[RISCV] Codegen support for XCVmem extension #76916
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-backend-risc-v Author: Liao Chunyu (ChunyuLiao) ChangesAll post-Increment load/store, register-register load/store spec: Full diff: https://github.com/llvm/llvm-project/pull/76916.diff 6 Files Affected:
diff --git a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
index 4759aa951664c7..b56a9a63bae51a 100644
--- a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
+++ b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
@@ -1210,7 +1210,7 @@ struct RISCVOperand final : public MCParsedAsmOperand {
}
void addRegRegOperands(MCInst &Inst, unsigned N) const {
- assert(N == 1 && "Invalid number of operands!");
+ assert(N == 2 && "Invalid number of operands!");
Inst.addOperand(MCOperand::createReg(RegReg.Reg1));
Inst.addOperand(MCOperand::createReg(RegReg.Reg2));
}
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
index bfa3bf3cc74e2b..3ba3000145784a 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
@@ -1417,6 +1417,67 @@ void RISCVDAGToDAGISel::Select(SDNode *Node) {
case ISD::LOAD: {
if (tryIndexedLoad(Node))
return;
+
+ if (Subtarget->hasVendorXCVmem()) {
+ // We hase to match post-incrementing load here
+ LoadSDNode *Load = cast<LoadSDNode>(Node);
+ if (Load->getAddressingMode() != ISD::POST_INC)
+ break;
+
+ SDValue Chain = Node->getOperand(0);
+ SDValue Base = Node->getOperand(1);
+ SDValue Offset = Node->getOperand(2);
+
+ bool simm12 = false;
+ bool signExtend = Load->getExtensionType() == ISD::SEXTLOAD;
+
+ if (auto ConstantOffset = dyn_cast<ConstantSDNode>(Offset)) {
+ int ConstantVal = ConstantOffset->getSExtValue();
+ simm12 = isInt<12>(ConstantVal);
+ if (simm12)
+ Offset = CurDAG->getTargetConstant(ConstantVal, SDLoc(Offset),
+ Offset.getValueType());
+ }
+
+ unsigned Opcode = 0;
+ switch (Load->getMemoryVT().getSimpleVT().SimpleTy) {
+ case MVT::i8:
+ if (simm12 && signExtend)
+ Opcode = RISCV::CV_LB_ri_inc;
+ else if (simm12 && !signExtend)
+ Opcode = RISCV::CV_LBU_ri_inc;
+ else if (!simm12 && signExtend)
+ Opcode = RISCV::CV_LB_rr_inc;
+ else
+ Opcode = RISCV::CV_LBU_rr_inc;
+ break;
+ case MVT::i16:
+ if (simm12 && signExtend)
+ Opcode = RISCV::CV_LH_ri_inc;
+ else if (simm12 && !signExtend)
+ Opcode = RISCV::CV_LHU_ri_inc;
+ else if (!simm12 && signExtend)
+ Opcode = RISCV::CV_LH_rr_inc;
+ else
+ Opcode = RISCV::CV_LHU_rr_inc;
+ break;
+ case MVT::i32:
+ if (simm12)
+ Opcode = RISCV::CV_LW_ri_inc;
+ else
+ Opcode = RISCV::CV_LW_rr_inc;
+ break;
+ default:
+ break;
+ }
+ if (!Opcode)
+ break;
+
+ ReplaceNode(Node, CurDAG->getMachineNode(Opcode, DL, XLenVT, XLenVT,
+ Chain.getSimpleValueType(), Base,
+ Offset, Chain));
+ return;
+ }
break;
}
case ISD::INTRINSIC_WO_CHAIN: {
@@ -2544,6 +2605,19 @@ bool RISCVDAGToDAGISel::SelectAddrRegImmLsb00000(SDValue Addr, SDValue &Base,
return true;
}
+bool RISCVDAGToDAGISel::SelectAddrRegReg(SDValue Addr, SDValue &Base,
+ SDValue &Offset) {
+ if (Addr.getOpcode() != ISD::ADD)
+ return false;
+
+ if (isa<ConstantSDNode>(Addr.getOperand(1)))
+ return false;
+
+ Base = Addr.getOperand(1);
+ Offset = Addr.getOperand(0);
+ return true;
+}
+
bool RISCVDAGToDAGISel::selectShiftMask(SDValue N, unsigned ShiftWidth,
SDValue &ShAmt) {
ShAmt = N;
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h
index 77e174135a599f..c813acf7325fec 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h
@@ -80,6 +80,8 @@ class RISCVDAGToDAGISel : public SelectionDAGISel {
return false;
}
+ bool SelectAddrRegReg(SDValue Addr, SDValue &Base, SDValue &Offset);
+
bool tryShrinkShlLogicImm(SDNode *Node);
bool trySignedBitfieldExtract(SDNode *Node);
bool tryIndexedLoad(SDNode *Node);
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index c8a94adcd91c6a..bdee4725243045 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1366,6 +1366,16 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
}
}
+ if (Subtarget.hasVendorXCVmem()) {
+ setIndexedLoadAction(ISD::POST_INC, MVT::i8, Legal);
+ setIndexedLoadAction(ISD::POST_INC, MVT::i16, Legal);
+ setIndexedLoadAction(ISD::POST_INC, MVT::i32, Legal);
+
+ setIndexedStoreAction(ISD::POST_INC, MVT::i8, Legal);
+ setIndexedStoreAction(ISD::POST_INC, MVT::i16, Legal);
+ setIndexedStoreAction(ISD::POST_INC, MVT::i32, Legal);
+ }
+
// Function alignments.
const Align FunctionAlignment(Subtarget.hasStdExtCOrZca() ? 2 : 4);
setMinFunctionAlignment(FunctionAlignment);
@@ -19324,6 +19334,26 @@ bool RISCVTargetLowering::getPostIndexedAddressParts(SDNode *N, SDNode *Op,
SDValue &Offset,
ISD::MemIndexedMode &AM,
SelectionDAG &DAG) const {
+ if (Subtarget.hasVendorXCVmem()) {
+ if (Op->getOpcode() != ISD::ADD)
+ return false;
+
+ if (LSBaseSDNode *LS = dyn_cast<LSBaseSDNode>(N)) {
+ Base = LS->getBasePtr();
+ } else {
+ return false;
+ }
+
+ if (Base == Op->getOperand(0)) {
+ Offset = Op->getOperand(1);
+ } else if (Base == Op->getOperand(1)) {
+ Offset = Op->getOperand(0);
+ } else {
+ return false;
+ }
+ AM = ISD::POST_INC;
+ return true;
+ }
EVT VT;
SDValue Ptr;
if (LoadSDNode *LD = dyn_cast<LoadSDNode>(N)) {
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoXCV.td b/llvm/lib/Target/RISCV/RISCVInstrInfoXCV.td
index 924e91e15c348f..228efe6e28567a 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoXCV.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoXCV.td
@@ -512,11 +512,13 @@ def CVrrAsmOperand : AsmOperandClass {
let DiagnosticType = "InvalidRegReg";
}
-def CVrr : Operand<OtherVT> {
+def CVrr : Operand<i32>,
+ ComplexPattern<i32, 2, "SelectAddrRegReg",[]> {
let ParserMatchClass = CVrrAsmOperand;
let EncoderMethod = "getRegReg";
let DecoderMethod = "decodeRegReg";
let PrintMethod = "printRegReg";
+ let MIOperandInfo = (ops GPR:$base, GPR:$offset);
}
class CVLoad_ri_inc<bits<3> funct3, string opcodestr>
@@ -659,6 +661,47 @@ let Predicates = [HasVendorXCVelw, IsRV32], hasSideEffects = 0,
def CV_ELW : CVLoad_ri<0b011, "cv.elw">;
}
+//===----------------------------------------------------------------------===//
+// Patterns for load & store operations
+//===----------------------------------------------------------------------===//
+class CVLdrrPat<PatFrag LoadOp, RVInst Inst>
+ : Pat<(XLenVT (LoadOp CVrr:$regreg)),
+ (Inst CVrr:$regreg)>;
+
+class CVStriPat<PatFrag StoreOp, RVInst Inst>
+ : Pat<(StoreOp (XLenVT GPR:$rs2), GPR:$rs1, simm12:$imm12),
+ (Inst GPR:$rs2, GPR:$rs1, simm12:$imm12)>;
+
+class CVStrriPat<PatFrag StoreOp, RVInst Inst>
+ : Pat<(StoreOp (XLenVT GPR:$rs2), GPR:$rs1, GPR:$rs3),
+ (Inst GPR:$rs2, GPR:$rs1, GPR:$rs3)>;
+
+class CVStrrPat<PatFrag StoreOp, RVInst Inst>
+ : Pat<(StoreOp (XLenVT GPR:$rs2), CVrr:$regreg),
+ (Inst GPR:$rs2, CVrr:$regreg)>;
+
+let Predicates = [HasVendorXCVmem, IsRV32], AddedComplexity = 1 in {
+ def : CVLdrrPat<sextloadi8, CV_LB_rr>;
+ def : CVLdrrPat<zextloadi8, CV_LBU_rr>;
+ def : CVLdrrPat<extloadi8, CV_LBU_rr>;
+ def : CVLdrrPat<sextloadi16, CV_LH_rr>;
+ def : CVLdrrPat<zextloadi16, CV_LHU_rr>;
+ def : CVLdrrPat<extloadi16, CV_LHU_rr>;
+ def : CVLdrrPat<load, CV_LW_rr>;
+
+ def : CVStriPat<post_truncsti8, CV_SB_ri_inc>;
+ def : CVStriPat<post_truncsti16, CV_SH_ri_inc>;
+ def : CVStriPat<post_store, CV_SW_ri_inc>;
+
+ def : CVStrriPat<post_truncsti8, CV_SB_rr_inc>;
+ def : CVStrriPat<post_truncsti16, CV_SH_ri_inc>;
+ def : CVStrriPat<post_store, CV_SW_rr_inc>;
+
+ def : CVStrrPat<truncstorei8, CV_SB_rr>;
+ def : CVStrrPat<truncstorei16, CV_SH_rr>;
+ def : CVStrrPat<store, CV_SW_rr>;
+}
+
def cv_tuimm2 : TImmLeaf<XLenVT, [{return isUInt<2>(Imm);}]>;
def cv_tuimm5 : TImmLeaf<XLenVT, [{return isUInt<5>(Imm);}]>;
def cv_uimm10 : ImmLeaf<XLenVT, [{return isUInt<10>(Imm);}]>;
diff --git a/llvm/test/CodeGen/RISCV/xcvmem.ll b/llvm/test/CodeGen/RISCV/xcvmem.ll
new file mode 100644
index 00000000000000..037e49b9b0df7d
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/xcvmem.ll
@@ -0,0 +1,295 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -O3 -mtriple=riscv32 -mattr=+xcvmem -verify-machineinstrs < %s \
+; RUN: | FileCheck %s --check-prefixes=CHECK
+
+define <2 x i32> @lb_ri_inc(i8* %a) {
+; CHECK-LABEL: lb_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lb a1, (a0), 42
+; CHECK-NEXT: ret
+ %1 = load i8, i8* %a
+ %2 = sext i8 %1 to i32
+ %3 = getelementptr i8, i8* %a, i32 42
+ %4 = ptrtoint i8* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define <2 x i32> @lb_rr_inc(i8* %a, i32 %b) {
+; CHECK-LABEL: lb_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lb a1, (a0), a1
+; CHECK-NEXT: ret
+ %1 = load i8, i8* %a
+ %2 = sext i8 %1 to i32
+ %3 = getelementptr i8, i8* %a, i32 %b
+ %4 = ptrtoint i8* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define i32 @lb_rr(i8* %a, i32 %b) {
+; CHECK-LABEL: lb_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lb a0, a1(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i8, i8* %a, i32 %b
+ %2 = load i8, i8* %1
+ %3 = sext i8 %2 to i32
+ ret i32 %3
+}
+
+define <2 x i32> @lbu_ri_inc(i8* %a) {
+; CHECK-LABEL: lbu_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lbu a1, (a0), 42
+; CHECK-NEXT: ret
+ %1 = load i8, i8* %a
+ %2 = zext i8 %1 to i32
+ %3 = getelementptr i8, i8* %a, i32 42
+ %4 = ptrtoint i8* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define <2 x i32> @lbu_rr_inc(i8* %a, i32 %b) {
+; CHECK-LABEL: lbu_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lbu a1, (a0), a1
+; CHECK-NEXT: ret
+ %1 = load i8, i8* %a
+ %2 = zext i8 %1 to i32
+ %3 = getelementptr i8, i8* %a, i32 %b
+ %4 = ptrtoint i8* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define i32 @lbu_rr(i8* %a, i32 %b) {
+; CHECK-LABEL: lbu_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lbu a0, a1(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i8, i8* %a, i32 %b
+ %2 = load i8, i8* %1
+ %3 = zext i8 %2 to i32
+ ret i32 %3
+}
+
+define <2 x i32> @lh_ri_inc(i16* %a) {
+; CHECK-LABEL: lh_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lh a1, (a0), 84
+; CHECK-NEXT: ret
+ %1 = load i16, i16* %a
+ %2 = sext i16 %1 to i32
+ %3 = getelementptr i16, i16* %a, i32 42
+ %4 = ptrtoint i16* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define <2 x i32> @lh_rr_inc(i16* %a, i32 %b) {
+; CHECK-LABEL: lh_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a1, a1, 1
+; CHECK-NEXT: cv.lh a1, (a0), a1
+; CHECK-NEXT: ret
+ %1 = load i16, i16* %a
+ %2 = sext i16 %1 to i32
+ %3 = getelementptr i16, i16* %a, i32 %b
+ %4 = ptrtoint i16* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define i32 @lh_rr(i16* %a, i32 %b) {
+; CHECK-LABEL: lh_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a1, a1, 1
