[libcxx-commits] [PATCH] D143158: [libcxx][test] Cover RISC-V in string.capacity test

[Xeonacid via Phabricator via libcxx-commits]

Xeonacid added a comment.

In D143158#4116199 <https://reviews.llvm.org/D143158#4116199>, @philnik wrote:

> In D143158#4116110 <https://reviews.llvm.org/D143158#4116110>, @asb wrote:
>
>> Thanks for the advice so far. One thing I wanted to flag early so as to avoid wasted effort is that the requirement on the linked doc <https://libcxx.llvm.org/AddingNewCIJobs.html#addingnewcijobs> "We may be reluctant to add and support CI jobs that take a long time to finish or that are too flaky." may be challenging for runners that need to use emulation. I haven't checked what proportion of libcxx tests are execution vs compile tests, but check-libcxx on a fast x86_64 host using qemu-user for to emulate a 'native' RISC-V environment is not fast.
>
> Do you have some estimate how long a run takes? Currently the slowest runners take ~45 minutes to complete a job.
>
> It would be an option to just have a minimal configuration running (i.e. set LIBCXX_ENABLE_RANDOM_DEVICE, LIBCXX_ENABLE_LOCALIZATION, LIBCXX_ENABLE_UNICODE, etc. to Off). That will probably reduce the runtime quite a bit (but of course also lower coverage). You can then enable these once there is native hardware available to run the tests (which should be the case soon-ish if I understand your commit message from D143172 <https://reviews.llvm.org/D143172> correctly).

It runs for ~2 hours both on 5950X with qemu-user and on SG2042 (64C RISC-V), which is likely the most powerful riscv hardware today.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D143158/new/

https://reviews.llvm.org/D143158