Preview for CUDA on WSL Updated for Performance

Recently, Microsoft announced their public preview program for their Windows Subsystem for Linux (WSL) capability on Microsoft Windows platforms. NVIDIA co-released GPU-accelerated support via CUDA on WSL, where AI frameworks run as Linux executables on Microsoft Windows platforms. 

We are happy to announce the release of the latest version of the preview driver with enhancements that will improve the experience for CUDA WSL users and enable more of the native CUDA features on WSL: 

  • Performance of CUDA on WSL2 was particularly lacking on small workloads without batching, as they incurred higher CPU overhead.  We have improved the performance of such workloads by optimizing the driver for launch and synchronization tasks and plan to deliver additional performance optimizations in upcoming drivers to improve the CUDA experience on WSL. 
  • DirectML API support for DirectX 12 based GPU acceleration 
  • One of the key CUDA features enabled in this driver is the support for PTX JIT. This will enable CUDA developers to run PTX code on WSL2. Note that unlike Native Linux, in WSL we load PTX JIT from the Driver Store directly, so you will not find the library in your lib folders.

Features planned in upcoming releases: 

  • Support for driver libraries such as OptiX 
  • Performance optimizations related to memory access
  • Support for NVML and Nvidia-smi for GPU management
  • Additional improvements for multi-GPU systems

Interested participants can register in both the NVIDIA Developer Program and the Microsoft Windows Insider Program to access driver installers and documentation via the Downloads area on our CUDA on WSL webpage. Developers can use our Forum for sharing their public preview feedback with the WSL community.

Leverage our developer blog, CUDA on Windows Subsystem for Linux, to learn much more about how CUDA works with WSL, including how to get started with running applications, including deep learning containers.

Tags: ,