Skip to content

[mlir][GPUToNVVM] enable fallback to generic LLVM lowering for math dialect in convert-gpu-to-nvvm pass #165728

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

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
+22 −4
Open

[mlir][GPUToNVVM] enable fallback to generic LLVM lowering for math dialect in convert-gpu-to-nvvm pass #165728

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
79e28d8
enable fallback to generic LLVM lowering for math dialect in convert-…
Oct 30, 2025
9089dba
remove unrelated test
Oct 30, 2025
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
6 changes: 2 additions & 4 deletions mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -383,16 +383,14 @@ struct LowerGpuOpsToNVVMOpsPass final
LLVMConversionTarget target(getContext());

// Set higher benefit, so patterns will run before generic LLVM lowering.
// Make sure the benefit here is higher than ArithToLLVMDialectInterface and
// MathToLLVMDialectInterface.
populateGpuToNVVMConversionPatterns(converter, llvmPatterns,
/*benefit=*/10);

llvm::SmallDenseSet<StringRef> allowedDialectsSet(allowedDialects.begin(),
allowedDialects.end());
for (Dialect *dialect : getContext().getLoadedDialects()) {
// Skip math patterns as nvvm needs custom math lowering.
if (isa<math::MathDialect>(dialect))
continue;

bool allowed = allowedDialectsSet.contains(dialect->getNamespace());
// Empty `allowedDialectsSet` means all dialects are allowed.
if (!allowedDialectsSet.empty() && !allowed)
Expand Down
20 changes: 20 additions & 0 deletions mlir/test/Conversion/GPUToNVVM/gpu-to-generic-llvm.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
// RUN: mlir-opt %s -convert-gpu-to-nvvm -split-input-file | FileCheck %s

/// Math/arith ops that are not supported by libdevice
/// should be converted by generic LLVM lowering patterns.

gpu.module @generic_llvm_test_module_0 {
// CHECK-LABEL: @math_abs_non_i32
func.func @math_abs_non_i32(%arg_i64: i64, %arg_i16: i16, %arg_i8: i8, %arg_i1: i1)
-> (i64, i16, i8, i1) {
// CHECK: "llvm.intr.abs"{{.*}} : (i64) -> i64
%abs_i64 = math.absi %arg_i64 : i64
// CHECK: "llvm.intr.abs"{{.*}} : (i16) -> i16
%abs_i16 = math.absi %arg_i16 : i16
// CHECK: "llvm.intr.abs"{{.*}} : (i8) -> i8
%abs_i8 = math.absi %arg_i8 : i8
// CHECK: "llvm.intr.abs"{{.*}} : (i1) -> i1
%abs_i1 = math.absi %arg_i1 : i1
return %abs_i64, %abs_i16, %abs_i8, %abs_i1 : i64, i16, i8, i1
}
}