ARM parsing for VLD1 all lanes, with writeback.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@145510 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 82 insertions, 43 deletions

diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 93153484de..77f7ee94aa 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -2446,9 +2446,12 @@ ARMBaseInstrInfo::getOperandLatency(const InstrItineraryData *ItinData,
     case ARM::VLD1DUPq8:
     case ARM::VLD1DUPq16:
     case ARM::VLD1DUPq32:
-    case ARM::VLD1DUPq8_UPD:
-    case ARM::VLD1DUPq16_UPD:
-    case ARM::VLD1DUPq32_UPD:
+    case ARM::VLD1DUPq8wb_fixed:
+    case ARM::VLD1DUPq16wb_fixed:
+    case ARM::VLD1DUPq32wb_fixed:
+    case ARM::VLD1DUPq8wb_register:
+    case ARM::VLD1DUPq16wb_register:
+    case ARM::VLD1DUPq32wb_register:
     case ARM::VLD2DUPd8:
     case ARM::VLD2DUPd16:
     case ARM::VLD2DUPd32:
@@ -2621,9 +2624,12 @@ ARMBaseInstrInfo::getOperandLatency(const InstrItineraryData *ItinData,
     case ARM::VLD1DUPq8Pseudo:
     case ARM::VLD1DUPq16Pseudo:
     case ARM::VLD1DUPq32Pseudo:
-    case ARM::VLD1DUPq8Pseudo_UPD:
-    case ARM::VLD1DUPq16Pseudo_UPD:
-    case ARM::VLD1DUPq32Pseudo_UPD:
+    case ARM::VLD1DUPq8PseudoWB_fixed:
+    case ARM::VLD1DUPq16PseudoWB_fixed:
+    case ARM::VLD1DUPq32PseudoWB_fixed:
+    case ARM::VLD1DUPq8PseudoWB_register:
+    case ARM::VLD1DUPq16PseudoWB_register:
+    case ARM::VLD1DUPq32PseudoWB_register:
     case ARM::VLD2DUPd8Pseudo:
     case ARM::VLD2DUPd16Pseudo:
     case ARM::VLD2DUPd32Pseudo:
diff --git a/lib/Target/ARM/ARMExpandPseudoInsts.cpp b/lib/Target/ARM/ARMExpandPseudoInsts.cpp
index 9daaf41285..6fe6fcde1a 100644
--- a/lib/Target/ARM/ARMExpandPseudoInsts.cpp
+++ b/lib/Target/ARM/ARMExpandPseudoInsts.cpp
@@ -130,11 +130,14 @@ namespace {
 
 static const NEONLdStTableEntry NEONLdStTable[] = {
 { ARM::VLD1DUPq16Pseudo,     ARM::VLD1DUPq16,     true, false, false, SingleSpc, 2, 4,false},
-{ ARM::VLD1DUPq16Pseudo_UPD, ARM::VLD1DUPq16_UPD, true, true, true,  SingleSpc, 2, 4,true},
+{ ARM::VLD1DUPq16PseudoWB_fixed, ARM::VLD1DUPq16wb_fixed, true, true, true,  SingleSpc, 2, 4,false},
+{ ARM::VLD1DUPq16PseudoWB_register, ARM::VLD1DUPq16wb_register, true, true, true,  SingleSpc, 2, 4,false},
 { ARM::VLD1DUPq32Pseudo,     ARM::VLD1DUPq32,     true, false, false, SingleSpc, 2, 2,false},
-{ ARM::VLD1DUPq32Pseudo_UPD, ARM::VLD1DUPq32_UPD, true, true, true,  SingleSpc, 2, 2,true},
+{ ARM::VLD1DUPq32PseudoWB_fixed, ARM::VLD1DUPq32wb_fixed, true, true, false,  SingleSpc, 2, 2,false},
+{ ARM::VLD1DUPq32PseudoWB_register, ARM::VLD1DUPq32wb_register, true, true, true,  SingleSpc, 2, 2,false},
 { ARM::VLD1DUPq8Pseudo,      ARM::VLD1DUPq8,      true, false, false, SingleSpc, 2, 8,false},
-{ ARM::VLD1DUPq8Pseudo_UPD,  ARM::VLD1DUPq8_UPD, true, true, true,  SingleSpc, 2, 8,true},
+{ ARM::VLD1DUPq8PseudoWB_fixed,  ARM::VLD1DUPq8wb_fixed, true, true, false,  SingleSpc, 2, 8,false},
+{ ARM::VLD1DUPq8PseudoWB_register,  ARM::VLD1DUPq8wb_register, true, true, true,  SingleSpc, 2, 8,false},
 
 { ARM::VLD1LNq16Pseudo,     ARM::VLD1LNd16,     true, false, false, EvenDblSpc, 1, 4 ,true},
 { ARM::VLD1LNq16Pseudo_UPD, ARM::VLD1LNd16_UPD, true, true, true,  EvenDblSpc, 1, 4 ,true},
@@ -1133,9 +1136,12 @@ bool ARMExpandPseudo::ExpandMI(MachineBasicBlock &MBB,
     case ARM::VLD1DUPq8Pseudo:
     case ARM::VLD1DUPq16Pseudo:
     case ARM::VLD1DUPq32Pseudo:
-    case ARM::VLD1DUPq8Pseudo_UPD:
-    case ARM::VLD1DUPq16Pseudo_UPD:
-    case ARM::VLD1DUPq32Pseudo_UPD:
+    case ARM::VLD1DUPq8PseudoWB_fixed:
+    case ARM::VLD1DUPq16PseudoWB_fixed:
+    case ARM::VLD1DUPq32PseudoWB_fixed:
+    case ARM::VLD1DUPq8PseudoWB_register:
+    case ARM::VLD1DUPq16PseudoWB_register:
+    case ARM::VLD1DUPq32PseudoWB_register:
     case ARM::VLD2DUPd8Pseudo:
     case ARM::VLD2DUPd16Pseudo:
     case ARM::VLD2DUPd32Pseudo:
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index f985d7ee72..406f163f89 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -1065,32 +1065,63 @@ def VLD1DUPq16 : VLD1QDUP<{0,1,1,?}, "16">;
 def VLD1DUPq32 : VLD1QDUP<{1,0,1,?}, "32">;
 
 // ...with address register writeback:
-class VLD1DUPWB<bits<4> op7_4, string Dt>
-  : NLdSt<1, 0b10, 0b1100, op7_4, (outs DPR:$Vd, GPR:$wb),
-          (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD1dupu,
-          "vld1", Dt, "\\{$Vd[]\\}, $Rn$Rm", "$Rn.addr = $wb", []> {
-  let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVLD1DupInstruction";
+multiclass VLD1DUPWB<bits<4> op7_4, string Dt> {
+  def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
+                     (outs VecListOneDAllLanes:$Vd, GPR:$wb),
+                     (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+                     "vld1", Dt, "$Vd, $Rn!",
+                     "$Rn.addr = $wb", []> {
+    let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+    let Inst{4} = Rn{4};
+    let DecoderMethod = "DecodeVLD1DupInstruction";
+    let AsmMatchConverter = "cvtVLDwbFixed";
+  }
+  def _register : NLdSt<1, 0b10, 0b1100, op7_4,
+                        (outs VecListOneDAllLanes:$Vd, GPR:$wb),
+                        (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+                        "vld1", Dt, "$Vd, $Rn, $Rm",
+                        "$Rn.addr = $wb", []> {
+    let Inst{4} = Rn{4};
+    let DecoderMethod = "DecodeVLD1DupInstruction";
+    let AsmMatchConverter = "cvtVLDwbRegister";
+  }
 }
-class VLD1QDUPWB<bits<4> op7_4, string Dt>
-  : NLdSt<1, 0b10, 0b1100, op7_4, (outs DPR:$Vd, DPR:$dst2, GPR:$wb),
-          (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD1dupu,
-          "vld1", Dt, "\\{$Vd[], $dst2[]\\}, $Rn$Rm", "$Rn.addr = $wb", []> {
-  let Inst{4} = Rn{4};
-  let DecoderMethod = "DecodeVLD1DupInstruction";
+multiclass VLD1QDUPWB<bits<4> op7_4, string Dt> {
+  def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
+                     (outs VecListTwoDAllLanes:$Vd, GPR:$wb),
+                     (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+                     "vld1", Dt, "$Vd, $Rn!",
+                     "$Rn.addr = $wb", []> {
+    let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+    let Inst{4} = Rn{4};
+    let DecoderMethod = "DecodeVLD1DupInstruction";
+    let AsmMatchConverter = "cvtVLDwbFixed";
+  }
+  def _register : NLdSt<1, 0b10, 0b1100, op7_4,
+                        (outs VecListTwoDAllLanes:$Vd, GPR:$wb),
+                        (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+                        "vld1", Dt, "$Vd, $Rn, $Rm",
+                        "$Rn.addr = $wb", []> {
+    let Inst{4} = Rn{4};
+    let DecoderMethod = "DecodeVLD1DupInstruction";
+    let AsmMatchConverter = "cvtVLDwbRegister";
+  }
 }
 
-def VLD1DUPd8_UPD  : VLD1DUPWB<{0,0,0,0}, "8">;
-def VLD1DUPd16_UPD : VLD1DUPWB<{0,1,0,?}, "16">;
-def VLD1DUPd32_UPD : VLD1DUPWB<{1,0,0,?}, "32">;
+defm VLD1DUPd8wb  : VLD1DUPWB<{0,0,0,0}, "8">;
+defm VLD1DUPd16wb : VLD1DUPWB<{0,1,0,?}, "16">;
+defm VLD1DUPd32wb : VLD1DUPWB<{1,0,0,?}, "32">;
 
-def VLD1DUPq8_UPD  : VLD1QDUPWB<{0,0,1,0}, "8">;
-def VLD1DUPq16_UPD : VLD1QDUPWB<{0,1,1,?}, "16">;
-def VLD1DUPq32_UPD : VLD1QDUPWB<{1,0,1,?}, "32">;
+defm VLD1DUPq8wb  : VLD1QDUPWB<{0,0,1,0}, "8">;
+defm VLD1DUPq16wb : VLD1QDUPWB<{0,1,1,?}, "16">;
+defm VLD1DUPq32wb : VLD1QDUPWB<{1,0,1,?}, "32">;
 
-def VLD1DUPq8Pseudo_UPD  : VLDQWBPseudo<IIC_VLD1dupu>;
-def VLD1DUPq16Pseudo_UPD : VLDQWBPseudo<IIC_VLD1dupu>;
-def VLD1DUPq32Pseudo_UPD : VLDQWBPseudo<IIC_VLD1dupu>;
+def VLD1DUPq8PseudoWB_fixed     : VLDQWBfixedPseudo<IIC_VLD1dupu>;
+def VLD1DUPq16PseudoWB_fixed    : VLDQWBfixedPseudo<IIC_VLD1dupu>;
+def VLD1DUPq32PseudoWB_fixed    : VLDQWBfixedPseudo<IIC_VLD1dupu>;
+def VLD1DUPq8PseudoWB_register  : VLDQWBregisterPseudo<IIC_VLD1dupu>;
+def VLD1DUPq16PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD1dupu>;
+def VLD1DUPq32PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD1dupu>;
 
 //   VLD2DUP  : Vector Load (single 2-element structure to all lanes)
 class VLD2DUP<bits<4> op7_4, string Dt>
diff --git a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
index fb96407929..2dc4d12f17 100644
--- a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
+++ b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
@@ -2415,10 +2415,6 @@ static DecodeStatus DecodeVLD1DupInstruction(llvm::MCInst &Inst, unsigned Insn,
 
   if (!Check(S, DecodeDPRRegisterClass(Inst, Rd, Address, Decoder)))
     return MCDisassembler::Fail;
-  if (regs == 2) {
-    if (!Check(S, DecodeDPRRegisterClass(Inst, (Rd+1)%32, Address, Decoder)))
-      return MCDisassembler::Fail;
-  }
   if (Rm != 0xF) {
     if (!Check(S, DecodeGPRRegisterClass(Inst, Rn, Address, Decoder)))
       return MCDisassembler::Fail;
@@ -2428,12 +2424,12 @@ static DecodeStatus DecodeVLD1DupInstruction(llvm::MCInst &Inst, unsigned Insn,
     return MCDisassembler::Fail;
   Inst.addOperand(MCOperand::CreateImm(align));
 
-  if (Rm == 0xD)
-    Inst.addOperand(MCOperand::CreateReg(0));
-  else if (Rm != 0xF) {
-    if (!Check(S, DecodeGPRRegisterClass(Inst, Rm, Address, Decoder)))
-      return MCDisassembler::Fail;
-  }
+  // The fixed offset post-increment encodes Rm == 0xd. The no-writeback
+  // variant encodes Rm == 0xf. Anything else is a register offset post-
+  // increment and we need to add the register operand to the instruction.
+  if (Rm != 0xD && Rm != 0xF &&
+      !Check(S, DecodeGPRRegisterClass(Inst, Rm, Address, Decoder)))
+    return MCDisassembler::Fail;
 
   return S;
 }