NEON VLD4(multiple 4 element structures) assembly parsing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@148762 91177308-0d34-0410-b5e6-96231b3b80d8

5 files changed, 218 insertions, 19 deletions

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index e369fc1e7d..4ec3bc82e0 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -133,6 +133,15 @@ def VecListThreeQAsmOperand : AsmOperandClass {
 def VecListThreeQ : RegisterOperand<DPR, "printVectorListThreeSpaced"> {
   let ParserMatchClass = VecListThreeQAsmOperand;
 }
+// Register list of three D registers spaced by 2 (three Q registers).
+def VecListFourQAsmOperand : AsmOperandClass {
+  let Name = "VecListFourQ";
+  let ParserMethod = "parseVectorList";
+  let RenderMethod = "addVecListOperands";
+}
+def VecListFourQ : RegisterOperand<DPR, "printVectorListFourSpaced"> {
+  let ParserMatchClass = VecListFourQAsmOperand;
+}
 
 // Register list of one D register, with "all lanes" subscripting.
 def VecListOneDAllLanesAsmOperand : AsmOperandClass {
@@ -6196,6 +6205,65 @@ def VST3qWB_register_Asm_32 :
 
 
 
+// VLD4 multiple structure pseudo-instructions. These need special handling for
+// the vector operands that the normal instructions don't yet model.
+// FIXME: Remove these when the register classes and instructions are updated.
+def VLD4dAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VLD4dAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VLD4dAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VLD4qAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VLD4qAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VLD4qAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+
+def VLD4dWB_fixed_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VLD4dWB_fixed_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VLD4dWB_fixed_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VLD4qWB_fixed_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VLD4qWB_fixed_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VLD4qWB_fixed_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VLD4dWB_register_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
+                  (ins VecListFourD:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VLD4dWB_register_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
+                  (ins VecListFourD:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VLD4dWB_register_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
+                  (ins VecListFourD:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VLD4qWB_register_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
+                  (ins VecListFourQ:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VLD4qWB_register_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
+                  (ins VecListFourQ:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VLD4qWB_register_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
+                  (ins VecListFourQ:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+
 // VMOV takes an optional datatype suffix
 defm : VFPDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
                          (VORRd DPR:$Vd, DPR:$Vm, DPR:$Vm, pred:$p)>;
diff --git a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
index db3cf1deb1..81c060f616 100644
--- a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -5317,6 +5317,26 @@ static unsigned getRealVLDOpcode(unsigned Opc, unsigned &Spacing) {
   case ARM::VLD3qAsm_8:  Spacing = 2; return ARM::VLD3q8;
   case ARM::VLD3qAsm_16: Spacing = 2; return ARM::VLD3q16;
   case ARM::VLD3qAsm_32: Spacing = 2; return ARM::VLD3q32;
+
+  // VLD4
+  case ARM::VLD4dWB_fixed_Asm_8:  Spacing = 1; return ARM::VLD4d8_UPD;
+  case ARM::VLD4dWB_fixed_Asm_16: Spacing = 1; return ARM::VLD4d16_UPD;
+  case ARM::VLD4dWB_fixed_Asm_32: Spacing = 1; return ARM::VLD4d32_UPD;
+  case ARM::VLD4qWB_fixed_Asm_8:  Spacing = 2; return ARM::VLD4q8_UPD;
+  case ARM::VLD4qWB_fixed_Asm_16: Spacing = 2; return ARM::VLD4q16_UPD;
+  case ARM::VLD4qWB_fixed_Asm_32: Spacing = 2; return ARM::VLD4q32_UPD;
+  case ARM::VLD4dWB_register_Asm_8:  Spacing = 1; return ARM::VLD4d8_UPD;
+  case ARM::VLD4dWB_register_Asm_16: Spacing = 1; return ARM::VLD4d16_UPD;
+  case ARM::VLD4dWB_register_Asm_32: Spacing = 1; return ARM::VLD4d32_UPD;
+  case ARM::VLD4qWB_register_Asm_8:  Spacing = 2; return ARM::VLD4q8_UPD;
+  case ARM::VLD4qWB_register_Asm_16: Spacing = 2; return ARM::VLD4q16_UPD;
+  case ARM::VLD4qWB_register_Asm_32: Spacing = 2; return ARM::VLD4q32_UPD;
+  case ARM::VLD4dAsm_8:  Spacing = 1; return ARM::VLD4d8;
+  case ARM::VLD4dAsm_16: Spacing = 1; return ARM::VLD4d16;
+  case ARM::VLD4dAsm_32: Spacing = 1; return ARM::VLD4d32;
+  case ARM::VLD4qAsm_8:  Spacing = 2; return ARM::VLD4q8;
+  case ARM::VLD4qAsm_16: Spacing = 2; return ARM::VLD4q16;
+  case ARM::VLD4qAsm_32: Spacing = 2; return ARM::VLD4q32;
   }
 }
 
@@ -5848,6 +5868,83 @@ processInstruction(MCInst &Inst,
     return true;
   }
 
+  // VLD4 multiple 3-element structure instructions.
+  case ARM::VLD4dAsm_8:
+  case ARM::VLD4dAsm_16:
+  case ARM::VLD4dAsm_32:
+  case ARM::VLD4qAsm_8:
+  case ARM::VLD4qAsm_16:
+  case ARM::VLD4qAsm_32: {
+    MCInst TmpInst;
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVLDOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(0)); // Vd
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing));
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing * 2));
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing * 3));
+    TmpInst.addOperand(Inst.getOperand(1)); // Rn
+    TmpInst.addOperand(Inst.getOperand(2)); // alignment
+    TmpInst.addOperand(Inst.getOperand(3)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(4));
+    Inst = TmpInst;
+    return true;
+  }
+
+  case ARM::VLD4dWB_fixed_Asm_8:
+  case ARM::VLD4dWB_fixed_Asm_16:
+  case ARM::VLD4dWB_fixed_Asm_32:
+  case ARM::VLD4qWB_fixed_Asm_8:
+  case ARM::VLD4qWB_fixed_Asm_16:
+  case ARM::VLD4qWB_fixed_Asm_32: {
+    MCInst TmpInst;
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVLDOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(0)); // Vd
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing));
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing * 2));
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing * 3));
+    TmpInst.addOperand(Inst.getOperand(1)); // Rn
+    TmpInst.addOperand(Inst.getOperand(1)); // Rn_wb == tied Rn
+    TmpInst.addOperand(Inst.getOperand(2)); // alignment
+    TmpInst.addOperand(MCOperand::CreateReg(0)); // Rm
+    TmpInst.addOperand(Inst.getOperand(3)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(4));
+    Inst = TmpInst;
+    return true;
+  }
+
+  case ARM::VLD4dWB_register_Asm_8:
+  case ARM::VLD4dWB_register_Asm_16:
+  case ARM::VLD4dWB_register_Asm_32:
+  case ARM::VLD4qWB_register_Asm_8:
+  case ARM::VLD4qWB_register_Asm_16:
+  case ARM::VLD4qWB_register_Asm_32: {
+    MCInst TmpInst;
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVLDOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(0)); // Vd
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing));
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing * 2));
+    TmpInst.addOperand(MCOperand::CreateReg(Inst.getOperand(0).getReg() +
+                                            Spacing * 3));
+    TmpInst.addOperand(Inst.getOperand(1)); // Rn
+    TmpInst.addOperand(Inst.getOperand(1)); // Rn_wb == tied Rn
+    TmpInst.addOperand(Inst.getOperand(2)); // alignment
+    TmpInst.addOperand(Inst.getOperand(3)); // Rm
+    TmpInst.addOperand(Inst.getOperand(4)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(5));
+    Inst = TmpInst;
+    return true;
+  }
+
   // VST3 multiple 3-element structure instructions.
   case ARM::VST3dAsm_8:
   case ARM::VST3dAsm_16:
diff --git a/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp b/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
index 734654e344..27d5de81e9 100644
--- a/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
+++ b/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
@@ -1096,3 +1096,15 @@ void ARMInstPrinter::printVectorListThreeSpaced(const MCInst *MI,
     << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << ", "
     << getRegisterName(MI->getOperand(OpNum).getReg() + 4) << "}";
 }
+
+void ARMInstPrinter::printVectorListFourSpaced(const MCInst *MI,
+                                                unsigned OpNum,
+                                                raw_ostream &O) {
+  // Normally, it's not safe to use register enum values directly with
+  // addition to get the next register, but for VFP registers, the
+  // sort order is guaranteed because they're all of the form D<n>.
+  O << "{" << getRegisterName(MI->getOperand(OpNum).getReg()) << ", "
+    << getRegisterName(MI->getOperand(OpNum).getReg() + 2) << ", "
+    << getRegisterName(MI->getOperand(OpNum).getReg() + 4) << ", "
+    << getRegisterName(MI->getOperand(OpNum).getReg() + 6) << "}";
+}
diff --git a/lib/Target/ARM/InstPrinter/ARMInstPrinter.h b/lib/Target/ARM/InstPrinter/ARMInstPrinter.h
index e189477c23..4af116c832 100644
--- a/lib/Target/ARM/InstPrinter/ARMInstPrinter.h
+++ b/lib/Target/ARM/InstPrinter/ARMInstPrinter.h
@@ -145,6 +145,8 @@ public:
                                         raw_ostream &O);
   void printVectorListThreeSpaced(const MCInst *MI, unsigned OpNum,
                                   raw_ostream &O);
+  void printVectorListFourSpaced(const MCInst *MI, unsigned OpNum,
+                                  raw_ostream &O);
 };
 
 } // end namespace llvm
diff --git a/test/MC/ARM/neon-vld-encoding.s b/test/MC/ARM/neon-vld-encoding.s
index 1b44d333af..1dc2fdc14c 100644
--- a/test/MC/ARM/neon-vld-encoding.s
+++ b/test/MC/ARM/neon-vld-encoding.s
@@ -196,25 +196,45 @@
 @ CHECK: vld3.32	{d5, d7, d9}, [r4]!     @ encoding: [0x8d,0x55,0x24,0xf4]
 
 
-@	vld4.8	{d16, d17, d18, d19}, [r0, :64]
-@	vld4.16	{d16, d17, d18, d19}, [r0, :128]
-@	vld4.32	{d16, d17, d18, d19}, [r0, :256]
-@	vld4.8	{d16, d18, d20, d22}, [r0, :256]!
-@	vld4.8	{d17, d19, d21, d23}, [r0, :256]!
-@	vld4.16	{d16, d18, d20, d22}, [r0]!
-@	vld4.16	{d17, d19, d21, d23}, [r0]!
-@	vld4.32	{d16, d18, d20, d22}, [r0]!
-@	vld4.32	{d17, d19, d21, d23}, [r0]!
-
-@ FIXME: vld4.8	{d16, d17, d18, d19}, [r0, :64]@ encoding: [0x1f,0x00,0x60,0xf4]
-@ FIXME: vld4.16 {d16, d17, d18, d19}, [r0,:128]@ encoding:[0x6f,0x00,0x60,0xf4]
-@ FIXME: vld4.32 {d16, d17, d18, d19}, [r0,:256]@ encoding:[0xbf,0x00,0x60,0xf4]
-@ FIXME: vld4.8	{d16, d18, d20, d22}, [r0,:256]!@ encoding:[0x3d,0x01,0x60,0xf4]
-@ FIXME: vld4.8	{d17, d19, d21, d23}, [r0,:256]!@ encoding:[0x3d,0x11,0x60,0xf4]
-@ FIXME: vld4.16 {d16, d18, d20, d22}, [r0]! @ encoding: [0x4d,0x01,0x60,0xf4]
-@ FIXME: vld4.16 {d17, d19, d21, d23}, [r0]! @ encoding: [0x4d,0x11,0x60,0xf4]
-@ FIXME: vld4.32 {d16, d18, d20, d22}, [r0]! @ encoding: [0x8d,0x01,0x60,0xf4]
-@ FIXME: vld4.32 {d17, d19, d21, d23}, [r0]! @ encoding: [0x8d,0x11,0x60,0xf4]
+	vld4.8 {d16, d17, d18, d19}, [r1, :64]
+	vld4.16 {d16, d17, d18, d19}, [r2, :128]
+	vld4.32 {d16, d17, d18, d19}, [r3, :256]
+	vld4.8 {d17, d19, d21, d23}, [r5, :256]
+	vld4.16 {d17, d19, d21, d23}, [r7]
+	vld4.32 {d16, d18, d20, d22}, [r8]
+
+	vld4.s8 {d16, d17, d18, d19}, [r1, :64]!
+	vld4.s16 {d16, d17, d18, d19}, [r2, :128]!
+	vld4.s32 {d16, d17, d18, d19}, [r3, :256]!
+	vld4.u8 {d17, d19, d21, d23}, [r5, :256]!
