NEON VST4(one lane) assembly parsing and encoding.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@148836 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 181 insertions, 11 deletions

diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 702de8b..4040db9 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -6364,6 +6364,55 @@ def VLD4qWB_register_Asm_32 :
                   (ins VecListFourQ:$list, addrmode6:$addr,
                        rGPR:$Rm, pred:$p)>;
 
+// VST4 single-lane pseudo-instructions. These need special handling for
+// the lane index that an InstAlias can't handle, so we use these instead.
+def VST4LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
+               (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
+               (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
+               (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
+               (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
+               (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+
+def VST4LNdWB_fixed_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
+               (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNdWB_fixed_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
+               (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNdWB_fixed_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
+               (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNqWB_fixed_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
+               (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNqWB_fixed_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
+               (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+def VST4LNdWB_register_Asm_8 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
+                  (ins VecListFourDByteIndexed:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4LNdWB_register_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
+                  (ins VecListFourDHWordIndexed:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4LNdWB_register_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
+                  (ins VecListFourDWordIndexed:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4LNqWB_register_Asm_16 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
+                  (ins VecListFourQHWordIndexed:$list, addrmode6:$addr,
+                       rGPR:$Rm, pred:$p)>;
+def VST4LNqWB_register_Asm_32 :
+        NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
+                  (ins VecListFourQWordIndexed:$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.
diff --git a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
index 9bc4e60..a3657db 100644
--- a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -5273,6 +5273,23 @@ static unsigned getRealVSTOpcode(unsigned Opc, unsigned &Spacing) {
   case ARM::VST3qAsm_16: Spacing = 2; return ARM::VST3q16;
   case ARM::VST3qAsm_32: Spacing = 2; return ARM::VST3q32;
 
+  // VST4LN
+  case ARM::VST4LNdWB_fixed_Asm_8:  Spacing = 1; return ARM::VST4LNd8_UPD;
+  case ARM::VST4LNdWB_fixed_Asm_16: Spacing = 1; return ARM::VST4LNd16_UPD;
+  case ARM::VST4LNdWB_fixed_Asm_32: Spacing = 1; return ARM::VST4LNd32_UPD;
+  case ARM::VST4LNqWB_fixed_Asm_16: Spacing = 1; return ARM::VST4LNq16_UPD;
+  case ARM::VST4LNqWB_fixed_Asm_32: Spacing = 2; return ARM::VST4LNq32_UPD;
+  case ARM::VST4LNdWB_register_Asm_8:  Spacing = 1; return ARM::VST4LNd8_UPD;
+  case ARM::VST4LNdWB_register_Asm_16: Spacing = 1; return ARM::VST4LNd16_UPD;
+  case ARM::VST4LNdWB_register_Asm_32: Spacing = 1; return ARM::VST4LNd32_UPD;
+  case ARM::VST4LNqWB_register_Asm_16: Spacing = 2; return ARM::VST4LNq16_UPD;
+  case ARM::VST4LNqWB_register_Asm_32: Spacing = 2; return ARM::VST4LNq32_UPD;
+  case ARM::VST4LNdAsm_8:  Spacing = 1; return ARM::VST4LNd8;
+  case ARM::VST4LNdAsm_16: Spacing = 1; return ARM::VST4LNd16;
+  case ARM::VST4LNdAsm_32: Spacing = 1; return ARM::VST4LNd32;
+  case ARM::VST4LNqAsm_16: Spacing = 2; return ARM::VST4LNq16;
+  case ARM::VST4LNqAsm_32: Spacing = 2; return ARM::VST4LNq32;
+
   // 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;
@@ -5493,6 +5510,34 @@ processInstruction(MCInst &Inst,
     return true;
   }
 
+  case ARM::VST4LNdWB_register_Asm_8:
+  case ARM::VST4LNdWB_register_Asm_16:
+  case ARM::VST4LNdWB_register_Asm_32:
+  case ARM::VST4LNqWB_register_Asm_16:
+  case ARM::VST4LNqWB_register_Asm_32: {
+    MCInst TmpInst;
+    // Shuffle the operands around so the lane index operand is in the
+    // right place.
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVSTOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(2)); // Rn_wb
+    TmpInst.addOperand(Inst.getOperand(2)); // Rn
+    TmpInst.addOperand(Inst.getOperand(3)); // alignment
+    TmpInst.addOperand(Inst.getOperand(4)); // 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(1)); // lane
+    TmpInst.addOperand(Inst.getOperand(5)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(6));
+    Inst = TmpInst;
+    return true;
+  }
+
   case ARM::VST1LNdWB_fixed_Asm_8:
   case ARM::VST1LNdWB_fixed_Asm_16:
   case ARM::VST1LNdWB_fixed_Asm_32: {
@@ -5563,6 +5608,34 @@ processInstruction(MCInst &Inst,
     return true;
   }
 
