[llvm] [RISCV] Refactor GPRF64 register class to make it usable for Zacas. (PR #77408)
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 9 00:03:54 PST 2024
https://github.com/topperc updated https://github.com/llvm/llvm-project/pull/77408
>From a69fbf642552ed3ad6dd0a726685f165c11f4a3d Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper at sifive.com>
Date: Mon, 8 Jan 2024 19:49:11 -0800
Subject: [PATCH 1/3] [RISCV] Refactor GPRF64 register class to make it usable
for Zacas.
-Rename to GPRPair.
-Rename registers to be named like X10_X11 instead of X10_PD. Except X0
which is now X0_Pair since it is not paired with X1.
-Use unknown size and offset for the subreg indices. This might
be a functional change, but does not affect any lit tests.
---
.../Target/RISCV/AsmParser/RISCVAsmParser.cpp | 2 +-
.../RISCV/Disassembler/RISCVDisassembler.cpp | 2 +-
.../Target/RISCV/RISCVExpandPseudoInsts.cpp | 12 ++++--
llvm/lib/Target/RISCV/RISCVISelLowering.cpp | 10 ++---
llvm/lib/Target/RISCV/RISCVInstrInfo.cpp | 17 ++++----
llvm/lib/Target/RISCV/RISCVInstrInfoD.td | 17 ++++----
llvm/lib/Target/RISCV/RISCVRegisterInfo.td | 42 +++++++++++--------
7 files changed, 57 insertions(+), 45 deletions(-)
diff --git a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
index d616aaeddf4114..4250950a917299 100644
--- a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
+++ b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
@@ -1295,7 +1295,7 @@ unsigned RISCVAsmParser::checkTargetMatchPredicate(MCInst &Inst) {
const MCInstrDesc &MCID = MII.get(Inst.getOpcode());
for (unsigned I = 0; I < MCID.NumOperands; ++I) {
- if (MCID.operands()[I].RegClass == RISCV::GPRPF64RegClassID) {
+ if (MCID.operands()[I].RegClass == RISCV::GPRPairRegClassID) {
const auto &Op = Inst.getOperand(I);
assert(Op.isReg());
diff --git a/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp b/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp
index ed80da14c79574..bc65cf2403b262 100644
--- a/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp
+++ b/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp
@@ -171,7 +171,7 @@ static DecodeStatus DecodeGPRCRegisterClass(MCInst &Inst, uint32_t RegNo,
return MCDisassembler::Success;
}
-static DecodeStatus DecodeGPRPF64RegisterClass(MCInst &Inst, uint32_t RegNo,
+static DecodeStatus DecodeGPRPairRegisterClass(MCInst &Inst, uint32_t RegNo,
uint64_t Address,
const MCDisassembler *Decoder) {
if (RegNo >= 32 || RegNo & 1)
diff --git a/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp b/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
index 24a13f93af880e..7592392de53b9c 100644
--- a/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
+++ b/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
@@ -300,8 +300,10 @@ bool RISCVExpandPseudo::expandRV32ZdinxStore(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MBBI) {
DebugLoc DL = MBBI->getDebugLoc();
const TargetRegisterInfo *TRI = STI->getRegisterInfo();
- Register Lo = TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_32);
- Register Hi = TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_32_hi);
+ Register Lo =
+ TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_gpr_even);
+ Register Hi =
+ TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_gpr_odd);
BuildMI(MBB, MBBI, DL, TII->get(RISCV::SW))
.addReg(Lo, getKillRegState(MBBI->getOperand(0).isKill()))
.addReg(MBBI->getOperand(1).getReg())
@@ -334,8 +336,10 @@ bool RISCVExpandPseudo::expandRV32ZdinxLoad(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MBBI) {
DebugLoc DL = MBBI->getDebugLoc();
const TargetRegisterInfo *TRI = STI->getRegisterInfo();
- Register Lo = TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_32);
- Register Hi = TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_32_hi);
+ Register Lo =
+ TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_gpr_even);
+ Register Hi =
+ TRI->getSubReg(MBBI->getOperand(0).getReg(), RISCV::sub_gpr_odd);
// If the register of operand 1 is equal to the Lo register, then swap the
// order of loading the Lo and Hi statements.