+; CHECK-NEXT: cv.lh a0, a1(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i16, i16* %a, i32 %b
+ %2 = load i16, i16* %1
+ %3 = sext i16 %2 to i32
+ ret i32 %3
+}
+
+define <2 x i32> @lhu_ri_inc(i16* %a) {
+; CHECK-LABEL: lhu_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lhu a1, (a0), 84
+; CHECK-NEXT: ret
+ %1 = load i16, i16* %a
+ %2 = zext i16 %1 to i32
+ %3 = getelementptr i16, i16* %a, i32 42
+ %4 = ptrtoint i16* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define <2 x i32> @lhu_rr_inc(i16* %a, i32 %b) {
+; CHECK-LABEL: lhu_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a1, a1, 1
+; CHECK-NEXT: cv.lhu a1, (a0), a1
+; CHECK-NEXT: ret
+ %1 = load i16, i16* %a
+ %2 = zext i16 %1 to i32
+ %3 = getelementptr i16, i16* %a, i32 %b
+ %4 = ptrtoint i16* %3 to i32
+ %5 = insertelement <2 x i32> undef, i32 %4, i32 0
+ %6 = insertelement <2 x i32> %5, i32 %2, i32 1
+ ret <2 x i32> %6
+}
+
+define i32 @lhu_rr(i16* %a, i32 %b) {
+; CHECK-LABEL: lhu_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a1, a1, 1
+; CHECK-NEXT: cv.lhu a0, a1(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i16, i16* %a, i32 %b
+ %2 = load i16, i16* %1
+ %3 = zext i16 %2 to i32
+ ret i32 %3
+}
+
+define <2 x i32> @lw_ri_inc(i32* %a) {
+; CHECK-LABEL: lw_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.lw a1, (a0), 168
+; CHECK-NEXT: ret
+ %1 = load i32, i32* %a
+ %2 = getelementptr i32, i32* %a, i32 42
+ %3 = ptrtoint i32* %2 to i32
+ %4 = insertelement <2 x i32> undef, i32 %3, i32 0
+ %5 = insertelement <2 x i32> %4, i32 %1, i32 1
+ ret <2 x i32> %5
+}
+
+define <2 x i32> @lw_rr_inc(i32* %a, i32 %b) {
+; CHECK-LABEL: lw_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a1, a1, 2
+; CHECK-NEXT: cv.lw a1, (a0), a1
+; CHECK-NEXT: ret
+ %1 = load i32, i32* %a
+ %2 = getelementptr i32, i32* %a, i32 %b
+ %3 = ptrtoint i32* %2 to i32
+ %4 = insertelement <2 x i32> undef, i32 %3, i32 0
+ %5 = insertelement <2 x i32> %4, i32 %1, i32 1
+ ret <2 x i32> %5
+}
+
+define i32 @lw_rr(i32* %a, i32 %b) {
+; CHECK-LABEL: lw_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a1, a1, 2
+; CHECK-NEXT: cv.lw a0, a1(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i32, i32* %a, i32 %b
+ %2 = load i32, i32* %1
+ ret i32 %2
+}
+
+define i8* @sb_ri_inc(i8* %a, i8 %b) {
+; CHECK-LABEL: sb_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.sb a1, (a0), 42
+; CHECK-NEXT: ret
+ store i8 %b, i8* %a
+ %1 = getelementptr i8, i8* %a, i32 42
+ ret i8* %1
+}
+
+define i8* @sb_rr_inc(i8* %a, i8 %b, i32 %c) {
+; CHECK-LABEL: sb_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.sb a1, (a0), a2
+; CHECK-NEXT: ret
+ store i8 %b, i8* %a
+ %1 = getelementptr i8, i8* %a, i32 %c
+ ret i8* %1
+}
+
+define void @sb_rr(i8* %a, i8 %b, i32 %c) {
+; CHECK-LABEL: sb_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.sb a1, a2(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i8, i8* %a, i32 %c
+ store i8 %b, i8* %1
+ ret void
+}
+
+define i16* @sh_ri_inc(i16* %a, i16 %b) {
+; CHECK-LABEL: sh_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.sh a1, (a0), 84
+; CHECK-NEXT: ret
+ store i16 %b, i16* %a
+ %1 = getelementptr i16, i16* %a, i32 42
+ ret i16* %1
+}
+
