RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 10:42:39 +01:00
Code Issues Projects Releases Wiki Activity

Remove extra truncs/exts around i32 bit operations on PPC64

Browse Source 
This generalizes the code to eliminate extra truncs/exts around i1 bit
operations to also do the same on PPC64 for i32 bit operations. This eliminates
a fairly prevalent code wart:

int foo(int a) {
  return a == 5 ? 7 : 8;
}

On PPC64, because of the extension implied by the ABI, this would generate:

	cmplwi 0, 3, 5
	li 12, 8
	li 4, 7
	isel 3, 4, 12, 2
	rldicl 3, 3, 0, 32
	blr

where the 'rldicl 3, 3, 0, 32', the extension, is completely unnecessary. At
least for the single-BB case (which is all that the DAG combine mechanism can
handle), this unnecessary extension is no longer generated.

llvm-svn: 202600
This commit is contained in:
Hal Finkel 2014-03-01 21:36:57 +00:00
parent 44a4d4b894
commit 4937443651
2 changed files with 112 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

88
lib/Target/PowerPC/PPCISelLowering.cpp
View File

@ -563,10 +563,11 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM)
setTargetDAGCombine(ISD::BSWAP);
setTargetDAGCombine(ISD::INTRINSIC_WO_CHAIN);
if (Subtarget->useCRBits()) {
setTargetDAGCombine(ISD::SIGN_EXTEND);
setTargetDAGCombine(ISD::ZERO_EXTEND);
setTargetDAGCombine(ISD::ANY_EXTEND);
if (Subtarget->useCRBits()) {
setTargetDAGCombine(ISD::TRUNCATE);
setTargetDAGCombine(ISD::SETCC);
setTargetDAGCombine(ISD::SELECT_CC);
@ -7294,6 +7295,20 @@ SDValue PPCTargetLowering::DAGCombineTruncBoolExt(SDNode *N,
SDNode *User = *UI;
if (User != N && !Visited.count(User))
return SDValue();
// Make sure that we're not going to promote the non-output-value
// operand(s) or SELECT or SELECT_CC.
// FIXME: Although we could sometimes handle this, and it does occur in
// practice that one of the condition inputs to the select is also one of
// the outputs, we currently can't deal with this.
if (User->getOpcode() == ISD::SELECT) {
if (User->getOperand(0) == Inputs[i])
return SDValue();
} else if (User->getOpcode() == ISD::SELECT_CC) {
if (User->getOperand(0) == Inputs[i] ||
User->getOperand(1) == Inputs[i])
return SDValue();
}
}
}
@ -7304,6 +7319,20 @@ SDValue PPCTargetLowering::DAGCombineTruncBoolExt(SDNode *N,
SDNode *User = *UI;
if (User != N && !Visited.count(User))
return SDValue();
// Make sure that we're not going to promote the non-output-value
// operand(s) or SELECT or SELECT_CC.
// FIXME: Although we could sometimes handle this, and it does occur in
// practice that one of the condition inputs to the select is also one of
// the outputs, we currently can't deal with this.
if (User->getOpcode() == ISD::SELECT) {
if (User->getOperand(0) == PromOps[i])
return SDValue();
} else if (User->getOpcode() == ISD::SELECT_CC) {
if (User->getOperand(0) == PromOps[i] ||
User->getOperand(1) == PromOps[i])
return SDValue();
}
}
}
@ -7391,8 +7420,6 @@ SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N,
SelectionDAG &DAG = DCI.DAG;
SDLoc dl(N);
assert(PPCSubTarget.useCRBits() &&
"Expecting to be tracking CR bits");
// If we're tracking CR bits, we need to be careful that we don't have:
// zext(binary-ops(trunc(x), trunc(y)))
// or
@ -7402,11 +7429,19 @@ SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N,
// bits are set as required by the final extension, we still may need to do
// some masking to get the proper behavior.
// This same functionality is important on PPC64 when dealing with
// 32-to-64-bit extensions; these occur often when 32-bit values are used as
// the return values of functions. Because it is so similar, it is handled
// here as well.
if (N->getValueType(0) != MVT::i32 &&
N->getValueType(0) != MVT::i64)
return SDValue();
if (N->getOperand(0).getValueType() != MVT::i1)
if (!((N->getOperand(0).getValueType() == MVT::i1 &&
PPCSubTarget.useCRBits()) ||
(N->getOperand(0).getValueType() == MVT::i32 &&
PPCSubTarget.isPPC64())))
return SDValue();
if (N->getOperand(0).getOpcode() != ISD::AND &&
@ -7468,6 +7503,20 @@ SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N,
SDNode *User = *UI;
if (User != N && !Visited.count(User))
return SDValue();
// Make sure that we're not going to promote the non-output-value
// operand(s) or SELECT or SELECT_CC.
// FIXME: Although we could sometimes handle this, and it does occur in
// practice that one of the condition inputs to the select is also one of
// the outputs, we currently can't deal with this.
if (User->getOpcode() == ISD::SELECT) {
if (User->getOperand(0) == Inputs[i])
return SDValue();
} else if (User->getOpcode() == ISD::SELECT_CC) {
if (User->getOperand(0) == Inputs[i] ||
User->getOperand(1) == Inputs[i])
return SDValue();
}
}
}
@ -7478,9 +7527,24 @@ SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N,
SDNode *User = *UI;
if (User != N && !Visited.count(User))
return SDValue();
// Make sure that we're not going to promote the non-output-value
// operand(s) or SELECT or SELECT_CC.
// FIXME: Although we could sometimes handle this, and it does occur in
// practice that one of the condition inputs to the select is also one of
// the outputs, we currently can't deal with this.
if (User->getOpcode() == ISD::SELECT) {
if (User->getOperand(0) == PromOps[i])
return SDValue();
} else if (User->getOpcode() == ISD::SELECT_CC) {
if (User->getOperand(0) == PromOps[i] ||
User->getOperand(1) == PromOps[i])
return SDValue();
}
}
}
unsigned PromBits = N->getOperand(0).getValueSizeInBits();
bool ReallyNeedsExt = false;
if (N->getOpcode() != ISD::ANY_EXTEND) {
// If all of the inputs are not already sign/zero extended, then
@ -7491,12 +7555,15 @@ SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N,
unsigned OpBits =
Inputs[i].getOperand(0).getValueSizeInBits();
assert(PromBits < OpBits && "Truncation not to a smaller bit count?");
if ((N->getOpcode() == ISD::ZERO_EXTEND &&
!DAG.MaskedValueIsZero(Inputs[i].getOperand(0),
APInt::getHighBitsSet(OpBits,
OpBits-1))) ||
OpBits-PromBits))) ||
(N->getOpcode() == ISD::SIGN_EXTEND &&
DAG.ComputeNumSignBits(Inputs[i].getOperand(0)) != OpBits)) {
DAG.ComputeNumSignBits(Inputs[i].getOperand(0)) <
(OpBits-(PromBits-1)))) {
ReallyNeedsExt = true;
break;
}
@ -7580,16 +7647,19 @@ SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N,
if (!ReallyNeedsExt)
return N->getOperand(0);
// To zero extend, just mask off everything except for the first bit.
// To zero extend, just mask off everything except for the first bit (in the
// i1 case).
if (N->getOpcode() == ISD::ZERO_EXTEND)
return DAG.getNode(ISD::AND, dl, N->getValueType(0), N->getOperand(0),
DAG.getConstant(1, N->getValueType(0)));
DAG.getConstant(APInt::getLowBitsSet(
N->getValueSizeInBits(0), PromBits),
N->getValueType(0)));
assert(N->getOpcode() == ISD::SIGN_EXTEND &&
"Invalid extension type");
EVT ShiftAmountTy = getShiftAmountTy(N->getValueType(0));
SDValue ShiftCst =
DAG.getConstant(N->getValueSizeInBits(0)-1, ShiftAmountTy);
DAG.getConstant(N->getValueSizeInBits(0)-PromBits, ShiftAmountTy);
return DAG.getNode(ISD::SRA, dl, N->getValueType(0),
DAG.getNode(ISD::SHL, dl, N->getValueType(0),
N->getOperand(0), ShiftCst), ShiftCst);

