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
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 @@
(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 @@
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 @@
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 @@
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 @@
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: