Index: llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp =================================================================== --- llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp +++ llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp @@ -244,7 +244,6 @@ bool tryInlineAsm(SDNode *N); - void SelectConcatVector(SDNode *N); void SelectCMPZ(SDNode *N, bool &SwitchEQNEToPLMI); void SelectCMP_SWAP(SDNode *N); @@ -2585,15 +2584,6 @@ CurDAG->RemoveDeadNode(N); } -void ARMDAGToDAGISel::SelectConcatVector(SDNode *N) { - // The only time a CONCAT_VECTORS operation can have legal types is when - // two 64-bit vectors are concatenated to a 128-bit vector. - EVT VT = N->getValueType(0); - if (!VT.is128BitVector() || N->getNumOperands() != 2) - llvm_unreachable("unexpected CONCAT_VECTORS"); - ReplaceNode(N, createDRegPairNode(VT, N->getOperand(0), N->getOperand(1))); -} - static Optional> getContiguousRangeOfSetBits(const APInt &A) { unsigned FirstOne = A.getBitWidth() - A.countLeadingZeros() - 1; @@ -2900,49 +2890,6 @@ break; } - case ARMISD::VMOVRRD: - ReplaceNode(N, CurDAG->getMachineNode(ARM::VMOVRRD, dl, MVT::i32, MVT::i32, - N->getOperand(0), getAL(CurDAG, dl), - CurDAG->getRegister(0, MVT::i32))); - return; - case ISD::UMUL_LOHI: { - if (Subtarget->isThumb1Only()) - break; - if (Subtarget->isThumb()) { - SDValue Ops[] = { N->getOperand(0), N->getOperand(1), - getAL(CurDAG, dl), CurDAG->getRegister(0, MVT::i32) }; - ReplaceNode( - N, CurDAG->getMachineNode(ARM::t2UMULL, dl, MVT::i32, MVT::i32, Ops)); - return; - } else { - SDValue Ops[] = { N->getOperand(0), N->getOperand(1), - getAL(CurDAG, dl), CurDAG->getRegister(0, MVT::i32), - CurDAG->getRegister(0, MVT::i32) }; - ReplaceNode(N, CurDAG->getMachineNode( - Subtarget->hasV6Ops() ? ARM::UMULL : ARM::UMULLv5, dl, - MVT::i32, MVT::i32, Ops)); - return; - } - } - case ISD::SMUL_LOHI: { - if (Subtarget->isThumb1Only()) - break; - if (Subtarget->isThumb()) { - SDValue Ops[] = { N->getOperand(0), N->getOperand(1), - getAL(CurDAG, dl), CurDAG->getRegister(0, MVT::i32) }; - ReplaceNode( - N, CurDAG->getMachineNode(ARM::t2SMULL, dl, MVT::i32, MVT::i32, Ops)); - return; - } else { - SDValue Ops[] = { N->getOperand(0), N->getOperand(1), - getAL(CurDAG, dl), CurDAG->getRegister(0, MVT::i32), - CurDAG->getRegister(0, MVT::i32) }; - ReplaceNode(N, CurDAG->getMachineNode( - Subtarget->hasV6Ops() ? ARM::SMULL : ARM::SMULLv5, dl, - MVT::i32, MVT::i32, Ops)); - return; - } - } case ARMISD::UMAAL: { unsigned Opc = Subtarget->isThumb() ? ARM::t2UMAAL : ARM::UMAAL; SDValue Ops[] = { N->getOperand(0), N->getOperand(1), @@ -3836,10 +3783,6 @@ return; } - case ISD::CONCAT_VECTORS: - SelectConcatVector(N); - return; - case ISD::ATOMIC_CMP_SWAP: SelectCMP_SWAP(N); return; Index: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td =================================================================== --- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td +++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td @@ -3956,27 +3956,35 @@ let isCommutable = 1 in { def SMULL : AsMul1I64<0b0000110, (outs GPR:$RdLo, GPR:$RdHi), (ins GPR:$Rn, GPR:$Rm), IIC_iMUL64, - "smull", "\t$RdLo, $RdHi, $Rn, $Rm", []>, + "smull", "\t$RdLo, $RdHi, $Rn, $Rm", + [(set GPR:$RdLo, GPR:$RdHi, + (smullohi GPR:$Rn, GPR:$Rm))]>, Requires<[IsARM, HasV6]>, Sched<[WriteMUL64Lo, WriteMUL64Hi, ReadMUL, ReadMUL]>; def UMULL : AsMul1I64<0b0000100, (outs GPR:$RdLo, GPR:$RdHi), (ins GPR:$Rn, GPR:$Rm), IIC_iMUL64, - "umull", "\t$RdLo, $RdHi, $Rn, $Rm", []>, + "umull", "\t$RdLo, $RdHi, $Rn, $Rm", + [(set GPR:$RdLo, GPR:$RdHi, + (umullohi GPR:$Rn, GPR:$Rm))]>, Requires<[IsARM, HasV6]>, Sched<[WriteMAC64Lo, WriteMAC64Hi, ReadMUL, ReadMUL]>; let Constraints = "@earlyclobber $RdLo,@earlyclobber $RdHi" in { def SMULLv5 : ARMPseudoExpand<(outs GPR:$RdLo, GPR:$RdHi), (ins GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s), - 4, IIC_iMUL64, [], + 4, IIC_iMUL64, + [(set GPR:$RdLo, GPR:$RdHi, + (smullohi GPR:$Rn, GPR:$Rm))], (SMULL GPR:$RdLo, GPR:$RdHi, GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s)>, Requires<[IsARM, NoV6]>, Sched<[WriteMUL64Lo, WriteMUL64Hi, ReadMUL, ReadMUL]>; def UMULLv5 : ARMPseudoExpand<(outs GPR:$RdLo, GPR:$RdHi), (ins GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s), - 4, IIC_iMUL64, [], + 4, IIC_iMUL64, + [(set GPR:$RdLo, GPR:$RdHi, + (umullohi GPR:$Rn, GPR:$Rm))], (UMULL GPR:$RdLo, GPR:$RdHi, GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s)>, Requires<[IsARM, NoV6]>, Sched<[WriteMUL64Lo, WriteMUL64Hi, ReadMUL, ReadMUL]>; Index: llvm/trunk/lib/Target/ARM/ARMInstrNEON.td =================================================================== --- llvm/trunk/lib/Target/ARM/ARMInstrNEON.td +++ llvm/trunk/lib/Target/ARM/ARMInstrNEON.td @@ -7139,6 +7139,17 @@ (f64 (IMPLICIT_DEF)), (i32 0)))), dsub_0)), dsub_0))>; } +def : Pat<(v2i64 (concat_vectors DPR:$Dn, DPR:$Dm)), + (REG_SEQUENCE QPR, DPR:$Dn, dsub_0, DPR:$Dm, dsub_1)>; +def : Pat<(v4i32 (concat_vectors DPR:$Dn, DPR:$Dm)), + (REG_SEQUENCE QPR, DPR:$Dn, dsub_0, DPR:$Dm, dsub_1)>; +def : Pat<(v8i16 (concat_vectors DPR:$Dn, DPR:$Dm)), + (REG_SEQUENCE QPR, DPR:$Dn, dsub_0, DPR:$Dm, dsub_1)>; +def : Pat<(v16i8 (concat_vectors DPR:$Dn, DPR:$Dm)), + (REG_SEQUENCE QPR, DPR:$Dn, dsub_0, DPR:$Dm, dsub_1)>; +def : Pat<(v4f32 (concat_vectors DPR:$Dn, DPR:$Dm)), + (REG_SEQUENCE QPR, DPR:$Dn, dsub_0, DPR:$Dm, dsub_1)>; + //===----------------------------------------------------------------------===// // Assembler aliases // Index: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td =================================================================== --- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td +++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td @@ -2603,8 +2603,12 @@ // Extra precision multiplies with low / high results let hasSideEffects = 0 in { let isCommutable = 1 in { -def t2SMULL : T2MulLong<0b000, 0b0000, "smull", []>; -def t2UMULL : T2MulLong<0b010, 0b0000, "umull", []>; +def t2SMULL : T2MulLong<0b000, 0b0000, "smull", + [(set rGPR:$RdLo, rGPR:$RdHi, + (smullohi rGPR:$Rn, rGPR:$Rm))]>; +def t2UMULL : T2MulLong<0b010, 0b0000, "umull", + [(set rGPR:$RdLo, rGPR:$RdHi, + (umullohi rGPR:$Rn, rGPR:$Rm))]>; } // isCommutable // Multiply + accumulate Index: llvm/trunk/lib/Target/ARM/ARMInstrVFP.td =================================================================== --- llvm/trunk/lib/Target/ARM/ARMInstrVFP.td +++ llvm/trunk/lib/Target/ARM/ARMInstrVFP.td @@ -14,11 +14,14 @@ def SDT_CMPFP0 : SDTypeProfile<0, 2, [SDTCisFP<0>, SDTCisVT<1, i32>]>; def SDT_VMOVDRR : SDTypeProfile<1, 2, [SDTCisVT<0, f64>, SDTCisVT<1, i32>, SDTCisSameAs<1, 2>]>; +def SDT_VMOVRRD : SDTypeProfile<2, 1, [SDTCisVT<0, i32>, SDTCisSameAs<0, 1>, + SDTCisVT<2, f64>]>; def arm_fmstat : SDNode<"ARMISD::FMSTAT", SDTNone, [SDNPInGlue, SDNPOutGlue]>; def arm_cmpfp : SDNode<"ARMISD::CMPFP", SDT_ARMFCmp, [SDNPOutGlue]>; def arm_cmpfp0 : SDNode<"ARMISD::CMPFPw0", SDT_CMPFP0, [SDNPOutGlue]>; def arm_fmdrr : SDNode<"ARMISD::VMOVDRR", SDT_VMOVDRR>; +def arm_fmrrd : SDNode<"ARMISD::VMOVRRD", SDT_VMOVRRD>; //===----------------------------------------------------------------------===// // Operand Definitions. @@ -1054,7 +1057,7 @@ def VMOVRRD : AVConv3I<0b11000101, 0b1011, (outs GPR:$Rt, GPR:$Rt2), (ins DPR:$Dm), IIC_fpMOVDI, "vmov", "\t$Rt, $Rt2, $Dm", - [/* FIXME: Can't write pattern for multiple result instr*/]>, + [(set GPR:$Rt, GPR:$Rt2, (arm_fmrrd DPR:$Dm))]>, Sched<[WriteFPMOV]> { // Instruction operands. bits<5> Dm;