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 79c16cf4c4c361..abadd08a4d90d9 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -138,7 +138,7 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
if (Subtarget.is64Bit())
addRegisterClass(MVT::f64, &RISCV::GPRRegClass);
else
- addRegisterClass(MVT::f64, &RISCV::GPRPF64RegClass);
+ addRegisterClass(MVT::f64, &RISCV::GPRPairRegClass);
}
static const MVT::SimpleValueType BoolVecVTs[] = {
@@ -16345,7 +16345,7 @@ static MachineBasicBlock *emitSplitF64Pseudo(MachineInstr &MI,
Register SrcReg = MI.getOperand(2).getReg();
const TargetRegisterClass *SrcRC = MI.getOpcode() == RISCV::SplitF64Pseudo_INX
- ? &RISCV::GPRPF64RegClass
+ ? &RISCV::GPRPairRegClass
: &RISCV::FPR64RegClass;
int FI = MF.getInfo<RISCVMachineFunctionInfo>()->getMoveF64FrameIndex(MF);
@@ -16384,7 +16384,7 @@ static MachineBasicBlock *emitBuildPairF64Pseudo(MachineInstr &MI,
Register HiReg = MI.getOperand(2).getReg();
const TargetRegisterClass *DstRC =
- MI.getOpcode() == RISCV::BuildPairF64Pseudo_INX ? &RISCV::GPRPF64RegClass
+ MI.getOpcode() == RISCV::BuildPairF64Pseudo_INX ? &RISCV::GPRPairRegClass
: &RISCV::FPR64RegClass;
int FI = MF.getInfo<RISCVMachineFunctionInfo>()->getMoveF64FrameIndex(MF);
@@ -18751,7 +18751,7 @@ RISCVTargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
if (VT == MVT::f32 && Subtarget.hasStdExtZfinx())
return std::make_pair(0U, &RISCV::GPRF32RegClass);
if (VT == MVT::f64 && Subtarget.hasStdExtZdinx() && !Subtarget.is64Bit())
- return std::make_pair(0U, &RISCV::GPRPF64RegClass);
+ return std::make_pair(0U, &RISCV::GPRPairRegClass);
return std::make_pair(0U, &RISCV::GPRNoX0RegClass);
case 'f':
if (Subtarget.hasStdExtZfhmin() && VT == MVT::f16)
@@ -18933,7 +18933,7 @@ RISCVTargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
// Subtarget into account.
if (Res.second == &RISCV::GPRF16RegClass ||
Res.second == &RISCV::GPRF32RegClass ||
- Res.second == &RISCV::GPRPF64RegClass)
+ Res.second == &RISCV::GPRPairRegClass)
return std::make_pair(Res.first, &RISCV::GPRRegClass);
return Res;
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 7f6a045a7d042f..388e0db27046e2 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -414,15 +414,16 @@ void RISCVInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
return;
}
- if (RISCV::GPRPF64RegClass.contains(DstReg, SrcReg)) {
- // Emit an ADDI for both parts of GPRPF64.
+ if (RISCV::GPRPairRegClass.contains(DstReg, SrcReg)) {
+ // Emit an ADDI for both parts of GPRPair.
BuildMI(MBB, MBBI, DL, get(RISCV::ADDI),
- TRI->getSubReg(DstReg, RISCV::sub_32))
- .addReg(TRI->getSubReg(SrcReg, RISCV::sub_32), getKillRegState(KillSrc))
+ TRI->getSubReg(DstReg, RISCV::sub_gpr_even))
+ .addReg(TRI->getSubReg(SrcReg, RISCV::sub_gpr_even),
+ getKillRegState(KillSrc))
.addImm(0);
BuildMI(MBB, MBBI, DL, get(RISCV::ADDI),
- TRI->getSubReg(DstReg, RISCV::sub_32_hi))
- .addReg(TRI->getSubReg(SrcReg, RISCV::sub_32_hi),
+ TRI->getSubReg(DstReg, RISCV::sub_gpr_odd))
+ .addReg(TRI->getSubReg(SrcReg, RISCV::sub_gpr_odd),
getKillRegState(KillSrc))
.addImm(0);
return;
@@ -607,7 +608,7 @@ void RISCVInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
Opcode = TRI->getRegSizeInBits(RISCV::GPRRegClass) == 32 ?
RISCV::SW : RISCV::SD;
IsScalableVector = false;
- } else if (RISCV::GPRPF64RegClass.hasSubClassEq(RC)) {
+ } else if (RISCV::GPRPairRegClass.hasSubClassEq(RC)) {
Opcode = RISCV::PseudoRV32ZdinxSD;
IsScalableVector = false;
} else if (RISCV::FPR16RegClass.hasSubClassEq(RC)) {
@@ -690,7 +691,7 @@ void RISCVInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
Opcode = TRI->getRegSizeInBits(RISCV::GPRRegClass) == 32 ?
