These are chat archives for dropbox/pyston

7th
Jan 2015
Bob Fang
@dorafmon
Jan 07 2015 00:55

Hi, so mine implmentation is slow, and the figures I have right now for running a simple fibonocci sequence program is:

slowpath_binop: 24157838
slowpath_callfunc: 24158499

while the master branch’s figure is

slowpath_binop: 36
slowpath_callfunc: 697

I just want to confirm my guess that it means that the IC is pretty much broken right? It is not related to the jitted code, but is the IC defuctioning causing the slow down

Kevin Modzelewski
@kmod
Jan 07 2015 01:00
Not quite sure what you mean
but those stats would mean that we're falling back on the slowpath a lot
Bob Fang
@dorafmon
Jan 07 2015 01:01
yeah I mean the only thing causes to fall back on the slow path is that the IC is not working properly right?
Kevin Modzelewski
@kmod
Jan 07 2015 01:01
usually, yeah
Bob Fang
@dorafmon
Jan 07 2015 01:01
cool
Kevin Modzelewski
@kmod
Jan 07 2015 01:01
There can be other reasons but that would make sense here
Bob Fang
@dorafmon
Jan 07 2015 01:02
cool I have been trying to trace down the cause of the slowdown all day in the IC part but have not got any clue, so I was thinking it might be something else. That’s why I ask to confirm my guess is on the right path :)
Bob Fang
@dorafmon
Jan 07 2015 23:11
hi in the optimised code I saw something like _ZN6pyston13interned_intsE
what does this mean?
Bob Fang
@dorafmon
Jan 07 2015 23:38
never mind it turns out I accidentally disabled IC for binop function, that’s why the performance is so poor.