+  case ARM::VST4LNdWB_fixed_Asm_8:
+  case ARM::VST4LNdWB_fixed_Asm_16:
+  case ARM::VST4LNdWB_fixed_Asm_32:
+  case ARM::VST4LNqWB_fixed_Asm_16:
+  case ARM::VST4LNqWB_fixed_Asm_32: {
+    MCInst TmpInst;
+    // Shuffle the operands around so the lane index operand is in the
+    // right place.
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVSTOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(2)); // Rn_wb
+    TmpInst.addOperand(Inst.getOperand(2)); // Rn
+    TmpInst.addOperand(Inst.getOperand(3)); // 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(1)); // lane
+    TmpInst.addOperand(Inst.getOperand(4)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(5));
+    Inst = TmpInst;
+    return true;
+  }
+
   case ARM::VST1LNdAsm_8:
   case ARM::VST1LNdAsm_16:
   case ARM::VST1LNdAsm_32: {
@@ -5627,6 +5700,32 @@ processInstruction(MCInst &Inst,
     return true;
   }
 
+  case ARM::VST4LNdAsm_8:
+  case ARM::VST4LNdAsm_16:
+  case ARM::VST4LNdAsm_32:
+  case ARM::VST4LNqAsm_16:
+  case ARM::VST4LNqAsm_32: {
+    MCInst TmpInst;
+    // Shuffle the operands around so the lane index operand is in the
+    // right place.
+    unsigned Spacing;
+    TmpInst.setOpcode(getRealVSTOpcode(Inst.getOpcode(), Spacing));
+    TmpInst.addOperand(Inst.getOperand(2)); // Rn
+    TmpInst.addOperand(Inst.getOperand(3)); // 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(1)); // lane
+    TmpInst.addOperand(Inst.getOperand(4)); // CondCode
+    TmpInst.addOperand(Inst.getOperand(5));
+    Inst = TmpInst;
+    return true;
+  }
+
   // Handle NEON VLD complex aliases.
   case ARM::VLD1LNdWB_register_Asm_8:
   case ARM::VLD1LNdWB_register_Asm_16:
diff --git a/test/MC/ARM/neon-vst-encoding.s b/test/MC/ARM/neon-vst-encoding.s
index ba2620f..2b14d37 100644
--- a/test/MC/ARM/neon-vst-encoding.s
+++ b/test/MC/ARM/neon-vst-encoding.s
@@ -210,17 +210,39 @@
 @ CHECK: vst3.32	{d5[1], d7[1], d9[1]}, [r4]! @ encoding: [0xcd,0x5a,0x84,0xf4]
 
 
-@	vst4.8	{d16[1], d17[1], d18[1], d19[1]}, [r0, :32]
-@	vst4.16	{d16[1], d17[1], d18[1], d19[1]}, [r0]
-@	vst4.32	{d16[1], d17[1], d18[1], d19[1]}, [r0, :128]
-@	vst4.16	{d17[3], d19[3], d21[3], d23[3]}, [r0, :64]
-@	vst4.32	{d17[0], d19[0], d21[0], d23[0]}, [r0]
-
-@ FIXME: vst4.8	{d16[1], d17[1], d18[1], d19[1]}, [r0, :32] @ encoding: [0x3f,0x03,0xc0,0xf4]
-@ FIXME: vst4.16 {d16[1], d17[1], d18[1], d19[1]}, [r0] @ encoding: [0x4f,0x07,0xc0,0xf4]
-@ FIXME: vst4.32 {d16[1], d17[1], d18[1], d19[1]}, [r0, :128] @ encoding: [0xaf,0x0b,0xc0,0xf4]
-@ FIXME: vst4.16 {d17[3], d19[3], d21[3], d23[3]}, [r0, :64] @ encoding: [0xff,0x17,0xc0,0xf4]
-@ FIXME: vst4.32 {d17[0], d19[0], d21[0], d23[0]}, [r0] @ encoding: [0x4f,0x1b,0xc0,0xf4]
+	vst4.8 {d16[1], d17[1], d18[1], d19[1]}, [r1]
+	vst4.16 {d16[1], d17[1], d18[1], d19[1]}, [r2]
+	vst4.32 {d16[1], d17[1], d18[1], d19[1]}, [r3]
+	vst4.16 {d17[1], d19[1], d21[1], d23[1]}, [r7]
+	vst4.32 {d16[1], d18[1], d20[1], d22[1]}, [r8]
+
+	vst4.s8 {d16[1], d17[1], d18[1], d19[1]}, [r1, :32]!
+	vst4.s16 {d16[1], d17[1], d18[1], d19[1]}, [r2, :64]!
+	vst4.s32 {d16[1], d17[1], d18[1], d19[1]}, [r3, :128]!
+	vst4.u16 {d17[1], d19[1], d21[1], d23[1]}, [r7]!
+	vst4.u32 {d16[1], d18[1], d20[1], d22[1]}, [r8]!
+
+	vst4.p8 {d16[1], d17[1], d18[1], d19[1]}, [r1, :32], r8
+	vst4.p16 {d16[1], d17[1], d18[1], d19[1]}, [r2], r7
+	vst4.f32 {d16[1], d17[1], d18[1], d19[1]}, [r3, :64], r5
+	vst4.i16 {d16[1], d18[1], d20[1], d22[1]}, [r6], r3
+	vst4.i32 {d17[1], d19[1], d21[1], d23[1]}, [r9], r4
+
+@ CHECK: vst4.8	{d16[1], d17[1], d18[1], d19[1]}, [r1] @ encoding: [0x2f,0x03,0xc1,0xf4]
+@ CHECK: vst4.16 {d16[1], d17[1], d18[1], d19[1]}, [r2] @ encoding: [0x4f,0x07,0xc2,0xf4]
+@ CHECK: vst4.32 {d16[1], d17[1], d18[1], d19[1]}, [r3] @ encoding: [0x8f,0x0b,0xc3,0xf4]
+@ CHECK: vst4.16 {d17[1], d19[1], d21[1], d23[1]}, [r7] @ encoding: [0x6f,0x17,0xc7,0xf4]
+@ CHECK: vst4.32 {d16[1], d18[1], d20[1], d22[1]}, [r8] @ encoding: [0xcf,0x0b,0xc8,0xf4]
+@ CHECK: vst4.8	{d16[1], d17[1], d18[1], d19[1]}, [r1, :32]! @ encoding: [0x3d,0x03,0xc1,0xf4]
+@ CHECK: vst4.16 {d16[1], d17[1], d18[1], d19[1]}, [r2, :64]! @ encoding: [0x5d,0x07,0xc2,0xf4]
+@ CHECK: vst4.32 {d16[1], d17[1], d18[1], d19[1]}, [r3, :128]! @ encoding: [0xad,0x0b,0xc3,0xf4]
+@ CHECK: vst4.16 {d17[1], d18[1], d19[1], d20[1]}, [r7]! @ encoding: [0x6d,0x17,0xc7,0xf4]
+@ CHECK: vst4.32 {d16[1], d18[1], d20[1], d22[1]}, [r8]! @ encoding: [0xcd,0x0b,0xc8,0xf4]
+@ CHECK: vst4.8	{d16[1], d17[1], d18[1], d19[1]}, [r1, :32], r8 @ encoding: [0x38,0x03,0xc1,0xf4]
+@ CHECK: vst4.16 {d16[1], d17[1], d18[1], d19[1]}, [r2], r7 @ encoding: [0x47,0x07,0xc2,0xf4]
+@ CHECK: vst4.32 {d16[1], d17[1], d18[1], d19[1]}, [r3, :64], r5 @ encoding: [0x95,0x0b,0xc3,0xf4]
+@ CHECK: vst4.16 {d16[1], d18[1], d20[1], d22[1]}, [r6], r3 @ encoding: [0x63,0x07,0xc6,0xf4]
+@ CHECK: vst4.32 {d17[1], d19[1], d21[1], d23[1]}, [r9], r4 @ encoding: [0xc4,0x1b,0xc9,0xf4]
 
 
 @ Spot-check additional size-suffix aliases.