Da
Show enters and exits. Hide enters and exits.
| 17:37:21 | dwaite | waves |
| 18:04:28 | brixen | hey dwaite |
| 18:19:42 | rue | Morning |
| 18:21:54 | dwaite | how is everyone? |
| 18:22:05 | dwaite | hasn't been hanging around here nearly enough |
| 18:28:06 | rue | Somewhat poorly, thanks, yourself? |
| 18:58:37 | fynn | Hey. A few weeks back, there was talk of a nice, comprehensive blog post about JIT compilation (by evan, I believe). Is there an ETA for it? |
| 19:07:24 | brixen | fynn: it's in the works |
| 19:07:38 | brixen | should be coming in the next couple weeks |
| 19:08:18 | brixen | it will probably start with an overview of the arch and later articles will go into greater details about particular aspects of the jit |
| 19:15:42 | Zoxc | it will be done once evan has nothing else to do? |
| 19:18:11 | Zoxc | that's what I said! |
| 19:18:54 | brixen | Zoxc: actually, that's different |
| 19:19:21 | brixen | your work queue could be saturated but you could still allocate a slice to one task |
| 19:19:31 | brixen | vs only working on the task if your work queue is empty |
| 19:21:41 | scoopr | is it pre-emptive or co-operating? :P |
| 19:22:28 | Zoxc | the poor post haven't even been assigned a time slice, so I guess it will be assigned when evan has nothing else to do |
| 19:23:43 | dbussink | evan`: did you have time to take a look at the issue i was seeing? |
| 19:23:49 | dbussink | evan`: just curious :) |
| 19:26:19 | dbussink | evan`: ah, was mainly curious if i actually got a good reproduceable example |
| 19:30:28 | dbussink | evan`: you're nick is a bit weird ;) |
| 19:30:46 | dwaite | evan`: whats your load level if you are doing time slicing? |
| 19:31:02 | evan | that was wierd. |
| 19:31:03 | evan | weird. |
| 19:31:42 | evan | dwaite: 0.3223 chromistoms per unit |
| 19:32:32 | boyscout | Fix JIT block inlining (restricted) - 4dc8991 - Evan Phoenix |
| 19:33:02 | dwaite | 0.32? you could handle a lot more load |
| 19:33:10 | dwaite | hands evan a movie to play back |
| 19:34:04 | evan | well, my direction-o-meter has an usually frequent poll rate |
| 19:34:12 | evan | so it takes up a lot of slices. |
| 19:34:36 | dbussink | evan: did that change fix the profile issue i was seeing? |
| 19:34:42 | dbussink | evan: or is that something else? |
| 19:34:52 | evan | it steals cycles from the longterm fact storage cache monitor |
| 19:35:02 | evan | which means my memorization skills suck. |
| 19:35:23 | evan | dbussink: which? |
| 19:35:26 | evan | the one I just did? |
| 19:35:28 | evan | it's unrelated |
| 19:35:30 | dwaite | whats slated for rc3? |
| 19:35:33 | evan | your fix is yet to come. |
| 19:35:34 | dbussink | evan: ah ok |
| 19:35:45 | dbussink | evan: ok, sorry if i'm too annoying with reporting issues :P |
| 19:39:30 | boyscout | CI: rubinius: 4dc8991 successful: 3024 files, 11747 examples, 35951 expectations, 0 failures, 0 errors |
| 19:40:31 | evan | dbussink: nope, no problem. |
| 19:43:56 | fynn | brixen: thanks, I look forward to it. |
| 19:44:30 | fynn | I'm especially curious about differences/similarities between Rubinius and Unladen Swallow in the JIT department. |
| 19:44:57 | fynn | Collin Winter (one of the lead developers of US) said something about how he envies you guys. |
| 19:46:41 | fynn | I'm not sure why because if anything, Ruby seems to be more dynamic than Python, as well as pose other harsher demands on the JIT (for instance, in the way that class declarations themselves are more powerful, etc. |
| 19:52:10 | dbussink | evan: hmm, after your change my repro doesn't work anymore :( |
| 19:52:20 | evan | doesn't work? |
| 19:52:40 | dbussink | evan: it doesn't segfault now anymore like it did |
| 19:52:45 | evan | thats good! |
| 19:52:49 | dbussink | it does show similar failures still |
| 19:52:53 | evan | oh. well. |
| 19:52:55 | dbussink | during spec runs |
| 19:52:59 | evan | well figure it out. |
| 19:53:06 | evan | rubyspec runs or DM spec runs?L |
| 19:53:11 | dbussink | dm spec runs |
| 19:53:22 | dbussink | some specs fail when the jit is enabled |
| 19:53:35 | dbussink | and i had a segfault before too, which i can imagine makes debugging easier |
| 20:51:16 | dbussink | evan: ah, i can still reproduce the segfault if i remove all rbc's in the dm-core specs first |
| 20:51:34 | dbussink | evan: so it starts compiling a lot of stuff which triggers the jit and subsequently exposes the issue |
| 21:31:33 | rue | Formalised git workflow: http://nvie.com/archives/323 [+❤] |
| 21:45:43 | jarib | rue: nice link |
| 21:52:59 | rue | Yeah |
| 21:53:20 | rue | It is a bit too formal for me, but a good guideline definitely |
| 23:28:51 | Defiler | Hah, this is excellent, and completely true: http://yugui.jp/articles/853 |
| 23:31:17 | ezmobius | haha awesome |
| 23:33:00 | brixen | nice |