NEON VST4(multiple 4 element structures) assembly parsing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@148764 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 196 insertions, 17 deletions

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 4ec3bc82e0..5d237a4566 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -6264,6 +6264,66 @@ def VLD4qWB_register_Asm_32 :
                   (ins VecListFourQ:$list, addrmode6:$addr,
                        rGPR:$Rm, pred:$p)>;
 
+
+// VST4 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 VST4dAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VST4dAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VST4dAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VST4qAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VST4qAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VST4qAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+
+def VST4dWB_fixed_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VST4dWB_fixed_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VST4dWB_fixed_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
+               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+def VST4qWB_fixed_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VST4qWB_fixed_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VST4qWB_fixed_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
+               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+def VST4dWB_register_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
+                  (ins VecListFourD:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4dWB_register_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
+                  (ins VecListFourD:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4dWB_register_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
+                  (ins VecListFourD:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4qWB_register_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
+                  (ins VecListFourQ:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4qWB_register_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
+                  (ins VecListFourQ:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4qWB_register_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${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 81c060f616..fb03a45251 100644
--- a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -5247,6 +5247,26 @@ static unsigned getRealVSTOpcode(unsigned Opc, unsigned &Spacing) {
   case ARM::VST3qAsm_8:  Spacing = 2; return ARM::VST3q8;
   case ARM::VST3qAsm_16: Spacing = 2; return ARM::VST3q16;
   case ARM::VST3qAsm_32: Spacing = 2; return ARM::VST3q32;
+
+  // VST4
+  case ARM::VST4dWB_fixed_Asm_8:  Spacing = 1; return ARM::VST4d8_UPD;
+  case ARM::VST4dWB_fixed_Asm_16: Spacing = 1; return ARM::VST4d16_UPD;
+  case ARM::VST4dWB_fixed_Asm_32: Spacing = 1; return ARM::VST4d32_UPD;
+  case ARM::VST4qWB_fixed_Asm_8:  Spacing = 2; return ARM::VST4q8_UPD;
+  case ARM::VST4qWB_fixed_Asm_16: Spacing = 2; return ARM::VST4q16_UPD;
+  case ARM::VST4qWB_fixed_Asm_32: Spacing = 2; return ARM::VST4q32_UPD;
+  case ARM::VST4dWB_register_Asm_8:  Spacing = 1; return ARM::VST4d8_UPD;
+  case ARM::VST4dWB_register_Asm_16: Spacing = 1; return ARM::VST4d16_UPD;
+  case ARM::VST4dWB_register_Asm_32: Spacing = 1; return ARM::VST4d32_UPD;
+  case ARM::VST4qWB_register_Asm_8:  Spacing = 2; return ARM::VST4q8_UPD;
+  case ARM::VST4qWB_register_Asm_16: Spacing = 2; return ARM::VST4q16_UPD;
+  case ARM::VST4qWB_register_Asm_32: Spacing = 2; return ARM::VST4q32_UPD;
+  case ARM::VST4dAsm_8:  Spacing = 1; return ARM::VST4d8;
+  case ARM::VST4dAsm_16: Spacing = 1; return ARM::VST4d16;
+  case ARM::VST4dAsm_32: Spacing = 1; return ARM::VST4d32;
+  case ARM::VST4qAsm_8:  Spacing = 2; return ARM::VST4q8;
+  case ARM::VST4qAsm_16: Spacing = 2; return ARM::VST4q16;
+  case ARM::VST4qAsm_32: Spacing = 2; return ARM::VST4q32;
   }
 }
 
@@ -6016,6 +6036,83 @@ processInstruction(MCInst &Inst,
     return true;
   }
 
+  // VST4 multiple 3-element structure instructions.
+  case ARM::VST4dAsm_8:
+  case ARM::VST4dAsm_16:
+  case ARM::VST4dAsm_32:
+  case ARM::VST4qAsm_8:
+  case ARM::VST4qAsm_16:
+  case ARM::VST4qAsm_32: {
+    MCInst TmpInst;
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVSTOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(1)); // Rn
+    TmpInst.addOperand(Inst.getOperand(2)); // alignment
+    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(3)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(4));
+    Inst = TmpInst;
+    return true;
+  }
+
+  case ARM::VST4dWB_fixed_Asm_8:
+  case ARM::VST4dWB_fixed_Asm_16:
+  case ARM::VST4dWB_fixed_Asm_32:
+  case ARM::VST4qWB_fixed_Asm_8:
+  case ARM::VST4qWB_fixed_Asm_16:
+  case ARM::VST4qWB_fixed_Asm_32: {
+    MCInst TmpInst;
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVSTOpcode(Inst.getOpcode(), Spacing));
+    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(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(3)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(4));
+    Inst = TmpInst;
+    return true;
+  }
+
+  case ARM::VST4dWB_register_Asm_8:
+  case ARM::VST4dWB_register_Asm_16:
+  case ARM::VST4dWB_register_Asm_32:
+  case ARM::VST4qWB_register_Asm_8:
+  case ARM::VST4qWB_register_Asm_16:
+  case ARM::VST4qWB_register_Asm_32: {
+    MCInst TmpInst;
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVSTOpcode(Inst.getOpcode(), Spacing));
+    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(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(4)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(5));
+    Inst = TmpInst;
+    return true;
+  }
+
   // Handle the Thumb2 mode MOV complex aliases.
   case ARM::t2MOVsr:
   case ARM::t2MOVSsr: {
diff --git a/test/MC/ARM/neon-vst-encoding.s b/test/MC/ARM/neon-vst-encoding.s
index fad4ae87de..ba2620ff93 100644
--- a/test/MC/ARM/neon-vst-encoding.s
+++ b/test/MC/ARM/neon-vst-encoding.s
@@ -99,23 +99,45 @@
 @ CHECK: vst3.32	{d5, d7, d9}, [r4]!     @ encoding: [0x8d,0x55,0x04,0xf4]
 
