Skip to content

Commit

Permalink
[NVPTX] Handle signext/zeroext attributes properly
Browse files Browse the repository at this point in the history 
Fix a case where we were incorrectly sign-extending a value when we should have been zero-extending the value.

Also change some SIGN_EXTEND to ANY_EXTEND because we really dont care and may have more opportunity to fold subexpressions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185331 91177308-0d34-0410-b5e6-96231b3b80d8
  • Loading branch information
@jholewinski
jholewinski committed Jul 1, 2013
1 parent 8834184 commit 03e5bb2
Show file tree
Hide file tree
Showing 2 changed files with 41 additions and 19 deletions.
44 changes: 25 additions & 19 deletions lib/Target/NVPTX/NVPTXISelLowering.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -713,7 +713,7 @@ SDValue NVPTXTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
sz = 8;
SDValue StVal = OutVals[OIdx];
if (elemtype.getSizeInBits() < 16) {
StVal = DAG.getNode(ISD::SIGN_EXTEND, dl, MVT::i16, StVal);
StVal = DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i16, StVal);
}
SDVTList CopyParamVTs = DAG.getVTList(MVT::Other, MVT::Glue);
SDValue CopyParamOps[] = { Chain,
Expand Down Expand Up @@ -947,7 +947,7 @@ SDValue NVPTXTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
MachinePointerInfo(), false, false, false,
0);
if (elemtype.getSizeInBits() < 16) {
theVal = DAG.getNode(ISD::SIGN_EXTEND, dl, MVT::i16, theVal);
theVal = DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i16, theVal);
}
SDVTList CopyParamVTs = DAG.getVTList(MVT::Other, MVT::Glue);
SDValue CopyParamOps[] = { Chain, DAG.getConstant(paramCount, MVT::i32),
Expand Down Expand Up @@ -1382,9 +1382,9 @@ NVPTXTargetLowering::LowerSTOREVector(SDValue Op, SelectionDAG &DAG) const {
// Since StoreV2 is a target node, we cannot rely on DAG type legalization.
// Therefore, we must ensure the type is legal. For i1 and i8, we set the
// stored type to i16 and propogate the "real" type as the memory type.
bool NeedSExt = false;
bool NeedExt = false;
if (EltVT.getSizeInBits() < 16)
NeedSExt = true;
NeedExt = true;

switch (NumElts) {
default:
Expand All @@ -1407,8 +1407,8 @@ NVPTXTargetLowering::LowerSTOREVector(SDValue Op, SelectionDAG &DAG) const {
for (unsigned i = 0; i < NumElts; ++i) {
SDValue ExtVal = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, EltVT, Val,
DAG.getIntPtrConstant(i));
if (NeedSExt)
ExtVal = DAG.getNode(ISD::SIGN_EXTEND, DL, MVT::i16, ExtVal);
if (NeedExt)
ExtVal = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i16, ExtVal);
Ops.push_back(ExtVal);
}

Expand Down Expand Up @@ -1614,15 +1614,18 @@ SDValue NVPTXTargetLowering::LowerFormalArguments(
aggregateIsPacked ? 1
: TD->getABITypeAlignment(
partVT.getTypeForEVT(F->getContext()));
SDValue p;
if (Ins[InsIdx].VT.getSizeInBits() > partVT.getSizeInBits())
p = DAG.getExtLoad(ISD::SEXTLOAD, dl, Ins[InsIdx].VT, Root, srcAddr,
SDValue p;
if (Ins[InsIdx].VT.getSizeInBits() > partVT.getSizeInBits()) {
ISD::LoadExtType ExtOp = Ins[InsIdx].Flags.isSExt() ?
ISD::SEXTLOAD : ISD::ZEXTLOAD;
p = DAG.getExtLoad(ExtOp, dl, Ins[InsIdx].VT, Root, srcAddr,
MachinePointerInfo(srcValue), partVT, false,
false, partAlign);
else
} else {
p = DAG.getLoad(partVT, dl, Root, srcAddr,
MachinePointerInfo(srcValue), false, false, false,
partAlign);
}
if (p.getNode())
p.getNode()->setIROrder(idx + 1);
InVals.push_back(p);
Expand Down Expand Up @@ -1657,7 +1660,7 @@ SDValue NVPTXTargetLowering::LowerFormalArguments(
P.getNode()->setIROrder(idx + 1);

if (Ins[InsIdx].VT.getSizeInBits() > EltVT.getSizeInBits())
P = DAG.getNode(ISD::SIGN_EXTEND, dl, Ins[InsIdx].VT, P);
P = DAG.getNode(ISD::ANY_EXTEND, dl, Ins[InsIdx].VT, P);
InVals.push_back(P);
Ofst += TD->getTypeAllocSize(EltVT.getTypeForEVT(F->getContext()));
++InsIdx;
Expand All @@ -1682,8 +1685,8 @@ SDValue NVPTXTargetLowering::LowerFormalArguments(
DAG.getIntPtrConstant(1));

if (Ins[InsIdx].VT.getSizeInBits() > EltVT.getSizeInBits()) {
Elt0 = DAG.getNode(ISD::SIGN_EXTEND, dl, Ins[InsIdx].VT, Elt0);
Elt1 = DAG.getNode(ISD::SIGN_EXTEND, dl, Ins[InsIdx].VT, Elt1);
Elt0 = DAG.getNode(ISD::ANY_EXTEND, dl, Ins[InsIdx].VT, Elt0);
Elt1 = DAG.getNode(ISD::ANY_EXTEND, dl, Ins[InsIdx].VT, Elt1);
}

InVals.push_back(Elt0);
Expand Down Expand Up @@ -1726,7 +1729,7 @@ SDValue NVPTXTargetLowering::LowerFormalArguments(
SDValue Elt = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, EltVT, P,
DAG.getIntPtrConstant(j));
if (Ins[InsIdx].VT.getSizeInBits() > EltVT.getSizeInBits())
Elt = DAG.getNode(ISD::SIGN_EXTEND, dl, Ins[InsIdx].VT, Elt);
Elt = DAG.getNode(ISD::ANY_EXTEND, dl, Ins[InsIdx].VT, Elt);
InVals.push_back(Elt);
}
Ofst += TD->getTypeAllocSize(VecVT.getTypeForEVT(F->getContext()));
Expand All @@ -1745,14 +1748,17 @@ SDValue NVPTXTargetLowering::LowerFormalArguments(
Value *srcValue = Constant::getNullValue(PointerType::get(
ObjectVT.getTypeForEVT(F->getContext()), llvm::ADDRESS_SPACE_PARAM));
SDValue p;
if (ObjectVT.getSizeInBits() < Ins[InsIdx].VT.getSizeInBits())
p = DAG.getExtLoad(ISD::SEXTLOAD, dl, Ins[InsIdx].VT, Root, Arg,
if (ObjectVT.getSizeInBits() < Ins[InsIdx].VT.getSizeInBits()) {
ISD::LoadExtType ExtOp = Ins[InsIdx].Flags.isSExt() ?
ISD::SEXTLOAD : ISD::ZEXTLOAD;
p = DAG.getExtLoad(ExtOp, dl, Ins[InsIdx].VT, Root, Arg,
MachinePointerInfo(srcValue), ObjectVT, false, false,
TD->getABITypeAlignment(ObjectVT.getTypeForEVT(F->getContext())));
else
TD->getABITypeAlignment(ObjectVT.getTypeForEVT(F->getContext())));
} else {
p = DAG.getLoad(Ins[InsIdx].VT, dl, Root, Arg,
MachinePointerInfo(srcValue), false, false, false,
TD->getABITypeAlignment(ObjectVT.getTypeForEVT(F->getContext())));
TD->getABITypeAlignment(ObjectVT.getTypeForEVT(F->getContext())));
}
if (p.getNode())
p.getNode()->setIROrder(idx + 1);
InVals.push_back(p);
Expand Down
16 changes: 16 additions & 0 deletions test/CodeGen/NVPTX/sext-params.ll
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@
; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s

target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64"


define i8 @foo(i8 signext %a) {
; CHECK: ld.param.s8
%ret = add i8 %a, 3
ret i8 %ret
}

define i8 @bar(i8 zeroext %a) {
; CHECK: ld.param.u8
%ret = add i8 %a, 3
ret i8 %ret
}

0 comments on commit 03e5bb2

Please sign in to comment.