bug#48007: computing derivations through inferior takes twice as long

2022-02-18 Thread Ludovic Courtès
Hi, Ludovic Courtès skribis: > Previously, each 'inferior-eval-with-store' would lead the inferior to > connect to the named socket the parent is listening to. With this > change, the connection is established once for all and reused > afterwards. > > * guix/inferior.scm ()[bridge-file-name]: R

bug#48007: computing derivations through inferior takes twice as long

2022-01-27 Thread Ludovic Courtès
Ricardo Wurmus skribis: > These patches look great to me and they work. > My real-world test case is now down to about 12 seconds. Good! Fixed the typo you mentioned on IRC and pushed as e778910bdfc68c60a5be59aac93049d32feae904. To summarize, the INFERIOR=y case still takes about twice as long

bug#48007: computing derivations through inferior takes twice as long

2022-01-27 Thread Ricardo Wurmus
Ludovic Courtès writes: > Ricardo Wurmus skribis: > >> I tried it in the GWL with the big RNAseq workflow I adopted from PiGx >> and the step to generate job scripts (which reference inferior packages) >> became considerably faster. There is still potential for improvement, >> but this change

bug#48007: computing derivations through inferior takes twice as long

2022-01-27 Thread Ludovic Courtès
Hi, Ricardo Wurmus skribis: > I tried it in the GWL with the big RNAseq workflow I adopted from PiGx > and the step to generate job scripts (which reference inferior packages) > became considerably faster. There is still potential for improvement, > but this change is the difference between not

bug#48007: computing derivations through inferior takes twice as long

2022-01-26 Thread Ricardo Wurmus
Ludovic Courtès writes: > The trick is to ensure the inferior maintains its object cache across > calls. The patch needs to be cleaned up because it peeks into > internals, but we should be able to do something along these lines and > optimize a couple of other things. Yeah, this makes sense.

bug#48007: computing derivations through inferior takes twice as long

2022-01-26 Thread Ludovic Courtès
Hi! Ricardo Wurmus skribis: > When INFERIOR is set to “y”, each package specification will be > looked up in the current Guix via an inferior. For any other > values of INFERIOR the specifications are resolved with the > current Guix (the very same Guix) directly. > > Here are the timings: >

bug#48007: computing derivations through inferior takes twice as long

2021-04-24 Thread Ricardo Wurmus
This bug report might be related to bug #48005. In the Guix Workflow Language we are always looking up packages through an inferior Guix. That Guix will in most cases be just the current Guix. As I was looking for ways to speed the GWL up, I noticed that the use of inferiors itself contribut