Spicy Compilation Long Times

Hi,

I’m running sudo zkg -vv unbundle bundle.zkg which takes quite some time to compile 24 packages.

I see most of the time is spent in spicyz on a single core:

Is it possible to optimise/use all cores for spicyz and friends?

I have in some of my other projects CMAKE_BUILD_PARALLEL_LEVEL=32 and MAKEFLAGS="-j 32".
I think, by default cmake does use all cores (please correct me if I’m wrong)?

Am I correct in stating that CFLAGS and CXXFLAGS are not applicable here?

During zkg unbundle packages will be compiled and installed one after another. Now in each package containing Spicy analyzers typically one CMake target per analyzer will be created which can and should be built in in parallel if there are multiple. Each analyzer gets created by a single invocation to spicyz which does not parallelize.

I suspect most of your packages contain exactly one Spicy analyzer in which case you would see exactly the reported behavior. There is not much you can do about this.

1 Like

Thank you for the confirmation. Am I being dumb by asking, can one run multiple zkg install commands concurrently to build packages in parallel?

can one run multiple zkg install commands concurrently to build packages in parallel

AFAIK zkg doesn’t prevent running multiple processes concurrently, but also does not protect its internal state in any way, so doing this could lead to failures or silently corrupted state.

1 Like