replaces ccache with sccache
#4534
Conversation
AyodeAwe
added
improvement
|
rerun tests |
|
rerun tests |
|
|
|
rerun tests |
|
PR is ready for review, but we'll wait to merge until all the other PRs are confirmed working as well. |
|
build after |
|
build where both architectures successfully used cache - https://gpuci.gpuopenanalytics.com/job/rapidsai/job/gpuci/job/cuml/job/prb/job/cuml-prb/10905/ |
BUILD.md
Outdated
| @@ -58,7 +58,7 @@ $ PARALLEL_LEVEL=8 ./build.sh libcuml # build and install libcuml limiting para | |||
| $ ./build.sh libcuml -n # build libcuml but do not install | |||
| $ ./build.sh prims --allgpuarch # build the ML prims tests for all supported GPU architectures | |||
| $ ./build.sh cuml --singlegpu # build the cuML python package without MNMG algorithms | |||
| $ ./build.sh --ccache # use ccache to cache compilations, speeding up subsequent builds | |||
| $ ./build.sh --sccache # use sccache to cache compilations, speeding up subsequent builds | |||
There was a problem hiding this comment.
Can we keep both options? sccache only really benefits docker users due to the absolute path requirement, ccache can still be benefitial for some developers
There was a problem hiding this comment.
I agree we should keep both options by transforming the flag into an option.
./build.sh --cache-tool=[ccache|sccache]
There was a problem hiding this comment.
Since sccache defaults to a local cache in the absence of a remote configuration, would that be sufficient? Or do you guys still think we should include both sccache and ccache?
There was a problem hiding this comment.
We're going to be removing ccache from our CI containers in favor of sccache. So if users are going to use our containers to build, then sccache will already be configured and they'll benefit from the shared cache by using absolute paths in the containers. If they build outside of our containers, then wouldn't sccache just behave similarly to ccache anyway (i.e. use a local cache)?
There was a problem hiding this comment.
If they build outside of our containers, then wouldn't sccache just behave similarly to ccache anyway (i.e. use a local cache)?
We shouldn't presume that sccache is sufficient to all developers / people building from source. sccache has more restrictions ( narrower compiler support, abs paths ) and these could make it a non-starter for these people. By making the option flexible by having the user specify the name instead of hardcoding it, we support more use cases at no real cost.
There was a problem hiding this comment.
@dantegd @robertmaynard Is it cool if we revert these changes made to BUILD.md and the associated files, and add them as a future enhancement? This way we don't hold up the rest of the PR from being merged.
There was a problem hiding this comment.
Sounds good by me, @AyodeAwe can you open a GH issue to track?
|
rerun tests |
Codecov Report
@@ Coverage Diff @@
## branch-22.04 #4534 +/- ##
===============================================
Coverage ? 85.71%
===============================================
Files ? 236
Lines ? 19365
Branches ? 0
===============================================
Hits ? 16599
Misses ? 2766
Partials ? 0
Flags with carried forward coverage won't be shown. Click here to find out more. Continue to review full report at Codecov.
|
|
@gpucibot merge |
This PR replaces `ccache` with `sccache`. Authors: - Jake Awe (https://github.com/AyodeAwe) - AJ Schmidt (https://github.com/ajschmidt8) Approvers: - Dante Gama Dessavre (https://github.com/dantegd) - AJ Schmidt (https://github.com/ajschmidt8) URL: rapidsai#4534
This PR replaces
ccachewithsccache.