IRC channel logs

2016-12-27.log

back to list of logs

<paroneayea>"let's use statprof to find out what the slowest procedure in 8sync is"
<paroneayea>% cumulative self
<paroneayea>time seconds seconds procedure
<paroneayea> 96.32 2.11 2.11 anon #x7f1c33c32fc0
<paroneayea> http://3.bp.blogspot.com/-TsDt67JnMro/UAa4bLXCjhI/AAAAAAAABE8/Qh8BNLPM0-8/s1600/delim.jpg
<paroneayea>(statprof is pretty great in general! no idea where to find that closure though)
<paroneayea>oh
<paroneayea>hrm.
<paroneayea>huh
<paroneayea>well looking at the 'tree display style gives some clues, maybe
<paroneayea>wingo: fibers only exposes time delays via sleep, but I assume usleep must be planned...
***logicmoo is now known as dmiles
<jmd>I want to create a soft-port which maintains some state. Any ideas where I can keep the state data?
<wingo>paroneayea: fibers' sleep procedure takes a real, not an integer
<wingo>so you can sleep for 1e-5 or whatever
<wingo>guile's sleep procedure should also do the same at some point; ArneBab had a patch
<paroneayea>wingo: ah nice
<paroneayea>I've become distracted from doc-writing by cutting down 8sync's agenda
<paroneayea>I've sure learned a lot about programming in scheme since I started 8sync about a year ago
<paroneayea>and programming in this domain
<paroneayea>enough to realize what a hot mess I made
<paroneayea>but progress I guess?
<paroneayea>ACTION cleans up some more
<paroneayea>"it was a pile of garbage, but hey! it ran"
<wingo>throwing away code is the best
<wingo>i have a work-stealing patch to land soon in fibers so that we will get automatic parallelism
<wingo>well
<wingo>i mean that you can just allocate N cores for fibers and mostly you get to take advantage of all cores
<wingo>devil is in the details of course
<paroneayea>I left a lot of unused code in 8sync for the potential future where the agenda would be completely functional / immutable
<paroneayea>and it was slowing things down
<paroneayea>in reality, I should have added that code when I was adding the feature
<paroneayea>not when I was speculatively possibly going to add it
<paroneayea>but at least it was a good mental exercise.
<paroneayea>wingo: cool about automatic parallelism in fibers
<paroneayea>I'm very curious about how it'll work out
<paroneayea>argh
<paroneayea>myseterious bug in my code, still thrashing CPU even after cleanup
<paroneayea>I just can't seem to track it down.
***holomorph is now known as Guest71080
***holomorp1 is now known as holomorph
<paroneayea>wow, ok.
<paroneayea>something about aborting to the prompt is what's slowing things down, but that makes no sense to me.
<paroneayea>ohh, I might know after all.
<paroneayea>I could be re-parameterizing things again...
<paroneayea>yep I'm pretty sure that's what's happening. harrrrr....
<paroneayea>yesssssss
<paroneayea>that was it
<paroneayea>and such an easy fix now that I'm less of a moron than I was when I started doing this stuff.
<paroneayea>wow, I sure didn't understand dynamic foo when I started 8sync :)
<amz3>#wiiiiiiiiiiiiiiiiiiinnnnnnnnnnnnnnnnnnnnnnnn
<amz3>there is a new version of the reasoned schemer coming out
<amz3> https://twitter.com/bodil/status/794479599812038657
<paroneayea>\\o/