RISCV::LW : RISCV::LD;
IsScalableVector = false;
- } else if (RISCV::GPRPF64RegClass.hasSubClassEq(RC)) {
+ } else if (RISCV::GPRPairRegClass.hasSubClassEq(RC)) {
Opcode = RISCV::PseudoRV32ZdinxLD;
IsScalableVector = false;
} else if (RISCV::FPR16RegClass.hasSubClassEq(RC)) {
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoD.td b/llvm/lib/Target/RISCV/RISCVInstrInfoD.td
index 418421b2a556f7..6e89934b4052c2 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoD.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoD.td
@@ -33,8 +33,8 @@ def AddrRegImmINX : ComplexPattern<iPTR, 2, "SelectAddrRegImmINX">;
// Zdinx
-def GPRPF64AsFPR : AsmOperandClass {
- let Name = "GPRPF64AsFPR";
+def GPRPairAsFPR : AsmOperandClass {
+ let Name = "GPRPairAsFPR";
let ParserMethod = "parseGPRAsFPR";
let PredicateMethod = "isGPRAsFPR";
let RenderMethod = "addRegOperands";
@@ -52,8 +52,9 @@ def FPR64INX : RegisterOperand<GPR> {
let DecoderMethod = "DecodeGPRRegisterClass";
}
-def FPR64IN32X : RegisterOperand<GPRPF64> {
- let ParserMatchClass = GPRPF64AsFPR;
+def FPR64IN32X : RegisterOperand<GPRPair> {
+ let ParserMatchClass = GPRPairAsFPR;
+ let DecoderMethod = "DecodeGPRPairRegisterClass";
}
def DExt : ExtInfo<"", "", [HasStdExtD], f64, FPR64, FPR32, FPR64, ?>;
@@ -515,15 +516,15 @@ def PseudoFROUND_D_IN32X : PseudoFROUND<FPR64IN32X, f64>;
/// Loads
let isCall = 0, mayLoad = 1, mayStore = 0, Size = 8, isCodeGenOnly = 1 in
-def PseudoRV32ZdinxLD : Pseudo<(outs GPRPF64:$dst), (ins GPR:$rs1, simm12:$imm12), []>;
+def PseudoRV32ZdinxLD : Pseudo<(outs GPRPair:$dst), (ins GPR:$rs1, simm12:$imm12), []>;
def : Pat<(f64 (load (AddrRegImmINX (XLenVT GPR:$rs1), simm12:$imm12))),
(PseudoRV32ZdinxLD GPR:$rs1, simm12:$imm12)>;
/// Stores
let isCall = 0, mayLoad = 0, mayStore = 1, Size = 8, isCodeGenOnly = 1 in
-def PseudoRV32ZdinxSD : Pseudo<(outs), (ins GPRPF64:$rs2, GPRNoX0:$rs1, simm12:$imm12), []>;
-def : Pat<(store (f64 GPRPF64:$rs2), (AddrRegImmINX (XLenVT GPR:$rs1), simm12:$imm12)),
- (PseudoRV32ZdinxSD GPRPF64:$rs2, GPR:$rs1, simm12:$imm12)>;
+def PseudoRV32ZdinxSD : Pseudo<(outs), (ins GPRPair:$rs2, GPRNoX0:$rs1, simm12:$imm12), []>;
+def : Pat<(store (f64 GPRPair:$rs2), (AddrRegImmINX (XLenVT GPR:$rs1), simm12:$imm12)),
+ (PseudoRV32ZdinxSD GPRPair:$rs2, GPR:$rs1, simm12:$imm12)>;
/// Pseudo-instructions needed for the soft-float ABI with RV32D
diff --git a/llvm/lib/Target/RISCV/RISCVRegisterInfo.td b/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
index a59d058382fe58..8d19d8e935a929 100644
--- a/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
+++ b/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
@@ -63,7 +63,10 @@ def sub_vrm1_5 : ComposedSubRegIndex<sub_vrm2_2, sub_vrm1_1>;
def sub_vrm1_6 : ComposedSubRegIndex<sub_vrm2_3, sub_vrm1_0>;
def sub_vrm1_7 : ComposedSubRegIndex<sub_vrm2_3, sub_vrm1_1>;
-def sub_32_hi : SubRegIndex<32, 32>;
+// GPR sizes change with HwMode.
+// FIXME: Support HwMode in SubRegIndex?
+def sub_gpr_even : SubRegIndex<-1>;
+def sub_gpr_odd : SubRegIndex<-1, -1>;
} // Namespace = "RISCV"
// Integer registers
@@ -546,33 +549,36 @@ def DUMMY_REG_PAIR_WITH_X0 : RISCVReg<0, "0">;
def GPRAll : GPRRegisterClass<(add GPR, DUMMY_REG_PAIR_WITH_X0)>;
let RegAltNameIndices = [ABIRegAltName] in {
- def X0_PD : RISCVRegWithSubRegs<0, X0.AsmName,
- [X0, DUMMY_REG_PAIR_WITH_X0],
- X0.AltNames> {
- let SubRegIndices = [sub_32, sub_32_hi];
+ def X0_Pair : RISCVRegWithSubRegs<0, X0.AsmName,
+ [X0, DUMMY_REG_PAIR_WITH_X0],
+ X0.AltNames> {
+ let SubRegIndices = [sub_gpr_even, sub_gpr_odd];
let CoveredBySubRegs = 1;
}
foreach I = 1-15 in {
defvar Index = !shl(I, 1);
+ defvar IndexP1 = !add(Index, 1);
defvar Reg = !cast<Register>("X"#Index);
- defvar RegP1 = !cast<Register>("X"#!add(Index,1));
- def X#Index#_PD : RISCVRegWithSubRegs<Index, Reg.AsmName,
- [Reg, RegP1],
- Reg.AltNames> {
- let SubRegIndices = [sub_32, sub_32_hi];
+ defvar RegP1 = !cast<Register>("X"#IndexP1);
+ def "X" # Index #"_X" # IndexP1 : RISCVRegWithSubRegs<Index,
+ Reg.AsmName,
+ [Reg, RegP1],
+ Reg.AltNames> {
+ let SubRegIndices = [sub_gpr_even, sub_gpr_odd];
let CoveredBySubRegs = 1;
}
}
}
-let RegInfos = RegInfoByHwMode<[RV64], [RegInfo<64, 64, 64>]> in
-def GPRPF64 : RegisterClass<"RISCV", [f64], 64, (add
- X10_PD, X12_PD, X14_PD, X16_PD,
- X6_PD,
- X28_PD, X30_PD,
- X8_PD,
- X18_PD, X20_PD, X22_PD, X24_PD, X26_PD,
- X0_PD, X2_PD, X4_PD
+let RegInfos = RegInfoByHwMode<[RV64], [RegInfo<64, 64, 64>]>,
+ DecoderMethod = "DecodeGPRPairRegisterClass" in
+def GPRPair : RegisterClass<"RISCV", [f64], 64, (add
+ X10_X11, X12_X13, X14_X15, X16_X17,
+ X6_X7,
+ X28_X29, X30_X31,
+ X8_X9,
+ X18_X19, X20_X21, X22_X23, X24_X25, X26_X27,
+ X0_Pair, X2_X3, X4_X5
)>;
// The register class is added for inline assembly for vector mask types.
>From 71d7cea73b1e1373a5fb7ec04d2ac07a23bc5be3 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper at sifive.com>
Date: Mon, 8 Jan 2024 20:31:08 -0800
Subject: [PATCH 2/3] fixup! remove unneeded change
---
llvm/lib/Target/RISCV/RISCVInstrInfoD.td | 1 -
1 file changed, 1 deletion(-)
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoD.td b/llvm/lib/Target/RISCV/RISCVInstrInfoD.td
index 6e89934b4052c2..fec43d814098ce 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoD.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoD.td
@@ -54,7 +54,6 @@ def FPR64INX : RegisterOperand<GPR> {
def FPR64IN32X : RegisterOperand<GPRPair> {
let ParserMatchClass = GPRPairAsFPR;
- let DecoderMethod = "DecodeGPRPairRegisterClass";
}
def DExt : ExtInfo<"", "", [HasStdExtD], f64, FPR64, FPR32, FPR64, ?>;
>From 47a8943cc17c8ff4b2cd40a42ad29c43deb12d10 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper at sifive.com>
Date: Tue, 9 Jan 2024 00:03:36 -0800
Subject: [PATCH 3/3] fixup! Update VT and RegInfos.
---
llvm/lib/Target/RISCV/RISCVRegisterInfo.td | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/llvm/lib/Target/RISCV/RISCVRegisterInfo.td b/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
index 8d19d8e935a929..5a4d8c4cfece7f 100644
--- a/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
+++ b/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
@@ -121,6 +121,8 @@ def XLenVT : ValueTypeByHwMode<[RV32, RV64],
// Allow f64 in GPR for ZDINX on RV64.
def XLenFVT : ValueTypeByHwMode<[RV64],
[f64]>;
+def XLenPairFVT : ValueTypeByHwMode<[RV32],
+ [f64]>;
def XLenRI : RegInfoByHwMode<
[RV32, RV64],
[RegInfo<32,32,32>, RegInfo<64,64,64>]>;
@@ -570,9 +572,10 @@ let RegAltNameIndices = [ABIRegAltName] in {
}
}
-let RegInfos = RegInfoByHwMode<[RV64], [RegInfo<64, 64, 64>]>,
+let RegInfos = RegInfoByHwMode<[RV32, RV64],
+ [RegInfo<64, 64, 64>, RegInfo<128, 128, 128>]>,
DecoderMethod = "DecodeGPRPairRegisterClass" in
-def GPRPair : RegisterClass<"RISCV", [f64], 64, (add
+def GPRPair : RegisterClass<"RISCV", [XLenPairFVT], 64, (add
X10_X11, X12_X13, X14_X15, X16_X17,
X6_X7,
X28_X29, X30_X31,
More information about the llvm-commits
mailing list