Parallelization of souffle

[EvgSkv]

Dear Souffle team,
I'm running souffle reach predicate on livejournal dataset in csv format written follows:

.decl g(a:number,b:number)
.decl reach(a:number)
.decl result(a:number)

.input g(filename="livejournal.csv",delimiter=",",headers=true)

reach(0).
reach(y) :- reach(x), g(x, y).

result(a) :- reach(a).

.output result(delimiter=",")

Linux time command shows that user and real time are the same:

real	0m30.796s
user	0m29.998s
sys	0m0.676s

This means that no parallelization is happening, right? Why would that be the case? Do I need to enable it somehow?

Thank you very much for the great language!

[cwarden]

Use --jobs auto to use all cores.

[EvgSkv]

Perfect, I see it parallelized. Thank you!

Now most time is spent reading data. Is there a better format to use for souffle data other than csv? I tried using souffle to compress csv and read that, but it didn't improve speed.

Maybe csv is optimial, that's totally fine. Please let me know! Thanks a lot!

[cwarden]

Try importing the csv into a sqlite database and using that as the input for souffle.

[EvgSkv]

I will try that! Thank you very much!