We just got a new machine for building/testing a fork of rust. The machine has loads of cores.
I was watching htop while
x.py test was running and noticed that whilst some suites are highly parallel (e.g. rustdoc), others are not.
Is there any reason the tests cannot be more parallel?
isn't there a flag you pass to increase the number of jobs used for tests?
yeah, its in the
x.py test --help output:
... -j, --jobs JOBS number of jobs to run in parallel ...
would that resolve your issue?
there is also a
RUST_TEST_THREADS environment variable that I believe tells
compiletest to run the tests on multiple threads ; you may need to use a combination of these two things to get the best utilization of a machine with many cores.
(oh, but I think we already set that by default to the value sent in for
--jobs, so the latter bit is probably unneccesary)