Skip to content

[mlir][AMDGPU] Add a scheduling barrier guard around inlineAsm lds.barrier #109678

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] Add a scheduling barrier guard around inlineAsm lds.barrier #109678

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
12 changes: 10 additions & 2 deletions mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -290,15 +290,23 @@ struct LDSBarrierOpLowering : public ConvertOpToLLVMPattern<LDSBarrierOp> {
if (requiresInlineAsm) {
auto asmDialectAttr = LLVM::AsmDialectAttr::get(rewriter.getContext(),
LLVM::AsmDialect::AD_ATT);
Location loc = op->getLoc();
// Ensure the inlineAsm is guarded with a scheduling region
// So it will not interfere with backend compilation more than
// it needs.
rewriter.create<amdgpu::SchedBarrierOp>(
loc, amdgpu::sched_barrier_opt_enum::none);
const char *asmStr =
";;;WARNING: BREAKS DEBUG WATCHES\ns_waitcnt lgkmcnt(0)\ns_barrier";
const char *constraints = "";
rewriter.replaceOpWithNewOp<LLVM::InlineAsmOp>(
op,
rewriter.create<LLVM::InlineAsmOp>(
loc,
/*resultTypes=*/TypeRange(), /*operands=*/ValueRange(),
/*asm_string=*/asmStr, constraints, /*has_side_effects=*/true,
/*is_align_stack=*/false, /*asm_dialect=*/asmDialectAttr,
/*operand_attrs=*/ArrayAttr());
rewriter.replaceOpWithNewOp<amdgpu::SchedBarrierOp>(
op, amdgpu::sched_barrier_opt_enum::none);
return success();
}
if (chipset.majorVersion < 12) {
Expand Down
4 changes: 4 additions & 0 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 @@ -239,14 +239,18 @@ func.func @amdgpu_raw_buffer_atomic_cmpswap_v2f16(%src : vector<2xf16>, %cmp : v

// CHECK-LABEL: func @lds_barrier
func.func @lds_barrier() {
// GFX908: rocdl.sched.barrier 0
// GFX908: llvm.inline_asm has_side_effects asm_dialect = att
// GFX908-SAME: ";;;WARNING: BREAKS DEBUG WATCHES\0As_waitcnt lgkmcnt(0)\0As_barrier"
// GFX908: rocdl.sched.barrier 0
// GFX90A: rocdl.waitcnt -7937
// GFX90A-NEXT: rocdl.s.barrier
// GFX10: rocdl.waitcnt -16129
// GFX10-NEXT: rocdl.s.barrier
// GFX11: rocdl.sched.barrier 0
// GFX11: llvm.inline_asm has_side_effects asm_dialect = att
// GFX11-SAME: ";;;WARNING: BREAKS DEBUG WATCHES\0As_waitcnt lgkmcnt(0)\0As_barrier"
// GFX11: rocdl.sched.barrier 0
// GFX12: rocdl.s.wait.dscnt 0
// GFX12-NEXT: rocdl.s.barrier.signal -1
// GFX12-NEXT: rocdl.s.barrier.wait -1
Expand Down
Loading