32
test/CodeGen/PowerPC/crbits.ll
View File

@ -124,11 +124,39 @@ entry:
; CHECK-LABEL: @test7
; CHECK: andi. {{[0-9]+}}, 3, 1
; CHECK: isel [[REG1:[0-9]+]], 4, 5, 1
; CHECK: extsw 3, [[REG1]]
; CHECK: isel 3, 4, 5, 1
; CHECK: blr
}
define signext i32 @exttest7(i32 signext %a) #0 {
entry:
%cmp = icmp eq i32 %a, 5
%cond = select i1 %cmp, i32 7, i32 8
ret i32 %cond
; CHECK-LABEL: @exttest7
; CHECK-DAG: cmplwi {{[0-9]+}}, 3, 5
; CHECK-DAG: li [[REG1:[0-9]+]], 8
; CHECK-DAG: li [[REG2:[0-9]+]], 7
; CHECK: isel 3, [[REG2]], [[REG1]],
; CHECK-NOT: rldicl
; CHECK: blr
}
define zeroext i32 @exttest8() #0 {
entry:
%v0 = load i64* undef, align 8
%sub = sub i64 80, %v0
%div = lshr i64 %sub, 1
%conv13 = trunc i64 %div to i32
%cmp14 = icmp ugt i32 %conv13, 80
%.conv13 = select i1 %cmp14, i32 0, i32 %conv13
ret i32 %.conv13
; CHECK-LABEL: @exttest8
; This is a don't-crash test: %conv13 is both one of the possible select output
; values and also an input to the conditional feeding it.
}
; Function Attrs: nounwind readnone
define float @test8(i1 zeroext %v2, float %v1, float %v3) #0 {
entry: