Skip to content

Use CUBIN generation instead of PTX for newer CUDA

Sebastian Ohlmann requested to merge cuda_cubin_generation into main

Description

This is supported from CUDA 11.1 on, see also https://docs.nvidia.com/cuda/cuda-c-best-practices-guide/index.html#dynamic-code-generation

It avoids incompatiblities between the PTX versions of the runtime library that we link against and the driver. This can happenif the runtime library is newer than the driver on the system.

News snippet

Use CUBIN generation instead of PTX for newer CUDA versions

Checklist

  • I have checked that my code follows the Octopus coding standards
  • I have added tests for all the new features added in this request.

Merge request reports