feat: support aarch64 and modern NVIDIA Blackwell (sm_100) architectures#2
Open
mikee-gwu wants to merge 1 commit intoscrya-com:mainfrom
Open
feat: support aarch64 and modern NVIDIA Blackwell (sm_100) architectures#2mikee-gwu wants to merge 1 commit intoscrya-com:mainfrom
mikee-gwu wants to merge 1 commit intoscrya-com:mainfrom
Conversation
|
I offer myself for tests on Blackwell SM120 (I'm running a RTX5080) |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Title: Support non-x86 architectures (aarch64) and modern NVIDIA GPUs (Blackwell/sm_100)
Description: This PR enables rotorquant to be built and run on NVIDIA Grace Blackwell (GB10) systems and other ARM64 (aarch64) environments.
Key Improvements:
Robust Build System: Updated setup.py to auto-detect the CUDA environment, properly set architecture lists (dynamically adding +PTX for compatibility).
Cross-Platform Compatibility: Replaced hardcoded .so extension references in cuda_backend.py with dynamic glob matching, allowing the library to load compiled extensions on any platform (e.g., aarch64).
Future-Proofing: Added support for NVIDIA Blackwell (sm_100) via PTX fallback, ensuring the kernels are JIT-compatible with modern NVIDIA driver architectures.
Testing:
Verified on NVIDIA DGX Spark (Grace Blackwell GB10, Ubuntu aarch64). The build process is now self-contained, and the end-to-end benchmark suite passes successfully.