Closed chm-diederichs closed 2 months ago
As it currently stands, this will make all builds sequential by default as
--parallel 1
is passed to CMake unless something else is specified. Let's use the number of available cores as the default instead; might as well put them to use!
I think by default cmake builds sequentially, maybe it is better to use os.availableParallelism()
then, what do you think?
I'm just worried it may be a little aggressive to allocate all cores by default.. So maybe Math.max(1, os.availableParallelism() >> 1)
?
i think the normal saturation default is cores + 1 from back in my day
The default depends on the build system used. Ninja, for example, will be as parallel as possible unless otherwise specified, whereas MSBuild will be strictly sequential.
Ok, in which case seems appropriate to defer to the build system default
Perfect, that works as well!