Support Nvidia GPUs in the MLIR backend#17
Merged
guillon merged 4 commits intoxtc-tools:mainfrom Jan 27, 2026
Merged
Conversation
ElectrikSpace
force-pushed
the
dev/snoiry/gpu
branch
from
2a48b57 to
10f3a2b
Compare
guillon
requested changes
Jan 20, 2026
This commit adds a new target to the MLIR backend for NVidia GPUS. It requires a specific build of llvm-project with options: - LLVM_TARGETS_TO_BUILD="host;NVPTX" - MLIR_ENABLE_CUDA_RUNNER=ON - MLIR_NVVM_EMBED_LIBDEVICE=ON The PTX version is currently hardcoded to sm_80 and ptx75+ (Ampere). Tested on Nvida A100, Nvidia L40S, and Nvidia L4.
These are now inferred from nvidia-smi and nvcc. Of course these tools need to be installed, but they are part of the Nvidia toolkit.
The implementation relies on the Nvidia CUpti profiling library. More specifically, it uses the range profiler. It integrates itself in the XTC runtime if preprocessor conditions and hooks.
- Vectorization is not supported for now - GPU kernels are generated only if there is one or multiple parallel axes for now.
ElectrikSpace
force-pushed
the
dev/snoiry/gpu
branch
from
10f3a2b to
4ae8099
Compare
guillon
approved these changes
Jan 27, 2026
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
Target Nvidia GPUs through the MLIR backend
Description
Add basic support for Nvidia GPUs, leveraging the existing support in MLIR. This also adds support for Cupti performance counters in the runtime.
TODO: