These are chat archives for dereneaton/ipyrad

18th
Feb 2016
Isaac Overcast
@isaacovercast
Feb 18 2016 04:33
I think it's an artifact of the old assembly format and how it saved. The CLI used to need to pass in the full path on ip.core.Assembly("wat") creation, but having the assembly name be the full path fsck'd things up. I think this was before we created the "assemblyname" parameter, and also before the json load/save, which I see does save to the project directory by default. I don't _think it's necessary anymore. We could probably get rid of it, might take a little testing, but i'd be happy to do it, it's a little hax
Deren Eaton
@dereneaton
Feb 18 2016 15:34
yeah, I removed it and everything seems to be still running fine in the API. That's what I thought from looking at the code.
Dude, I'm running on a full lane of RAD data right now on 10 processors and its crushing. After only 12 hours its already on step5.
and it could actually be running faster. I forgot I had some other things running so my CPUs are maxed.
Deren Eaton
@dereneaton
Feb 18 2016 17:34
It actually seems like step5 is sometimes the slowest step.
Deren Eaton
@dereneaton
Feb 18 2016 17:54
it just took 2 hours to finish one sample in step5. I mean, it will be faster with more processors, but still, that's a big place to look for speed improvement.