scripts: gen_boards_catalog.py: add verbose flag to Twister #97733
Conversation
Pass verbose flag to Twister to get better insights into what boards are being built and see progress. Signed-off-by: Benjamin Cabé <[email protected]>
kartben
added
the
Trivial
zephyrbot
added
area: Documentation Infrastructure
size: XS
|
| *[arg for path in EDT_PICKLE_PATHS for arg in ('--keep-artifacts', path)], | ||
| *[arg for path in RUNNERS_YAML_PATHS for arg in ('--keep-artifacts', path)], | ||
| "--cmake-only", | ||
| "-v", |
There was a problem hiding this comment.
based on my tests I don't think it's necessary - while it might be an issue for "real" Twister test runs, -j 96 seems to be just fine for the CMake-only hello world builds, and not noticeably faster/slower than e.g. -j32
What I would love to know though is why the Sphinx part of the build is so slow, despite Sphinx being given 8 threads. In fact it's even slower with more threads, and (marginally) faster with e.g 4. It looks like there is some kind of bottleneck somewhere (filesystem??) and I really hope @stephanosio maybe has an idea about what might be going on. FWIW while the Twister stuff is done in like 2m30s, the sphinx part is actually slower than we had on the Github runner!(?)
There was a problem hiding this comment.
the sphinx part is actually slower than we had on the Github runner!(?)
Are you sure? 4xlarge zephyr-runners are allocated 16 vCPUs, 32GB RAM and have local NVMe SSDs for storage; so, they should not be any slower than the GitHub-hosted runners ... if any, they should be faster.
4 participants
Pass verbose flag to Twister to get better insights into what boards are being built and see progress.