|
01:33
TEttinger joined
06:00
FROGGS joined
09:08
vendethiel joined
09:16
brrt joined
|
|||
| brrt | MoarVM *almost* builds with gcc on windows | 09:22 | |
| but not quite :-( | |||
| dyncall issues | |||
| timotimo | so try libffi instead? | 09:43 | |
| nebuchadnezzar | o/ | ||
| timotimo | heyo | 09:44 | |
| jnthn | o/ | 09:45 | |
| nebuchadnezzar | time to start having nqp up to date in Debian | 09:48 | |
| timotimo | oh, cool :) | 09:49 | |
| jnthn | \o/ | 09:50 | |
| FROGGS | brrt: it builds without problems using gcc on my Windows 7 | 11:48 | |
| timotimo: you'd probably need libffi prebuild dlls | |||
|
12:18
TimToady joined
13:18
brrt joined
|
|||
| brrt | fwiw: gcc 4.9.2 on windows 10 enterprise | 13:19 | |
| (evaluation license ;-)) | |||
| gcc supplied by strawberry perl :-) | 13:20 | ||
|
13:39
vendethiel joined
13:54
FROGGS_ joined
17:21
Peter_R joined
|
|||
| nwc10 | jnthn: in github.com/MoarVM/MoarVM/commit/d4ac6afa75 why use MVMint64 for a 1 bit flag? | 19:11 | |
| I think that's somewhat less than awesomely efficient on some 32 bit platforms. | |||
| jnthn | Depends how smart the C compiler in question is I guess... | 19:12 | |
| nwc10 | the ABI may force thigns | ||
| jnthn | True | 19:13 | |
| nwc10 | I don't actually happen to have anythign to hand to confirm/deny this assumption | ||
| but I *think* that on a 32 bit system, the caller will have to write two 32 bit words | |||
| because the caller has no idea that the callee only cares for 1 bit of 1 of them. | |||
| "write two 32 bit words" is probably actually "assign to 2 registers" | 19:14 | ||
| jnthn | I guess the situation on 64-bit isn't pessimized by making it an MVMint32 | ||
| nwc10 | that would be my hunch | ||
| jnthn | Given this code is called once per GC run, I'm not sure we stand to lose much in this case :) | 19:15 | |
| nwc10 | it was more a "eh, why is it that way?" question | ||
| agree, we've burned more time in this converstation that likely will ever be saved by fixing it | 19:16 | ||
| across all MoarVM ever. Or at least "this year" | |||
| jnthn | It's that way 'cus I'd just spent a bunch of time hunting a GC race condition, and didn't consider this relatively minor detail... | ||
| nwc10 | oh, and I comment on the bike shed sized commit? | 19:17 | |
| Not the nuclear reactor? | |||
| (which is far more likely to go boom) | |||
| so boom | much messy | wah :-( | |||
| jnthn | Well, the patch fixes the race | 19:18 | |
| One of those where the time to work out what needing fixing took a lot longer than fixing it. | 19:20 | ||
| dalek | arVM: c40e8a7 | jnthn++ | src/gc/collect. (2 files): 32 bits will be sufficient for this flag; nwc10++. |
19:23 | |
| jnthn | Now, if you can find one of those on a hot-path... :) | ||
| nwc10 | then you will buy me a beer*? [* offer only valid in Köln :-)] | 19:31 | |
| [for the local beer. In the correct glass] | 19:32 | ||
| Oh my. There is such a thing as a Doppelmaß. I didn't know this. | 19:34 | ||
| jnthn | A...what???! | 19:40 | |
| Either my Germany is rusty of that's a LOT of beer... | |||
| timotimo | it's probably a lot of beer | ||
| nwc10: is it more than a Stiefel? | |||
| jnthn | Heh, funny to call my German rusty, I never learned the language. :P | ||
| I can just order beer, be polite to train conductors, and know some bad words used in Rammstein songs I listened to as a teenager :P | 19:41 | ||
| nwc10 | I don't know what a Stiefel is, in this context. | ||
| The source was en.wikipedia.org/wiki/Austrian_cuisine#Beer -- ... two litre Doppelmaß in the Bavarian style | 19:42 | ||
| jnthn | Maybe it's named so 'cus if you drink all of it, it'll stiefel your productivity... | ||
| timotimo | hahaha | 19:44 | |
| it's an actual boot-shaped glass | |||
| nwc10 | oh, Stiefel. Yes. I should know that word | ||
| Gummistiefel | |||
| timotimo | :) | ||
| well, yeah, except the stiefel you'll drink the beer out of isn't a gummistiefel | 19:45 | ||
|
19:50
FROGGS joined
20:48
TEttinger joined
|
|||