|
github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm Set by AlexDaniel on 12 June 2018. |
|||
|
00:24
lizmat joined
00:25
p6bannerbot sets mode: +v lizmat
06:46
MasterDuke left
07:42
domidumont joined
07:43
p6bannerbot sets mode: +v domidumont
07:49
domidumont left
08:35
Kaiepi joined
08:36
p6bannerbot sets mode: +v Kaiepi
08:38
brrt joined
08:39
p6bannerbot sets mode: +v brrt
|
|||
| brrt | \o | 08:39 | |
| nine++ for hllbool(for) ops | |||
| Geth | MoarVM: fb8976ab19 | (Bart Wiegmans)++ | src/jit/x64/emit.dasc [JIT] Pointers must be loaded by mov64 Because 'normal' x86 operands can only be 4 bytes wide. |
08:40 | |
| timotimo | oh, is that the cause of the crash? | 08:41 | |
| brrt | maybe? | 08:43 | |
| what crash | |||
| timotimo | segfaults during stage parse i think? | 08:44 | |
| brrt | let me try | ||
| timotimo | i can't reproduce 'em | ||
| but liz can. so maybe needs a mac to tickle it? | |||
| brrt | hmm, i'm not seeing a crash either | 08:48 | |
| .ask lizmat can you try again with moar HEAD? | |||
| yoleaux | brrt: I'll pass your message to lizmat. | ||
| timotimo | JIT_DISABLE=1 supposedly makes it go away | ||
|
08:51
fake_space_whale left
|
|||
| lizmat | seems to build ok now | 08:52 | |
| yoleaux | 08:48Z <brrt> lizmat: can you try again with moar HEAD? | ||
| timotimo | cool. | 08:53 | |
| brrt | \o/ | 08:56 | |
| lizmat | spectesting now, and then really off to breakfast (if there is still any) | ||
| brrt | with apologies to nine++, this is one of the reasons I'd really rather get rid of the lego JIT | ||
| timotimo | :) | ||
| how come using mov there instead of mov64 doesn't explode violently? is it fine with pointers that have the upper 32bit zeroed out? | 08:57 | ||
| lizmat | spectests clean also, brrt++ nine++ | 09:03 | |
| breakfast& | |||
| nine | ooooooooh thanks a lot brrt! | ||
| brrt: but why the change to uintptr_t? | 09:04 | ||
|
09:05
lizmat_ joined
09:06
p6bannerbot sets mode: +v lizmat_
09:08
lizmat left
|
|||
| brrt | because safety :-) | 09:08 | |
| uintptr_t is guaranteed by the standard to hold a pointer in an integer | |||
| an MVMuint64 will also hold it, but is not guaranteed to | 09:09 | ||
| so 'C as you understand it' will always work, 'C as specified' might find an edge case where it won't; and I'm being paranoid | |||
| nine | But I've used an MVMObject* which...is a pointer type? | 09:41 | |
| brrt | yes | 10:01 | |
| under the hood, what happens is that dynasm passes this as two 32 bit integers to the assembler, which are stitched together | 10:02 | ||
| this will complain if the 64 bit number is not an integer | |||
| hence the cast is necessary | |||
| nine | aaah...thanks for clearing that up :) | 10:05 | |
| brrt | :-) | 10:32 | |
| I kind of dislike how you have to know the under-the-hood details of dynasm and x86 for that.... it's a pretty leaky abstraction | |||
|
10:40
brrt left
11:47
MasterDuke joined,
p6bannerbot sets mode: +v MasterDuke,
MasterDuke left,
MasterDuke joined,
herbert.freenode.net sets mode: +v MasterDuke,
p6bannerbot sets mode: +v MasterDuke
12:02
brrt joined
12:03
p6bannerbot sets mode: +v brrt
12:10
brrt left
13:49
lizmat_ left
13:54
lizmat joined,
p6bannerbot sets mode: +v lizmat
15:00
zakharyas joined
15:01
p6bannerbot sets mode: +v zakharyas
15:04
brrt joined
15:05
p6bannerbot sets mode: +v brrt
15:22
dogbert17 joined
15:23
p6bannerbot sets mode: +v dogbert17
16:07
brrt left
16:13
lizmat left
16:39
ggoebel_ joined
16:40
p6bannerbot sets mode: +v ggoebel_
17:02
AlexDani` joined
17:03
p6bannerbot sets mode: +v AlexDani`
17:06
AlexDaniel left
17:12
AlexDani` is now known as AlexDaniel,
AlexDaniel left,
AlexDaniel joined,
hitchcock.freenode.net sets mode: +v AlexDaniel,
p6bannerbot sets mode: +v AlexDaniel
17:20
brrt joined
17:21
p6bannerbot sets mode: +v brrt
|
|||
| brrt | \o | 17:48 | |
| diakopter | :D brrt | 17:51 | |
|
17:53
domidumont joined
17:54
p6bannerbot sets mode: +v domidumont
|
|||
| timotimo | o/ | 18:05 | |
| brrt | ohai timotimo | 18:15 | |
| I'm suspicious that our current implementation of trunc_i8, etc. don't do what we think they do | 18:20 | ||
|
19:15
brrt left
19:17
domidumont left
19:31
fake_space_whale joined
19:32
p6bannerbot sets mode: +v fake_space_whale
19:40
lizmat joined
19:41
p6bannerbot sets mode: +v lizmat
19:57
robertle left
20:36
zakharyas left
20:38
fake_space_whale left
22:03
MasterDuke left
22:52
reportable6 left,
reportable6 joined
22:53
p6bannerbot sets mode: +v reportable6
|
|||
| AlexDaniel | blocker reminder: github.com/MoarVM/MoarVM/issues/951 | 22:58 | |
| timotimo | hum. do we do alignment in regular create, but not in fastcreate, perhaps? | 23:02 | |
| no, sp_fastcreate also just calls to MVM_gc_allocate_nursery | 23:04 | ||
|
23:05
reportable6 left,
reportable6 joined
23:06
p6bannerbot sets mode: +v reportable6
23:41
MasterDuke joined,
p6bannerbot sets mode: +v MasterDuke
23:42
MasterDuke left,
MasterDuke joined,
herbert.freenode.net sets mode: +v MasterDuke,
p6bannerbot sets mode: +v MasterDuke
|
|||