Handle combines with zero and constant canonicalization for adds.
llvm-svn: 351024
SDValue visitADD(SDNode *N);
SDValue visitADDLike(SDValue N0, SDValue N1, SDNode *LocReference);
SDValue visitSUB(SDNode *N);
+ SDValue visitADDSAT(SDNode *N);
+ SDValue visitSUBSAT(SDNode *N);
SDValue visitADDC(SDNode *N);
SDValue visitUADDO(SDNode *N);
SDValue visitUADDOLike(SDValue N0, SDValue N1, SDNode *N);
case ISD::MERGE_VALUES: return visitMERGE_VALUES(N);
case ISD::ADD: return visitADD(N);
case ISD::SUB: return visitSUB(N);
+ case ISD::SADDSAT:
+ case ISD::UADDSAT: return visitADDSAT(N);
+ case ISD::SSUBSAT:
+ case ISD::USUBSAT: return visitSUBSAT(N);
case ISD::ADDC: return visitADDC(N);
case ISD::UADDO: return visitUADDO(N);
case ISD::SUBC: return visitSUBC(N);
return SDValue();
}
+SDValue DAGCombiner::visitADDSAT(SDNode *N) {
+ unsigned Opcode = N->getOpcode();
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ EVT VT = N0.getValueType();
+ SDLoc DL(N);
+
+ // fold vector ops
+ if (VT.isVector()) {
+ // TODO SimplifyVBinOp
+
+ // fold (add_sat x, 0) -> x, vector edition
+ if (ISD::isBuildVectorAllZeros(N1.getNode()))
+ return N0;
+ if (ISD::isBuildVectorAllZeros(N0.getNode()))
+ return N1;
+ }
+
+ if (DAG.isConstantIntBuildVectorOrConstantInt(N0)) {
+ // canonicalize constant to RHS
+ if (!DAG.isConstantIntBuildVectorOrConstantInt(N1))
+ return DAG.getNode(Opcode, DL, VT, N1, N0);
+
+ // TODO Constant Folding
+ }
+
+ // fold (add_sat x, 0) -> x
+ if (isNullConstant(N1))
+ return N0;
+
+ return SDValue();
+}
+
static SDValue getAsCarry(const TargetLowering &TLI, SDValue V) {
bool Masked = false;
return SDValue();
}
+SDValue DAGCombiner::visitSUBSAT(SDNode *N) {
+ unsigned Opcode = N->getOpcode();
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ EVT VT = N0.getValueType();
+ SDLoc DL(N);
+
+ // fold vector ops
+ if (VT.isVector()) {
+ // TODO SimplifyVBinOp
+
+ // fold (sub_sat x, 0) -> x, vector edition
+ if (ISD::isBuildVectorAllZeros(N1.getNode()))
+ return N0;
+ }
+
+ // TODO Constant Folding
+
+ // fold (sub_sat x, 0) -> x
+ if (isNullConstant(N1))
+ return N0;
+
+ return SDValue();
+}
+
SDValue DAGCombiner::visitSUBC(SDNode *N) {
SDValue N0 = N->getOperand(0);
SDValue N1 = N->getOperand(1);
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: movl %edi, %ecx
-; CHECK-NEXT: addl $1, %ecx
+; CHECK-NEXT: incl %ecx
; CHECK-NEXT: setns %al
; CHECK-NEXT: addl $2147483647, %eax # imm = 0x7FFFFFFF
-; CHECK-NEXT: addl $1, %edi
+; CHECK-NEXT: incl %edi
; CHECK-NEXT: cmovnol %edi, %eax
; CHECK-NEXT: retq
%res = call i32 @llvm.sadd.sat.i32(i32 1, i32 %a0);
define i32 @combine_zero_i32(i32 %a0) {
; CHECK-LABEL: combine_zero_i32:
; CHECK: # %bb.0:
-; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: movl %edi, %ecx
-; CHECK-NEXT: addl $0, %ecx
-; CHECK-NEXT: setns %al
-; CHECK-NEXT: addl $2147483647, %eax # imm = 0x7FFFFFFF
-; CHECK-NEXT: addl $0, %edi
-; CHECK-NEXT: cmovnol %edi, %eax
+; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = call i32 @llvm.sadd.sat.i32(i32 %a0, i32 0);
ret i32 %1
}
define <8 x i16> @combine_zero_v8i16(<8 x i16> %a0) {
-; SSE-LABEL: combine_zero_v8i16:
-; SSE: # %bb.0:
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: paddsw %xmm1, %xmm0
-; SSE-NEXT: retq
-;
-; AVX-LABEL: combine_zero_v8i16:
-; AVX: # %bb.0:
-; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vpaddsw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; CHECK-LABEL: combine_zero_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: retq
%1 = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %a0, <8 x i16> zeroinitializer);
ret <8 x i16> %1
}
define i32 @combine_zero_i32(i32 %a0) {
; CHECK-LABEL: combine_zero_i32:
; CHECK: # %bb.0:
-; CHECK-NEXT: addl $0, %edi
-; CHECK-NEXT: movl $-1, %eax
-; CHECK-NEXT: cmovael %edi, %eax
+; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = call i32 @llvm.uadd.sat.i32(i32 %a0, i32 0);
ret i32 %1
}
define <8 x i16> @combine_zero_v8i16(<8 x i16> %a0) {
-; SSE-LABEL: combine_zero_v8i16:
-; SSE: # %bb.0:
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: paddusw %xmm1, %xmm0
-; SSE-NEXT: retq
-;
-; AVX-LABEL: combine_zero_v8i16:
-; AVX: # %bb.0:
-; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vpaddusw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; CHECK-LABEL: combine_zero_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: retq
%1 = call <8 x i16> @llvm.uadd.sat.v8i16(<8 x i16> %a0, <8 x i16> zeroinitializer);
ret <8 x i16> %1
}
define i32 @combine_zero_i32(i32 %a0) {
; CHECK-LABEL: combine_zero_i32:
; CHECK: # %bb.0:
-; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: movl %edi, %ecx
-; CHECK-NEXT: subl $0, %ecx
-; CHECK-NEXT: setns %al
-; CHECK-NEXT: addl $2147483647, %eax # imm = 0x7FFFFFFF
-; CHECK-NEXT: subl $0, %edi
-; CHECK-NEXT: cmovnol %edi, %eax
+; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = call i32 @llvm.ssub.sat.i32(i32 %a0, i32 0);
ret i32 %1
}
define <8 x i16> @combine_zero_v8i16(<8 x i16> %a0) {
-; SSE-LABEL: combine_zero_v8i16:
-; SSE: # %bb.0:
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: psubsw %xmm1, %xmm0
-; SSE-NEXT: retq
-;
-; AVX-LABEL: combine_zero_v8i16:
-; AVX: # %bb.0:
-; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vpsubsw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; CHECK-LABEL: combine_zero_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: retq
%1 = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %a0, <8 x i16> zeroinitializer);
ret <8 x i16> %1
}
define i32 @combine_zero_i32(i32 %a0) {
; CHECK-LABEL: combine_zero_i32:
; CHECK: # %bb.0:
-; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: subl $0, %edi
-; CHECK-NEXT: cmovael %edi, %eax
+; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = call i32 @llvm.usub.sat.i32(i32 %a0, i32 0);
ret i32 %1
}
define <8 x i16> @combine_zero_v8i16(<8 x i16> %a0) {
-; SSE-LABEL: combine_zero_v8i16:
-; SSE: # %bb.0:
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: psubusw %xmm1, %xmm0
-; SSE-NEXT: retq
-;
-; AVX-LABEL: combine_zero_v8i16:
-; AVX: # %bb.0:
-; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vpsubusw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; CHECK-LABEL: combine_zero_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: retq
%1 = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> %a0, <8 x i16> zeroinitializer);
ret <8 x i16> %1
}