+	vld4.u16 {d17, d19, d21, d23}, [r7]!
+	vld4.u32 {d16, d18, d20, d22}, [r8]!
+
+	vld4.p8 {d16, d17, d18, d19}, [r1, :64], r8
+	vld4.p16 {d16, d17, d18, d19}, [r2], r7
+	vld4.f32 {d16, d17, d18, d19}, [r3, :64], r5
+	vld4.i8 {d16, d18, d20, d22}, [r4, :256], r2
+	vld4.i16 {d16, d18, d20, d22}, [r6], r3
+	vld4.i32 {d17, d19, d21, d23}, [r9], r4
+
+@ CHECK: vld4.8 {d16, d17, d18, d19}, [r1, :64] @ encoding: [0x1f,0x00,0x61,0xf4]
+@ CHECK: vld4.16 {d16, d17, d18, d19}, [r2, :128] @ encoding: [0x6f,0x00,0x62,0xf4]
+@ CHECK: vld4.32 {d16, d17, d18, d19}, [r3, :256] @ encoding: [0xbf,0x00,0x63,0xf4]
+@ CHECK: vld4.8 {d17, d19, d21, d23}, [r5, :256] @ encoding: [0x3f,0x11,0x65,0xf4]
+@ CHECK: vld4.16 {d17, d19, d21, d23}, [r7] @ encoding: [0x4f,0x11,0x67,0xf4]
+@ CHECK: vld4.32 {d16, d18, d20, d22}, [r8] @ encoding: [0x8f,0x01,0x68,0xf4]
+@ CHECK: vld4.8 {d16, d17, d18, d19}, [r1, :64]! @ encoding: [0x1d,0x00,0x61,0xf4]
+@ CHECK: vld4.16 {d16, d17, d18, d19}, [r2, :128]! @ encoding: [0x6d,0x00,0x62,0xf4]
+@ CHECK: vld4.32 {d16, d17, d18, d19}, [r3, :256]! @ encoding: [0xbd,0x00,0x63,0xf4]
+@ CHECK: vld4.8 {d17, d19, d21, d23}, [r5, :256]! @ encoding: [0x3d,0x11,0x65,0xf4]
+@ CHECK: vld4.16 {d17, d19, d21, d23}, [r7]! @ encoding: [0x4d,0x11,0x67,0xf4]
+@ CHECK: vld4.32 {d16, d18, d20, d22}, [r8]! @ encoding: [0x8d,0x01,0x68,0xf4]
+@ CHECK: vld4.8 {d16, d17, d18, d19}, [r1, :64], r8 @ encoding: [0x18,0x00,0x61,0xf4]
+@ CHECK: vld4.16 {d16, d17, d18, d19}, [r2], r7 @ encoding: [0x47,0x00,0x62,0xf4]
+@ CHECK: vld4.32 {d16, d17, d18, d19}, [r3, :64], r5 @ encoding: [0x95,0x00,0x63,0xf4]
+@ CHECK: vld4.8 {d16, d18, d20, d22}, [r4, :256], r2 @ encoding: [0x32,0x01,0x64,0xf4]
+@ CHECK: vld4.16 {d16, d18, d20, d22}, [r6], r3 @ encoding: [0x43,0x01,0x66,0xf4]
+@ CHECK: vld4.32 {d17, d19, d21, d23}, [r9], r4 @ encoding: [0x84,0x11,0x69,0xf4]
 
 
 	vld1.8 {d4[]}, [r1]