Skip to content

[MLIR][NVVM][Refactor] Refactor intrinsic lowering for NVVM Ops #157079

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 11 commits into
base: main
Choose a base branch
from
Open

[MLIR][NVVM][Refactor] Refactor intrinsic lowering for NVVM Ops #157079

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
7c7d739
[MLIR][NVVM][Refactor] Refactor intrinsic lowering for NVVM Ops
Wolfram70 Sep 4, 2025
81d3f9a
fix typo
Wolfram70 Sep 4, 2025
3fe6daa
move functions in NVVMToLLVMIRTranslation.cpp to NVVMDialect.cpp
Wolfram70 Sep 5, 2025
8e6b622
simplify names
Wolfram70 Sep 5, 2025
04310fc
fix format
Wolfram70 Sep 5, 2025
a4d09da
fix warnings
Wolfram70 Sep 5, 2025
f6a90df
use struct instead of tuple
Wolfram70 Sep 5, 2025
0ae224f
replace unused variables with _
Wolfram70 Sep 5, 2025
84b8e76
replace undef with poison
Wolfram70 Sep 5, 2025
78c6e07
fix warnings
Wolfram70 Sep 8, 2025
5697c92
fix tma_store_reduce test
Wolfram70 Sep 8, 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
19 changes: 15 additions & 4 deletions mlir/include/mlir/Dialect/LLVMIR/NVVMDialect.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,10 +55,21 @@ enum NVVMMemorySpace {
kSharedClusterMemorySpace = 7,
};

/// A pair type of LLVM's Intrinsic ID and args (which are llvm values).
/// This type is returned by the getIntrinsicIDAndArgs() methods.
using IDArgPair =
std::pair<llvm::Intrinsic::ID, llvm::SmallVector<llvm::Value *>>;
/// A struct of LLVM's Intrinsic ID, args (which are llvm values),
/// and args types (which are llvm types).
/// Args types are only required for overloaded intrinsics to provide the
/// correct argument types to the createIntrinsicCall() method.
/// This type is returned by the getIIDAndArgsWithTypes() methods.
struct IIDArgsWithTypes {
IIDArgsWithTypes(llvm::Intrinsic::ID id,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why we need a constructor?

Copy link
Contributor Author

@Wolfram70 Wolfram70 Sep 16, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is because when returning this struct with the ternary ? operator, aggregate initialization doesn't work and gives an error, so I added this constructor to explicitly construct the object in those cases instead.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see, can't we use below syntax?

return cond == some_cond ? S1{a1, b1} : S1{a2, b2};

llvm::SmallVector<llvm::Value *> args,
llvm::SmallVector<llvm::Type *> types)
: id(id), args(args), types(types) {}

llvm::Intrinsic::ID id;
llvm::SmallVector<llvm::Value *> args;
llvm::SmallVector<llvm::Type *> types;
};

/// Return the element type and number of elements associated with a wmma matrix
/// of given chracteristics. This matches the logic in IntrinsicsNVVM.td
Expand Down
Loading