replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[FastISel][X86] Fix smul.with.overflow.i8 lowering. 

Add custom lowering code for signed multiply instruction selection, because the
default FastISel instruction selection for ISD::MUL will use unsigned multiply
for the i8 type and signed multiply for all other types. This would set the
incorrect flags for the overflow check.

This fixes <rdar://problem/17549300>

llvm-svn: 212493
This commit is contained in:
Juergen Ributzka 2014-07-07 21:52:21 +00:00
parent 4c5b4054b2
commit 665ea71fcd
2 changed files with 80 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
llvm/lib/Target/X86/X86FastISel.cpp
View File

@ -2402,7 +2402,7 @@ bool X86FastISel::X86VisitIntrinsicCall(const IntrinsicInst &I) {
case Intrinsic::usub_with_overflow:
BaseOpc = ISD::SUB; CondOpc = X86::SETBr; break;
case Intrinsic::smul_with_overflow:
BaseOpc = ISD::MUL; CondOpc = X86::SETOr; break;
BaseOpc = X86ISD::SMUL; CondOpc = X86::SETOr; break;
case Intrinsic::umul_with_overflow:
BaseOpc = X86ISD::UMUL; CondOpc = X86::SETOr; break;
}
@ -2430,7 +2430,8 @@ bool X86FastISel::X86VisitIntrinsicCall(const IntrinsicInst &I) {
RHSIsKill);
}
// FastISel doesn't have a pattern for X86::MUL*r. Emit it manually.
// FastISel doesn't have a pattern for all X86::MUL*r and X86::IMUL*r. Emit
// it manually.
if (BaseOpc == X86ISD::UMUL && !ResultReg) {
static const unsigned MULOpc[] =
{ X86::MUL8r, X86::MUL16r, X86::MUL32r, X86::MUL64r };
@ -2442,6 +2443,21 @@ bool X86FastISel::X86VisitIntrinsicCall(const IntrinsicInst &I) {
.addReg(LHSReg, getKillRegState(LHSIsKill));
ResultReg = FastEmitInst_r(MULOpc[VT.SimpleTy-MVT::i8],
TLI.getRegClassFor(VT), RHSReg, RHSIsKill);
} else if (BaseOpc == X86ISD::SMUL && !ResultReg) {
static const unsigned MULOpc[] =
{ X86::IMUL8r, X86::IMUL16rr, X86::IMUL32rr, X86::IMUL64rr };
if (VT == MVT::i8) {
// Copy the first operand into AL, which is an implicit input to the
// X86::IMUL8r instruction.
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
TII.get(TargetOpcode::COPY), X86::AL)
.addReg(LHSReg, getKillRegState(LHSIsKill));
ResultReg = FastEmitInst_r(MULOpc[0], TLI.getRegClassFor(VT), RHSReg,
RHSIsKill);
} else
ResultReg = FastEmitInst_rr(MULOpc[VT.SimpleTy-MVT::i8],
TLI.getRegClassFor(VT), LHSReg, LHSIsKill,
RHSReg, RHSIsKill);
}
if (!ResultReg)

60
llvm/test/CodeGen/X86/xaluo.ll
View File

@ -261,6 +261,34 @@ entry:
}
; SMULO
define zeroext i1 @smulo.i8(i8 %v1, i8 %v2, i8* %res) {
entry:
; FAST-LABEL: smulo.i8
; FAST: movb %dil, %al
; FAST-NEXT: imulb %sil
; FAST-NEXT: seto %cl
%t = call {i8, i1} @llvm.smul.with.overflow.i8(i8 %v1, i8 %v2)
%val = extractvalue {i8, i1} %t, 0
%obit = extractvalue {i8, i1} %t, 1
store i8 %val, i8* %res
ret i1 %obit
}
define zeroext i1 @smulo.i16(i16 %v1, i16 %v2, i16* %res) {
entry:
; DAG-LABEL: smulo.i16
; DAG: imulw %si, %di
; DAG-NEXT: seto %al
; FAST-LABEL: smulo.i16
; FAST: imulw %si, %di
; FAST-NEXT: seto %al
%t = call {i16, i1} @llvm.smul.with.overflow.i16(i16 %v1, i16 %v2)
%val = extractvalue {i16, i1} %t, 0
%obit = extractvalue {i16, i1} %t, 1
store i16 %val, i16* %res
ret i1 %obit
}
define zeroext i1 @smulo.i32(i32 %v1, i32 %v2, i32* %res) {
entry:
; DAG-LABEL: smulo.i32
@ -292,6 +320,34 @@ entry:
}
; UMULO
define zeroext i1 @umulo.i8(i8 %v1, i8 %v2, i8* %res) {
entry:
; FAST-LABEL: umulo.i8
; FAST: movb %dil, %al
; FAST-NEXT: mulb %sil
; FAST-NEXT: seto %cl
%t = call {i8, i1} @llvm.umul.with.overflow.i8(i8 %v1, i8 %v2)
%val = extractvalue {i8, i1} %t, 0
%obit = extractvalue {i8, i1} %t, 1
store i8 %val, i8* %res
ret i1 %obit
}
define zeroext i1 @umulo.i16(i16 %v1, i16 %v2, i16* %res) {
entry:
; DAG-LABEL: umulo.i16
; DAG: mulw %si
; DAG-NEXT: seto
; FAST-LABEL: umulo.i16
; FAST: mulw %si
; FAST-NEXT: seto
%t = call {i16, i1} @llvm.umul.with.overflow.i16(i16 %v1, i16 %v2)
%val = extractvalue {i16, i1} %t, 0
%obit = extractvalue {i16, i1} %t, 1
store i16 %val, i16* %res
ret i1 %obit
}
define zeroext i1 @umulo.i32(i32 %v1, i32 %v2, i32* %res) {
entry:
; DAG-LABEL: umulo.i32
@ -675,8 +731,12 @@ declare {i32, i1} @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone
declare {i64, i1} @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone
declare {i32, i1} @llvm.usub.with.overflow.i32(i32, i32) nounwind readnone
declare {i64, i1} @llvm.usub.with.overflow.i64(i64, i64) nounwind readnone
declare {i8, i1} @llvm.smul.with.overflow.i8 (i8, i8 ) nounwind readnone
declare {i16, i1} @llvm.smul.with.overflow.i16(i16, i16) nounwind readnone
declare {i32, i1} @llvm.smul.with.overflow.i32(i32, i32) nounwind readnone
declare {i64, i1} @llvm.smul.with.overflow.i64(i64, i64) nounwind readnone
declare {i8, i1} @llvm.umul.with.overflow.i8 (i8, i8 ) nounwind readnone
declare {i16, i1} @llvm.umul.with.overflow.i16(i16, i16) nounwind readnone
declare {i32, i1} @llvm.umul.with.overflow.i32(i32, i32) nounwind readnone
declare {i64, i1} @llvm.umul.with.overflow.i64(i64, i64) nounwind readnone