Skip to content

Conversation

@Lunderberg
Copy link
Contributor

For older versions of ccache that do not support nvcc, this has no effect. For newer versions of ccache, this enables caching the compiled nvcc outputs. This was tested using ccache versions 3.7.7 (does not cache nvcc output) and 4.8.3 (caches nvcc output).

This is primarily intended to reduce compilation time of cutlass kernels, which can take several minutes after switching branches.

For older versions of `ccache` that do not support `nvcc`, this has no
effect.  For newer versions of `ccache`, this enables caching the
compiled `nvcc` outputs.  This was tested using `ccache` versions
3.7.7 (does not cache `nvcc` output) and 4.8.3 (caches `nvcc` output).

This is primarily intended to reduce compilation time of cutlass
kernels, which can take several minutes after switching branches.
Copy link
Contributor

@csullivan csullivan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Many many thanks! Very happy to see the unnecessary lengthy nvcc spawned ptxas recompiles of cutlass kernels for different archs go away

@Lunderberg Lunderberg merged commit 56bdcee into apache:main Jan 4, 2024
@Lunderberg Lunderberg deleted the cmake_use_ccache_for_nvcc branch January 4, 2024 17:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants