OK, I've pushed a parallel-rendering routine which you can access by passing 
the `-p` or `--parallel` flag to any invocation of `raco pollen render`. (I 
will leave it undocumented for now until it has proven itself).

If you just do `raco pollen render -p` instead of `raco pollen render`, as I 
would expect, it's more wasteful in terms of overall rendering time (because it 
can't rely on the cache as heavily) but that the net time is shorter (because 
the waste is spread across multiple cores).

That being so, I got the best performance by doing `raco pollen setup` (which 
preheats the cache in parallel) followed by `raco pollen render -p` (which 
renders in parallel). By setting up the cache before starting any rendering, 
you minimize the possibility of duplication of work between parallel renders.

I have seen "cache lock" warnings from time to time in the console — AFAIK 
these are just notices from the cache subsystem that it's doing its job by 
blocking attempts at simultaneous disk writes by parallel jobs. It shouldn't 
affect the result.

Try it and see what you think.



> On Apr 23, 2019, at 9:02 AM, [email protected] wrote:
> 
> I'm still interested in possible gains from parallelizing the template step. 
> My current project will likely start out with around 400 separate documents 
> and grow from there, hence my interest in improving build times.

-- 
You received this message because you are subscribed to the Google Groups 
"Pollen" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to