Hello all! I have a very large document that I am building (25k pages - I know, I know, I didn’t actually create it) and am looking to better understand the updates from 0.12.0 and explicit pagebreaks.
Based off of this blog post about release 0.12.0 is mentions that under the right circumstances you could make use of all your CPU cores. The bulk of my document is my appendix (pages 931 - 25k). Initially there were no explicit pagebreaks, so I added 18 at various points in the appendix depending on the heading depth to see if it would do anything to the compilation time. When I monitor the resources of the server I’m using it’s currently using one thread and it’s spiked at around 100%. It’s possible that it could jump from one thread to another, but doesn’t seem to utilize more than one the bulk of the compile.
When I initially kick off the typst compile command it seems like it does utilize more than one core, but eventually settles on the one. Anything I should do differently or try? Or is this the expected behavior?
The use of --timings has been helpful in debugging this. I built a slimmed down version of the document in a few different scenarios - no pagebreaks in appendix, a few, and many. It does appear when additional pagebreaks are added it does indead spread it out amongst the thread but it actually slows down the compile time the more pagebreaks I add.
I think ultimately I need to revisit the styling / syntax I’m using to see if I can better optimize there and evaluate the timings after the changes.
Hello. After reading the changelog, it says that it’s multithreaded and not multi-cored, but apparently it uses the OS threads, which are on modern processors can be parallelized. I’m not familiar with all concepts, but I just checked, and it can use 700% of CPU for a moment. For comparison, you can use the --jobs flag and see if there is a magic number that makes it the fastest. Without the actual document, I can’t really say about how you can optimize the compilation time. But --timings is indeed your best friend for profiling the compilation.
Ahh, thank you for the clarification on the multithreaded vs. multi-cored! I’ll investigate my hardware and see if that’s an option. I’ll also try the --jobs flag. Thanks for the follow up!