|
01:56
ilbot3 joined
03:05
evalable6 joined
03:32
jpf joined
04:34
evalable6 joined
05:32
domidumont joined
05:33
brrt joined
|
|||
| brrt | good * #moarvm | 05:33 | |
|
05:53
evalable6 joined
|
|||
| brrt | okay, my current hypothesis is that there is a frame handler annotation on a phi node | 05:56 | |
| that's a new thing, i think | |||
|
05:59
domidumont joined
|
|||
| brrt | yep, that's the thing | 06:00 | |
| that means my MVM_spesh_iterator_skip_phi is the thing that's broken | |||
| okay, ehm. | |||
| i have a solution but i'm not happy | |||
|
06:02
domidumont joined
|
|||
| brrt | meh, i may have find a nicer fix | 06:11 | |
| *found | 06:12 | ||
|
06:18
patrickz joined
|
|||
| Geth | MoarVM: e001239a14 | (Bart Wiegmans)++ | tools/jit-dump.pl jit-dump.pl - rename option From 'dir' to 'directory' |
06:25 | |
| MoarVM: a51cd95365 | (Bart Wiegmans)++ | 2 files JIT [log] more noisy log We misnamed a dynamic_label control as a throwish control, which was very confusing. And - if we're going to be noisy, why not be really noisy. |
|||
| MoarVM: bedc851138 | (Bart Wiegmans)++ | 2 files JIT [expr] process annotations on PHI node Presumably due to code elimination, (frame handler) annotations can sometimes end up on PHI nodes. Previously the expr JIT did not process these because the iterator skipped such nodes. That was not such a good idea as it seems to have led to a regression in Net::HTTP, which was fortunately caught by integration tests. |
|||
|
07:04
brrt joined
|
|||
| brrt | .tell AlexDaniel I believe commit bedc851138 fixes the blocker | 07:07 | |
| yoleaux | brrt: I'll pass your message to AlexDaniel. | ||
| AlexDaniel` | . | ||
| AlexDaniel | . | ||
| yoleaux | 07:07Z <brrt> AlexDaniel: I believe commit bedc851138 fixes the blocker | ||
| AlexDaniel` | I can easily test it once nqp and rakudo bumps are in place | 07:08 | |
| nwc10 | brrt: you're doing it wrong! When you push a load of commits, the commit bot is supposed to fall off the channel. Not the author. OK? :-) | ||
| brrt notes that the timestamp coincides with the arrival of my train | 07:09 | ||
| AlexDaniel` | let's not assume who's a bot and who is not :) | ||
.oO( stoprobotabuse.com/ ) |
07:10 | ||
| nwc10 | well, I have seen brrt give a talk at a conference, so I'm pretty sure he's not | 07:17 | |
| brrt | that's a fair point | ||
| nwc10 | I have not yet seen Geth give a talk | ||
| brrt | on the other hand | ||
| nwc10 | or make a commit | ||
| brrt | you have not seen me make a conference talk whilst on irc | ||
| so those might be separate identities | 07:18 | ||
| nwc10 | that is a good point | ||
| Geth | MoarVM/jit-expr-optimizer: f5b16ba01e | (Bart Wiegmans)++ | src/jit/expr.c Add expression tree equivalence function This is intended to be used in the optimizer. I'm not sure if I should add a 'depth' parameter. |
||
| MoarVM/jit-expr-optimizer: f1d6abf0f6 | (Bart Wiegmans)++ | 6 files [JIT] Simple tree-rewriting optimizer For some reason the multiple-referenced-loads optimization never seems to run, which is funny because I really expected it to. But the const idx -> addr optimization does. This probably needs a few more layers of abstraction, but for now it works. |
|||
| brrt | those are not new commits though, those are rebased | 07:19 | |
|
07:25
evalable6 joined
|
|||
| brrt | i think i've found why the optimizer couldn't optimize away double loads | 07:25 | |
|
08:25
TimToady_ joined,
[Coke]_ joined,
jnthn_ joined,
Util_ joined
08:27
timo joined
08:43
robertle joined
08:44
AlexDaniel` joined
08:48
greppable6 joined,
bloatable6 joined,
coverable6 joined,
evalable6 joined,
squashable6 joined
08:54
coverable6 joined
|
|||
| Geth | MoarVM: 4eadf94599 | (Stefan Seifert)++ | src/core/nativecall.c JIT compile native calls with at most 1 string argument |
09:05 | |
| MoarVM: 008612fded | (Stefan Seifert)++ | src/core/nativecall.c Support unsinged ints when JIT compiling native calls |
|||
| MoarVM: e051ee3cd4 | (Stefan Seifert)++ | 3 files JIT compile native calls with VMArray arguments |
|||
| brrt | not to be outdone :-P | 09:06 | |
| nine++ | |||
| nine | Turns out, thanks to jnthn++ moving string encoding into the high level code, Inline::Perl5 doesn't actually pass strings to C code at all. It's just bufs. And VMArrays are trivial to support :) | ||
| brrt | that's fairly cool | 09:07 | |
| nine | So what's missing now is rw args. I've made some decent progress however on the way back from my vacation :) | ||
| brrt | that's what holidays are for | 09:08 | |
| by the way | |||
| am i mistaken in having seen a call-for-speakers at fosdem 2018? | |||
| jnthn | No, there was one | 09:09 | |
| moritz | it was mentioned in a p6weekly the past few weeks | 09:15 | |
|
10:11
zakharyas joined
12:30
dogbert2 joined
12:42
travis-ci joined
|
|||
| travis-ci | MoarVM build errored. Bart Wiegmans 'JIT [expr] process annotations on PHI node | 12:42 | |
| travis-ci.org/MoarVM/MoarVM/builds/289362912 github.com/MoarVM/MoarVM/compare/6...dc8511387a | |||
|
12:42
travis-ci left
12:46
releasable6 joined
|
|||
| brrt | i'm not sure that's actually true, it appears to have simply no output at all | 12:49 | |
|
12:57
AlexDaniel` joined
|
|||
| Guest2271 | easy enough to restart that job | 13:33 | |
|
14:02
zakharyas joined
15:03
brrt joined
|
|||
| Geth | MoarVM/spesh_faster_shutdown: f75dfa72f9 | (Timo Paulssen)++ | 3 files allow VM shutdown to cause early spesh exit. |
15:20 | |
| timotimo | this will also want to kick off a submission from the main thread or something to cause the worker to run into its "should i keep running" check | 15:40 | |
| jnthn | Why bother? If it's sat waiting on a queue, it's not using any resources | 15:41 | |
| timotimo | i'm not entirely sure how the quit process really works in this case | ||
| running to the end of the entry point seems a surefire way to get the thread to disappear nicely | 15:42 | ||
|
15:42
domidumont joined
16:25
TimToady joined
16:33
robertle joined
17:49
domidumont joined
17:50
patrickz joined
18:15
patrickz joined
19:39
evalable6 joined
21:19
brrt joined
21:24
zakharyas joined
22:21
brrt joined
|
|||