The Rust Programming Language Blog: Raising the baseline for the `nvptx64-nvidia-cuda` target

The nvptx64-nvidia-cuda target is a compilation target for NVIDIA GPUs. When using this target, the final output is . These changes affect both the Rust compiler (rustc) and related host tooling, and they make it impossible to generate PTX artifacts compatible with older GPUs and older CUDA drivers.

The new minimum supported versions will be:

• PTX ISA 7.0 (requires a CUDA 11 driver or newer)


• SM 7.0 (GPUs with compute capability below 7.0 are no longer supported)

What happens when I update to Rust 1.97?

If you need to target a CUDA driver that does not support PTX ISA 7.0 (CUDA 10-era drivers and older), Rust 1.97 will no longer be able to generate PTX compatible with that environment. Similarly, if you need to run on GPUs with compute capability below 7.0 (for example, Maxwell or Pascal), Rust 1.97 will no longer be able to generate compatible PTX for those GPUs.

Assuming you are targeting a CUDA driver compatible with CUDA 11 or newer and using GPUs with compute capability 7.0 or newer:

• If you do not specify -C target-cpu, the new default will be sm_70, and your build should continue to work (but will no longer be compatible with pre-Volta GPUs).


• If you currently specify an older -C target-cpu (for example, sm_60), you will need to either:
  
  
  
  • remove that flag and let it default to sm_70, or
  
  
  • update it to sm_70 or a newer architecture.
  
  
  
  


• If you already specify -C target-cpu=sm_70 (or newer), there should be no behavioral changes from this update.

For more details on building and configuring nvptx64-nvidia-cuda, see the platform support documentation.