Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Disabling on-disk cache #4549

Open
sklam opened this issue Sep 10, 2019 · 4 comments
Open

Disabling on-disk cache #4549

sklam opened this issue Sep 10, 2019 · 4 comments

Comments

@sklam
Copy link
Member

sklam commented Sep 10, 2019

Although NUMBA_CACHE_DIR allows us to move the cache directory to another location, I think that for HPC systems we actually want to have a mechanism for disabling caching completely. We sometimes run mpi4py applications with >100K processes, and it is much better to have all those processes independently spend a few seconds compiling things rather than trying to read them from disk.

Is there any hope of a "NUMBA_CACHE_DISABLE" switch to forcibly override any cache=True options globally?

Originally posted by @tskisner in #3470 (comment)

@ickc
Copy link
Contributor

ickc commented Sep 10, 2019

To reiterate, I think I read somewhere that when Numba reaches v1.0, Numba will default to cache if the decorator doesn't specifically disabled it. So it would be important to have this env. var. option (or any sort of "global kill switch") to override this behavior before Numba reaching v1.0. Could we add this issue as part of the v1.0 milestone?

@seibert seibert added this to the Numba 1.0 milestone Sep 10, 2019
@seibert
Copy link
Contributor

seibert commented Sep 10, 2019

Likely we will do this sooner, but good to note the need before caching is enabled by default.

@leofang
Copy link
Contributor

leofang commented Sep 13, 2019

On the GPU side, I would view this feature request as more or less the opposite of that in #4389, in which we don't want any compiling to happen. It'd be good to have both features in 1.0 and let users decide the deployment strategy in a HPC system.

@seibert
Copy link
Contributor

seibert commented Sep 16, 2019

I think we'll want to do this as NUMBA_CACHE_OVERRIDE=always|never

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants