This test is excessively slow on GPU targets, taking anywhere beween 5
and 60 seconds to complete each time it's run. See
https://lab.llvm.org/buildbot/#/builders/55/builds/52203/steps/12/logs/stdio
for an example on the NVPTX buildbot. Simply disable testing this on the
GPU for now.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Try disabling just the third set of loops, since the size of the checks grows exponentially it's likely that they're taking up most of the time.
I'd rather not limit the digits to just 0 and 1. If you want to shorten the tests maybe try 16? Otherwise probably just disable the whole test. Having the test enabled and running but not actually checking anything seems like a recipe for confusion.
Well we at least get the first loop's checks. This version takes 5ms on my machine, the original takes 6000ms, changing it to 16 takes 500ms.
How long does it take with a limit of 36 but disabling the loop that goes to the third digit?
I think that's the option I would prefer then. That way we get to test all of the 1 and 2 digit cases
libc/test/src/stdlib/StrtolTest.h | ||
---|---|---|
202 | nit: Clarify this comment with the new design |
nit: Clarify this comment with the new design