CF308
CF308: ; preds = %CF283
store i1 %Cmp164, i1* undef
br i1 undef, label %CF283, label %CF293
CF293: ; preds = %CF308
ret void
}