| samcv | good * moarvm | 00:03 | |
|
01:48
ilbot3 joined
|
|||
| MasterDuke | .tell nine i think i got the same segfault as you when trying to profile compiling rakudo, rt.perl.org/Ticket/Display.html?id=131653 | 03:15 | |
| yoleaux | MasterDuke: I'll pass your message to nine. | ||
|
04:22
geekosaur joined
|
|||
| samcv | well this could work gist.github.com/f4db547ef71ad9572f...b5c54dedfb trying it and the calculation is pretty close without hiving to do binary search afterward | 04:59 | |
| yoleaux | 02:53Z <[Coke]> samcv: node-gyp is borked: github.com/nodejs/node-gyp/issues/...-310951976 | ||
|
05:59
brrt joined
|
|||
| brrt | ohai #moarvm | 06:20 | |
| so, i've figured out that the expr JIT doesn't support OSR yet | |||
| i suppose that shoul'dve been more obvious than i though | 06:21 | ||
| t* | |||
|
06:23
domidumont joined
06:29
domidumont joined
07:16
zakharyas joined,
brrt joined
07:48
KDr2__ joined
07:51
dogbert2 joined
08:03
ilbot3 joined
08:18
domidumont joined
|
|||
| jnthn | morning, #moarvm o/ | 08:45 | |
| brrt | morning jnthn | ||
| say, i've been mulling a bit about the OSR problem | 08:46 | ||
| two conclusions stand out | |||
| - i should have a function to add an OSR label callable from expr tree buidling | |||
| - i should make the expr tree builder support such intenral labels | |||
| and the corollary to that second point | 08:47 | ||
| - once we have internal labels, there is very little stopping us from having the expr tree builder consuming entire graphs | |||
| jnthn ponders what the OSR stuff actually needs from the JIT | 08:48 | ||
| brrt | well, it's the other way around really | ||
| the JIT needs to set up stuff so that it can invoke OSR | |||
| eh | |||
| OSR can work on it | |||
| jnthn | It's the interpreter that triggers OSR though | ||
| Then we just jump in to the code | |||
| brrt | yeah. and i have reason to believe that's currently broken with the expr JIT | ||
| jnthn | So yeah, the JIT needs to make it so we can do that | ||
| brrt | :-) | 08:49 | |
| jnthn | Though I figured that wasn't much different than being able to jump back into code when we took it off the stack for other reasons. | ||
| brrt | it is the same process | ||
| that jit_entry_label was one lucky guess, really, as far as design decisions go | |||
| jnthn | Aye, so I guess my surprise boils down to "wouldn't the expr JIT already be handling that label stuff anyway" :-) | 08:51 | |
| Or "if those are left to the lego JIT then why isn't the OSR case also" :) | |||
| brrt | well, then we are facing the same question, really :-) | 08:52 | |
| because, i don't know why that is just yet | |||
| jnthn | ah, OK | 08:53 | |
| brrt | okay, hmmm | 08:56 | |
| the sp_fastcreate i was thinking of, doesn't have an OSR annotation associated at all | |||
| Geth | MoarVM: 3a4c23eca4 | (Jonathan Worthington)++ | src/io/syncfile.c Remove pipeable implementation in syncfile. This is dead code after the removal of synchronous processes. Leave the I/O vtable slot behind in io.h, since it'll soon be repurposed. |
08:58 | |
| brrt | and, the bindkey_o later, doesn't either | 09:10 | |
| hmm | 09:11 | ||
| so this is something fishier still | |||
| Geth | MoarVM: a4f7c84693 | (Jonathan Worthington)++ | src/types.h Clean up some leftovers in types.h. |
09:32 | |
| MoarVM: f46a06eb24 | (Jonathan Worthington)++ | 4 files Provide way to get async handle from IO handle. So that we can make `permit` work correctly on the async proc handle. |
|||
|
10:11
robertle joined
|
|||
| Geth | MoarVM: d0dd989589 | (Jonathan Worthington)++ | src/gc/roots.c Make sure that we mark the permit queue. |
10:14 | |
| MoarVM: 10f4840ad9 | (Jonathan Worthington)++ | src/io/eventloop.c Correct thinkos in permit handling. |
|||
|
10:48
DeadDelta joined
15:08
ilbot3 joined
15:47
robertle_ joined
15:59
greppable6 joined
16:47
greppable6 joined
17:11
domidumont joined
19:35
ChanServ joined
20:41
vendethiel joined
20:46
Ven joined
20:57
Ven_ joined
22:18
vendethiel- joined
|
|||
| lizmat | and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2017/06/26/...way-there/ | 22:36 | |
|
22:37
dogbert2 joined
|
|||