bool tryInlineAsm(SDNode *N);
void SelectConcatVector(SDNode *N);
-
+ void SelectCMPZ(SDNode *N, bool &SwitchEQNEToPLMI);
+
bool trySMLAWSMULW(SDNode *N);
void SelectCMP_SWAP(SDNode *N);
ReplaceNode(N, createDRegPairNode(VT, N->getOperand(0), N->getOperand(1)));
}
+static Optional<std::pair<unsigned, unsigned>>
+getContiguousRangeOfSetBits(const APInt &A) {
+ unsigned FirstOne = A.getBitWidth() - A.countLeadingZeros() - 1;
+ unsigned LastOne = A.countTrailingZeros();
+ if (A.countPopulation() != (FirstOne - LastOne + 1))
+ return Optional<std::pair<unsigned,unsigned>>();
+ return std::make_pair(FirstOne, LastOne);
+}
+
+void ARMDAGToDAGISel::SelectCMPZ(SDNode *N, bool &SwitchEQNEToPLMI) {
+ assert(N->getOpcode() == ARMISD::CMPZ);
+ SwitchEQNEToPLMI = false;
+
+ if (!Subtarget->isThumb())
+ // FIXME: Work out whether it is profitable to do this in A32 mode - LSL and
+ // LSR don't exist as standalone instructions - they need the barrel shifter.
+ return;
+ // select (cmpz (and X, C), #0) -> (LSLS X) or (LSRS X) or (LSRS (LSLS X))
+ SDValue And = N->getOperand(0);
+ SDValue Zero = N->getOperand(1);
+ if (!isa<ConstantSDNode>(Zero) || !cast<ConstantSDNode>(Zero)->isNullValue() ||
+ And->getOpcode() != ISD::AND)
+ return;
+ SDValue X = And.getOperand(0);
+ auto C = dyn_cast<ConstantSDNode>(And.getOperand(1));
+
+ if (!C || !X->hasOneUse())
+ return;
+ auto Range = getContiguousRangeOfSetBits(C->getAPIntValue());
+ if (!Range)
+ return;
+
+ // There are several ways to lower this:
+ SDNode *NewN;
+ SDLoc dl(N);
+
+ auto EmitShift = [&](unsigned Opc, SDValue Src, unsigned Imm) -> SDNode* {
+ if (Subtarget->isThumb2()) {
+ Opc = (Opc == ARM::tLSLri) ? ARM::t2LSLri : ARM::t2LSRri;
+ SDValue Ops[] = { Src, CurDAG->getTargetConstant(Imm, dl, MVT::i32),
+ getAL(CurDAG, dl), CurDAG->getRegister(0, MVT::i32),
+ CurDAG->getRegister(0, MVT::i32) };
+ return CurDAG->getMachineNode(Opc, dl, MVT::i32, Ops);
+ } else {
+ SDValue Ops[] = {CurDAG->getRegister(ARM::CPSR, MVT::i32), Src,
+ CurDAG->getTargetConstant(Imm, dl, MVT::i32),
+ getAL(CurDAG, dl), CurDAG->getRegister(0, MVT::i32)};
+ return CurDAG->getMachineNode(Opc, dl, MVT::i32, Ops);
+ }
+ };
+
+ if (Range->second == 0) {
+ // 1. Mask includes the LSB -> Simply shift the top N bits off
+ NewN = EmitShift(ARM::tLSLri, X, 31 - Range->first);
+ ReplaceNode(And.getNode(), NewN);
+ } else if (Range->first == 31) {
+ // 2. Mask includes the MSB -> Simply shift the bottom N bits off
+ NewN = EmitShift(ARM::tLSRri, X, Range->second);
+ ReplaceNode(And.getNode(), NewN);
+ } else if (Range->first == Range->second) {
+ // 3. Only one bit is set. We can shift this into the sign bit and use a
+ // PL/MI comparison.
