NVIDIA cuFFT LTO EA preview documentation

Welcome to the cuFFT LTO EA (cuFFT with Link-Time Optimization Early Access) preview.

Here you can find:

  • A Quick start guide with a sample snippet.

  • A How to use cuFFT LTO EA section, with an explanation of how to use this preview version of cuFFT with LTO.

  • An API reference section, with a description of the extension that this cuFFT LTO EA APIs brings to cuFFT.

The cuFFT LTO EA preview can be found in NVIDIA cuFFT LTO EA Preview.

Highlights

  • Extension to the callback API to support LTO callback routines.

  • No offline device-linking required to use callbacks.

  • Adds callback support to the dynamic cuFFT library.

  • Adds callback support to Windows.

  • Compatible with existing callback device code.

  • Increased performance vs. the non-LTO callback routines for many cases.

Note

The cuFFT LTO EA preview, unlike the version of cuFFT shipped in the CUDA Toolkit, is not a full production binary. It is meant as a way for users to test LTO-enabled callback functions on both Linux and Windows, and provide us with feedback so that we can improve the experience before this feature makes into production as part of cuFFT. While we do our best to ship these binaries without issues, occasional bugs might be encountered.

Please direct any feedback you might have to Miguel Ferrer Avila <mferreravila@nvidia.com>, Lukasz Ligowski <lligowski@nvidia.com> or Arthy Sundaram <asundaram@nvidia.com>.