On Tue, Aug 30, 2022 at 12:38:20PM +0200, Jean-Marc Lasgouttes wrote: > Le 30/08/2022 à 12:02, Pavel Sanda a écrit : > > The bottlenecks I typically experience are of two kinds: > > 1) pictures conversion (having parallel code would be great even for > > document load when ton of pics gets converted!) > > Yes, the speed of this is questionable right now.
OK, sounds like this is a candidate then for parallelization. > > 2) final pdflatex compilation > > > > The latex generation part is just neligible. > > If it is not, it should be profiled before trying to fix something that has > not been measured. Makes sense. > > Point 1 seems to be a good candidate for paralelization. External calls > > (like knitr) might be as well in case, again, its evaluation can't be > > dependent on the previous insets. But in this case I am almost sure some > > previous knitr inset might produce something needed for the subsequent ones > > ... > > Concerning knitr, we pass everything in one big file, and indeed, this is > usually highly sequential. I don't think there's hope of speeding up knitr within a single .lyx file. Actually, there are ways to use parallelization in R itself and those should be preferred for a single .lyx file. Scott
signature.asc
Description: PGP signature
-- lyx-devel mailing list lyx-devel@lists.lyx.org http://lists.lyx.org/mailman/listinfo/lyx-devel