 
-@	vst4.8	{d16, d17, d18, d19}, [r0, :64]
-@	vst4.16	{d16, d17, d18, d19}, [r0, :128]
-@	vst4.8	{d16, d18, d20, d22}, [r0, :256]!
-@	vst4.8	{d17, d19, d21, d23}, [r0, :256]!
-@	vst4.16	{d16, d18, d20, d22}, [r0]!
-@	vst4.16	{d17, d19, d21, d23}, [r0]!
-@	vst4.32	{d16, d18, d20, d22}, [r0]!
-@	vst4.32	{d17, d19, d21, d23}, [r0]!
-
-@ FIXME: vst4.8	{d16, d17, d18, d19}, [r0, :64] @ encoding: [0x1f,0x00,0x40,0xf4]
-@ FIXME: vst4.16 {d16, d17, d18, d19}, [r0, :128] @ encoding: [0x6f,0x00,0x40,0xf4]
-@ FIXME: vst4.8	{d16, d18, d20, d22}, [r0, :256]! @ encoding: [0x3d,0x01,0x40,0xf4]
-@ FIXME: vst4.8	{d17, d19, d21, d23}, [r0, :256]! @ encoding: [0x3d,0x11,0x40,0xf4]
-@ FIXME: vst4.16 {d16, d18, d20, d22}, [r0]! @ encoding: [0x4d,0x01,0x40,0xf4]
-@ FIXME: vst4.16 {d17, d19, d21, d23}, [r0]! @ encoding: [0x4d,0x11,0x40,0xf4]
-@ FIXME: vst4.32 {d16, d18, d20, d22}, [r0]! @ encoding: [0x8d,0x01,0x40,0xf4]
-@ FIXME: vst4.32 {d17, d19, d21, d23}, [r0]! @ encoding: [0x8d,0x11,0x40,0xf4]
+	vst4.8 {d16, d17, d18, d19}, [r1, :64]
+	vst4.16 {d16, d17, d18, d19}, [r2, :128]
+	vst4.32 {d16, d17, d18, d19}, [r3, :256]
+	vst4.8 {d17, d19, d21, d23}, [r5, :256]
+	vst4.16 {d17, d19, d21, d23}, [r7]
+	vst4.32 {d16, d18, d20, d22}, [r8]
+
+	vst4.s8 {d16, d17, d18, d19}, [r1, :64]!
+	vst4.s16 {d16, d17, d18, d19}, [r2, :128]!
+	vst4.s32 {d16, d17, d18, d19}, [r3, :256]!
+	vst4.u8 {d17, d19, d21, d23}, [r5, :256]!
+	vst4.u16 {d17, d19, d21, d23}, [r7]!
+	vst4.u32 {d16, d18, d20, d22}, [r8]!
+
+	vst4.p8 {d16, d17, d18, d19}, [r1, :64], r8
+	vst4.p16 {d16, d17, d18, d19}, [r2], r7
+	vst4.f32 {d16, d17, d18, d19}, [r3, :64], r5
+	vst4.i8 {d16, d18, d20, d22}, [r4, :256], r2
+	vst4.i16 {d16, d18, d20, d22}, [r6], r3
+	vst4.i32 {d17, d19, d21, d23}, [r9], r4
+
+@ CHECK: vst4.8 {d16, d17, d18, d19}, [r1, :64] @ encoding: [0x1f,0x00,0x41,0xf4]
+@ CHECK: vst4.16 {d16, d17, d18, d19}, [r2, :128] @ encoding: [0x6f,0x00,0x42,0xf4]
+@ CHECK: vst4.32 {d16, d17, d18, d19}, [r3, :256] @ encoding: [0xbf,0x00,0x43,0xf4]
+@ CHECK: vst4.8 {d17, d19, d21, d23}, [r5, :256] @ encoding: [0x3f,0x11,0x45,0xf4]
+@ CHECK: vst4.16 {d17, d19, d21, d23}, [r7] @ encoding: [0x4f,0x11,0x47,0xf4]
+@ CHECK: vst4.32 {d16, d18, d20, d22}, [r8] @ encoding: [0x8f,0x01,0x48,0xf4]
+@ CHECK: vst4.8 {d16, d17, d18, d19}, [r1, :64]! @ encoding: [0x1d,0x00,0x41,0xf4]
+@ CHECK: vst4.16 {d16, d17, d18, d19}, [r2, :128]! @ encoding: [0x6d,0x00,0x42,0xf4]
+@ CHECK: vst4.32 {d16, d17, d18, d19}, [r3, :256]! @ encoding: [0xbd,0x00,0x43,0xf4]
+@ CHECK: vst4.8 {d17, d19, d21, d23}, [r5, :256]! @ encoding: [0x3d,0x11,0x45,0xf4]
+@ CHECK: vst4.16 {d17, d19, d21, d23}, [r7]! @ encoding: [0x4d,0x11,0x47,0xf4]
+@ CHECK: vst4.32 {d16, d18, d20, d22}, [r8]! @ encoding: [0x8d,0x01,0x48,0xf4]
+@ CHECK: vst4.8 {d16, d17, d18, d19}, [r1, :64], r8 @ encoding: [0x18,0x00,0x41,0xf4]
+@ CHECK: vst4.16 {d16, d17, d18, d19}, [r2], r7 @ encoding: [0x47,0x00,0x42,0xf4]
+@ CHECK: vst4.32 {d16, d17, d18, d19}, [r3, :64], r5 @ encoding: [0x95,0x00,0x43,0xf4]
+@ CHECK: vst4.8 {d16, d18, d20, d22}, [r4, :256], r2 @ encoding: [0x32,0x01,0x44,0xf4]
+@ CHECK: vst4.16 {d16, d18, d20, d22}, [r6], r3 @ encoding: [0x43,0x01,0x46,0xf4]
+@ CHECK: vst4.32 {d17, d19, d21, d23}, [r9], r4 @ encoding: [0x84,0x11,0x49,0xf4]
 
 
 	vst2.8	{d16[1], d17[1]}, [r0, :16]