Amalgamated vs parallel build time.

[zevv]

When working on the Janet source, I often find my edit-to-run turnaround time just a bit too high to be comfy; even though I have a fairly modern 24 core machine, the final build of the amalgamated janet.c still takes up to 6 seconds because this last step can not be parallelized - shame!

As a workaround I have some local changes to my Makefile and boot.janet (see below) that adds a target for generating the boot code without amalgamating all the sources, and just building and linking all the individual objects the "old fashioned" way. This results in a build time of approx 1.5 seconds, instead of the 8 seconds of the normal target.

Are other people also bothered by the compilation time, and would it make sense to add a parallelizable build target to the Makefile for this?

zevv@8f43bf3

[sogaiu]

Nice speedup!

Perhaps it's because my machine isn't as nice (or may be I didn't do it right), but my speed up wasn't so dramatic. Somewhere in the 1-2 seconds range.

Oh wait...no, it doesn't look faster here. Must be doing something wrong.

[zevv]

You're doing a parallel build, right? try (replace NCPUS with a sane value for your machine)

make clean; time make -j NCPUS

vs

make clean; time make -j NCPUS quicky=1

[sogaiu]

Oh yes, pilot error indeed.

I get the same sort of numbers you reported...nice :)

Thanks!