+define i16* @sh_rr_inc(i16* %a, i16 %b, i32 %c) {
+; CHECK-LABEL: sh_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a2, a2, 1
+; CHECK-NEXT: cv.sh a1, (a0), a2
+; CHECK-NEXT: ret
+ store i16 %b, i16* %a
+ %1 = getelementptr i16, i16* %a, i32 %c
+ ret i16* %1
+}
+
+define void @sh_rr(i16* %a, i16 %b, i32 %c) {
+; CHECK-LABEL: sh_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a2, a2, 1
+; CHECK-NEXT: cv.sh a1, a2(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i16, i16* %a, i32 %c
+ store i16 %b, i16* %1
+ ret void
+}
+
+define i32* @sw_ri_inc(i32* %a, i32 %b) {
+; CHECK-LABEL: sw_ri_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: cv.sw a1, (a0), 168
+; CHECK-NEXT: ret
+ store i32 %b, i32* %a
+ %1 = getelementptr i32, i32* %a, i32 42
+ ret i32* %1
+}
+
+define i32* @sw_rr_inc(i32* %a, i32 %b, i32 %c) {
+; CHECK-LABEL: sw_rr_inc:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a2, a2, 2
+; CHECK-NEXT: cv.sw a1, (a0), a2
+; CHECK-NEXT: ret
+ store i32 %b, i32* %a
+ %1 = getelementptr i32, i32* %a, i32 %c
+ ret i32* %1
+}
+
+define void @sw_rr(i32* %a, i32 %b, i32 %c) {
+; CHECK-LABEL: sw_rr:
+; CHECK: # %bb.0:
+; CHECK-NEXT: slli a2, a2, 2
+; CHECK-NEXT: cv.sw a1, a2(a0)
+; CHECK-NEXT: ret
+ %1 = getelementptr i32, i32* %a, i32 %c
+ store i32 %b, i32* %1
+ ret void
+}
|
topperc
reviewed
Jan 4, 2024
| @@ -1210,7 +1210,7 @@ struct RISCVOperand final : public MCParsedAsmOperand { | |||
| } | |||
|
|
|||
| void addRegRegOperands(MCInst &Inst, unsigned N) const { | |||
| assert(N == 1 && "Invalid number of operands!"); | |||
| assert(N == 2 && "Invalid number of operands!"); | |||
There was a problem hiding this comment.
How does this relate to the rest of the patch? A .ll test shouldn't use the AsmParser.
There was a problem hiding this comment.
I think it's because I'm using let MIOperandInfo = (ops GPR:$base, GPR:$offset) for CVrr operand, so it triggers the assert.
topperc
reviewed
Jan 4, 2024
| @@ -1417,6 +1417,67 @@ void RISCVDAGToDAGISel::Select(SDNode *Node) { | |||
| case ISD::LOAD: { | |||
| if (tryIndexedLoad(Node)) | |||
| return; | |||
|
|
|||
| if (Subtarget->hasVendorXCVmem()) { | |||
| // We hase to match post-incrementing load here | |||
| SDValue Base = Node->getOperand(1); | ||
| SDValue Offset = Node->getOperand(2); | ||
|
|
||
| bool simm12 = false; |
11 tasks
topperc
reviewed
Jan 4, 2024
| if (Op->getOpcode() != ISD::ADD) | ||
| return false; | ||
|
|
||
| if (LSBaseSDNode *LS = dyn_cast<LSBaseSDNode>(N)) { |
| return false; | ||
| } | ||
|
|
||
| if (Base == Op->getOperand(0)) { |
| } | ||
| AM = ISD::POST_INC; | ||
| return true; | ||
| } | ||
| EVT VT; |
All post-Increment load/store, register-register load/store spec: https://github.com/openhwgroup/cv32e40p/blob/master/docs/source/instruction_set_extensions.rst
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
All post-Increment load/store, register-register load/store
spec:
https://github.com/openhwgroup/cv32e40p/blob/master/docs/source/instruction_set_extensions.rst
Contributors: @CharKeaney, @jeremybennett, @lewis-revill,
@NandniJamnadas, @PaoloS02, @serkm, @simonpcook, @xingmingjie, @realqhc