+ NewN = EmitShift(ARM::tLSLri, X, 31 - Range->first);
+ ReplaceNode(And.getNode(), NewN);
+
+ SwitchEQNEToPLMI = true;
+ } else if (!Subtarget->hasV6T2Ops()) {
+ // 4. Do a double shift to clear bottom and top bits, but only in
+ // thumb-1 mode as in thumb-2 we can use UBFX.
+ NewN = EmitShift(ARM::tLSLri, X, 31 - Range->first);
+ NewN = EmitShift(ARM::tLSRri, SDValue(NewN, 0),
+ Range->second + (31 - Range->first));
+ ReplaceNode(And.getNode(), NewN);
+ }
+
+}
+
void ARMDAGToDAGISel::Select(SDNode *N) {
SDLoc dl(N);
return;
}
}
+
break;
}
case ARMISD::VMOVRRD:
assert(N2.getOpcode() == ISD::Constant);
assert(N3.getOpcode() == ISD::Register);
- SDValue Tmp2 = CurDAG->getTargetConstant(((unsigned)
- cast<ConstantSDNode>(N2)->getZExtValue()), dl,
- MVT::i32);
+ unsigned CC = (unsigned) cast<ConstantSDNode>(N2)->getZExtValue();
+
+ if (InFlag.getOpcode() == ARMISD::CMPZ) {
+ bool SwitchEQNEToPLMI;
+ SelectCMPZ(InFlag.getNode(), SwitchEQNEToPLMI);
+ InFlag = N->getOperand(4);
+
+ if (SwitchEQNEToPLMI) {
+ switch ((ARMCC::CondCodes)CC) {
+ default: llvm_unreachable("CMPZ must be either NE or EQ!");
+ case ARMCC::NE:
+ CC = (unsigned)ARMCC::MI;
+ break;
+ case ARMCC::EQ:
+ CC = (unsigned)ARMCC::PL;
+ break;
+ }
+ }
+ }
+
+ SDValue Tmp2 = CurDAG->getTargetConstant(CC, dl, MVT::i32);
SDValue Ops[] = { N1, Tmp2, N3, Chain, InFlag };
SDNode *ResNode = CurDAG->getMachineNode(Opc, dl, MVT::Other,
MVT::Glue, Ops);
// Other cases are autogenerated.
break;
}
+
+ case ARMISD::CMOV: {
+ SDValue InFlag = N->getOperand(4);
+
+ if (InFlag.getOpcode() == ARMISD::CMPZ) {
+ bool SwitchEQNEToPLMI;
+ SelectCMPZ(InFlag.getNode(), SwitchEQNEToPLMI);
+
+ if (SwitchEQNEToPLMI) {
+ SDValue ARMcc = N->getOperand(2);
+ ARMCC::CondCodes CC =
+ (ARMCC::CondCodes)cast<ConstantSDNode>(ARMcc)->getZExtValue();
+
+ switch (CC) {
+ default: llvm_unreachable("CMPZ must be either NE or EQ!");
+ case ARMCC::NE:
+ CC = ARMCC::MI;
+ break;
+ case ARMCC::EQ:
+ CC = ARMCC::PL;
+ break;
+ }
+ SDValue NewARMcc = CurDAG->getConstant((unsigned)CC, dl, MVT::i32);
+ SDValue Ops[] = {N->getOperand(0), N->getOperand(1), NewARMcc,
+ N->getOperand(3), N->getOperand(4)};
+ CurDAG->MorphNodeTo(N, ARMISD::CMOV, N->getVTList(), Ops);
+ }
+
+ }
+ // Other cases are autogenerated.
+ break;
+ }
case ARMISD::VZIP: {
unsigned Opc = 0;
--- /dev/null
+; RUN: llc -mtriple=thumbv7m-linux-gnu < %s | FileCheck %s --check-prefix=CHECK --check-prefix=T2
+; RUN: llc -mtriple=thumbv6m-linux-gnu < %s | FileCheck %s --check-prefix=CHECK --check-prefix=T1
+
+; CHECK-LABEL: single_bit:
+; CHECK: lsls r0, r0, #23
+; T2-NEXT: mov
+; T2-NEXT: it
+; T1-NEXT: bmi
+define i32 @single_bit(i32 %p) {
+ %a = and i32 %p, 256
+ %b = icmp eq i32 %a, 0
+ br i1 %b, label %true, label %false
+
+true:
+ ret i32 1
+
+false:
+ ret i32 2
+}
+
+; CHECK-LABEL: multi_bit_lsb_ubfx:
+; CHECK: lsls r0, r0, #24
+; T2-NEXT: mov
+; T2-NEXT: it
+; T1-NEXT: beq
+define i32 @multi_bit_lsb_ubfx(i32 %p) {
+ %a = and i32 %p, 255
+ %b = icmp eq i32 %a, 0
+ br i1 %b, label %true, label %false
+
+true:
+ ret i32 1
+
+false:
+ ret i32 2
+}
+
+; CHECK-LABEL: multi_bit_msb:
+; CHECK: lsrs r0, r0, #24
+; T2-NEXT: mov
+; T2-NEXT: it
+; T1-NEXT: beq
+define i32 @multi_bit_msb(i32 %p) {
+ %a = and i32 %p, 4278190080 ; 0xff000000
+ %b = icmp eq i32 %a, 0
+ br i1 %b, label %true, label %false
+
+true:
+ ret i32 1
+
+false:
+ ret i32 2
+}
+
+; CHECK-LABEL: multi_bit_nosb:
+; T1: lsls r0, r0, #8
+; T1-NEXT: lsrs r0, r0, #24
+; T2: tst.w
+; T2-NEXT: it
+; T1-NEXT: beq
+define i32 @multi_bit_nosb(i32 %p) {
+ %a = and i32 %p, 16711680 ; 0x00ff0000
+ %b = icmp eq i32 %a, 0
+ br i1 %b, label %true, label %false
+
+true:
+ ret i32 1
+
+false:
+ ret i32 2
+}
define float @select_f(float %a, float %b, i1 %c) {
; CHECK-LABEL: select_f:
-; NONE: tst.w r2, #1
+; NONE: lsls r2, r2, #31
; NONE: moveq r0, r1
-; HARD: tst.w r0, #1
+; HARD: lsls r0, r0, #31
; VFP4-ALL: vmovne.f32 s1, s0
; VFP4-ALL: vmov.f32 s0, s1
; FP-ARMv8: vseleq.f32 s0, s1, s0
define double @select_d(double %a, double %b, i1 %c) {
; CHECK-LABEL: select_d:
-; NONE: ldr.w [[REG:r[0-9]+]], [sp]
-; NONE: ands [[REG]], [[REG]], #1
+; NONE: ldr{{(.w)?}} [[REG:r[0-9]+]], [sp]
+; NONE: lsls{{(.w)?}} [[REG]], [[REG]], #31
; NONE: moveq r0, r2
; NONE: moveq r1, r3
-; SP: ands r0, r0, #1
+; SP: lsls r0, r0, #31
; SP-DAG: vmov [[ALO:r[0-9]+]], [[AHI:r[0-9]+]], d0
; SP-DAG: vmov [[BLO:r[0-9]+]], [[BHI:r[0-9]+]], d1
; SP: itt ne
; SP-DAG: movne [[BLO]], [[ALO]]
; SP-DAG: movne [[BHI]], [[AHI]]
; SP: vmov d0, [[BLO]], [[BHI]]
-; DP: tst.w r0, #1
+; DP: lsls r0, r0, #31
; VFP4-DP: vmovne.f64 d1, d0
; VFP4-DP: vmov.f64 d0, d1
; FP-ARMV8: vseleq.f64 d0, d1, d0