[RISCV] Add SMLoc to expanded vector pseudoinstructions in AsmParser. #84875
+97
−18
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
This is needed for llvm-mca to correctly apply vsetvli instruments to these instructions. Fixes llvm#84799.
|
@llvm/pr-subscribers-backend-risc-v Author: Craig Topper (topperc) ChangesThis is needed for llvm-mca to correctly apply vsetvli instruments to these instructions. Fixes #84799. Full diff: https://github.com/llvm/llvm-project/pull/84875.diff 2 Files Affected:
diff --git a/llvm/include/llvm/MC/MCInstBuilder.h b/llvm/include/llvm/MC/MCInstBuilder.h
index 6e5e9dd69018f4..d06ed4c6c840a9 100644
--- a/llvm/include/llvm/MC/MCInstBuilder.h
+++ b/llvm/include/llvm/MC/MCInstBuilder.h
@@ -27,6 +27,12 @@ class MCInstBuilder {
Inst.setOpcode(Opcode);
}
+ /// Set the location.
+ MCInstBuilder &setLoc(SMLoc SM) {
+ Inst.setLoc(SM);
+ return *this;
+ }
+
/// Add a new register operand.
MCInstBuilder &addReg(unsigned Reg) {
Inst.addOperand(MCOperand::createReg(Reg));
diff --git a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
index d83979a873f2a3..caff0e8fcefe23 100644
--- a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
+++ b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
@@ -3271,11 +3271,13 @@ void RISCVAsmParser::emitVMSGE(MCInst &Inst, unsigned Opcode, SMLoc IDLoc,
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(1))
.addOperand(Inst.getOperand(2))
- .addReg(RISCV::NoRegister));
+ .addReg(RISCV::NoRegister)
+ .setLoc(IDLoc));
emitToStreamer(Out, MCInstBuilder(RISCV::VMNAND_MM)
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(0))
- .addOperand(Inst.getOperand(0)));
+ .addOperand(Inst.getOperand(0))
+ .setLoc(IDLoc));
} else if (Inst.getNumOperands() == 4) {
// masked va >= x, vd != v0
//
@@ -3287,11 +3289,13 @@ void RISCVAsmParser::emitVMSGE(MCInst &Inst, unsigned Opcode, SMLoc IDLoc,
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(1))
.addOperand(Inst.getOperand(2))
- .addOperand(Inst.getOperand(3)));
+ .addOperand(Inst.getOperand(3))
+ .setLoc(IDLoc));
emitToStreamer(Out, MCInstBuilder(RISCV::VMXOR_MM)
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(0))
- .addReg(RISCV::V0));
+ .addReg(RISCV::V0)
+ .setLoc(IDLoc));
} else if (Inst.getNumOperands() == 5 &&
Inst.getOperand(0).getReg() == RISCV::V0) {
// masked va >= x, vd == v0
@@ -3306,11 +3310,13 @@ void RISCVAsmParser::emitVMSGE(MCInst &Inst, unsigned Opcode, SMLoc IDLoc,
.addOperand(Inst.getOperand(1))
.addOperand(Inst.getOperand(2))
.addOperand(Inst.getOperand(3))
- .addReg(RISCV::NoRegister));
+ .addReg(RISCV::NoRegister)
+ .setLoc(IDLoc));
emitToStreamer(Out, MCInstBuilder(RISCV::VMANDN_MM)
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(0))
- .addOperand(Inst.getOperand(1)));
+ .addOperand(Inst.getOperand(1))
+ .setLoc(IDLoc));
} else if (Inst.getNumOperands() == 5) {
// masked va >= x, any vd
//
@@ -3323,19 +3329,23 @@ void RISCVAsmParser::emitVMSGE(MCInst &Inst, unsigned Opcode, SMLoc IDLoc,
.addOperand(Inst.getOperand(1))
.addOperand(Inst.getOperand(2))
.addOperand(Inst.getOperand(3))
- .addReg(RISCV::NoRegister));
+ .addReg(RISCV::NoRegister)
+ .setLoc(IDLoc));
emitToStreamer(Out, MCInstBuilder(RISCV::VMANDN_MM)
.addOperand(Inst.getOperand(1))
.addReg(RISCV::V0)
- .addOperand(Inst.getOperand(1)));
+ .addOperand(Inst.getOperand(1))
+ .setLoc(IDLoc));
emitToStreamer(Out, MCInstBuilder(RISCV::VMANDN_MM)
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(0))
- .addReg(RISCV::V0));
+ .addReg(RISCV::V0)
+ .setLoc(IDLoc));
emitToStreamer(Out, MCInstBuilder(RISCV::VMOR_MM)
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(1))
- .addOperand(Inst.getOperand(0)));
+ .addOperand(Inst.getOperand(0))
+ .setLoc(IDLoc));
}
}
@@ -3637,7 +3647,8 @@ bool RISCVAsmParser::processInstruction(MCInst &Inst, SMLoc IDLoc,
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(1))
.addImm(Imm - 1)
- .addOperand(Inst.getOperand(3)));
+ .addOperand(Inst.getOperand(3))
+ .setLoc(IDLoc));
return false;
}
case RISCV::PseudoVMSGEU_VI:
@@ -3655,7 +3666,8 @@ bool RISCVAsmParser::processInstruction(MCInst &Inst, SMLoc IDLoc,
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(1))
.addOperand(Inst.getOperand(1))
- .addOperand(Inst.getOperand(3)));
+ .addOperand(Inst.getOperand(3))
+ .setLoc(IDLoc));
} else {
// Other immediate values can subtract one like signed.
unsigned Opc = Inst.getOpcode() == RISCV::PseudoVMSGEU_VI
@@ -3665,7 +3677,8 @@ bool RISCVAsmParser::processInstruction(MCInst &Inst, SMLoc IDLoc,
.addOperand(Inst.getOperand(0))
.addOperand(Inst.getOperand(1))
.addImm(Imm - 1)
- .addOperand(Inst.getOperand(3)));
+ .addOperand(Inst.getOperand(3))
+ .setLoc(IDLoc));
}
return false;
|
|
I suspect we have the same problem with any other pseudo which can occur in assembly (as opposed to being compiler generated). One example would be LI. This would be a followup though, not a change to this patch. |
preames
approved these changes
Mar 13, 2024
Possibly. I wasn't sure if it mattered for scalar. Do we have something like the SEW LMUL instrument for scalar |
I don't believe so, but if this also caused the BEGIN/END marker behavior, then the scalar case is still relevant. |
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.
This is needed for llvm-mca to correctly apply vsetvli instruments to these instructions.
Fixes #84799.