Skip to content

[mlir] [amdgpu] Remove s_wait_loadcnt from amdgpu.lds_barrier on gfx12 #152778

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

[mlir] [amdgpu] Remove s_wait_loadcnt from amdgpu.lds_barrier on gfx12 #152778

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 11 additions & 9 deletions mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -535,18 +535,22 @@ struct LDSBarrierOpLowering : public ConvertOpToLLVMPattern<LDSBarrierOp> {
LogicalResult
matchAndRewrite(LDSBarrierOp op, LDSBarrierOp::Adaptor adaptor,
ConversionPatternRewriter &rewriter) const override {
bool requiresInlineAsm = chipset < kGfx90a || chipset.majorVersion == 11;
bool requiresInlineAsm = chipset < kGfx90a || chipset.majorVersion >= 11;

if (requiresInlineAsm) {
auto asmDialectAttr = LLVM::AsmDialectAttr::get(rewriter.getContext(),
LLVM::AsmDialect::AD_ATT);
const char *asmStr =
const char *asmStrPreGfx12 =
";;;WARNING: BREAKS DEBUG WATCHES\ns_waitcnt lgkmcnt(0)\ns_barrier";
const char *asmStr =
";;;WARNING: BREAKS DEBUG WATCHES\n"
"s_wait_dscnt 0x0\ns_barrier_signal -1\ns_barrier_wait -1";
const char *constraints = "";
rewriter.replaceOpWithNewOp<LLVM::InlineAsmOp>(
op,
/*resultTypes=*/TypeRange(), /*operands=*/ValueRange(),
/*asm_string=*/asmStr, constraints, /*has_side_effects=*/true,
/*asm_string=*/chipset.majorVersion >= 12 ? asmStr : asmStrPreGfx12,
constraints, /*has_side_effects=*/true,
/*is_align_stack=*/false, LLVM::TailCallKind::None,
/*asm_dialect=*/asmDialectAttr,
/*operand_attrs=*/ArrayAttr());
Expand Down Expand Up @@ -574,14 +578,12 @@ struct LDSBarrierOpLowering : public ConvertOpToLLVMPattern<LDSBarrierOp> {
Location loc = op->getLoc();
ROCDL::SWaitcntOp::create(rewriter, loc, ldsOnlyBits);
rewriter.replaceOpWithNewOp<ROCDL::SBarrierOp>(op);
return success();
} else {
Location loc = op->getLoc();
ROCDL::WaitDscntOp::create(rewriter, loc, 0);
ROCDL::BarrierSignalOp::create(rewriter, loc, -1);
rewriter.replaceOpWithNewOp<ROCDL::BarrierWaitOp>(op, -1);
return op.emitOpError(
"don't know how to lower this for chipset major version")
<< chipset.majorVersion;
}

return success();
}
};

Expand Down
5 changes: 2 additions & 3 deletions mlir/test/Conversion/AMDGPUToROCDL/amdgpu-to-rocdl.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -424,9 +424,8 @@ func.func @lds_barrier() {
// GFX10-NEXT: rocdl.s.barrier
// GFX11: llvm.inline_asm has_side_effects asm_dialect = att
// GFX11-SAME: ";;;WARNING: BREAKS DEBUG WATCHES\0As_waitcnt lgkmcnt(0)\0As_barrier"
// GFX12: rocdl.s.wait.dscnt 0
// GFX12-NEXT: rocdl.s.barrier.signal -1
// GFX12-NEXT: rocdl.s.barrier.wait -1
// GFX12: llvm.inline_asm has_side_effects asm_dialect = att
// GFX12-SAME: ";;;WARNING: BREAKS DEBUG WATCHES\0As_wait_dscnt 0x0\0As_barrier_signal -1\0As_barrier_wait -1"
amdgpu.lds_barrier
func.return
}
Expand Down