|
04:06
Geth joined,
lizmat joined
05:43
ShimmerFairy left
05:44
ShimmerFairy joined
10:15
ShimmerFairy left,
ShimmerFairy joined
10:22
librasteve_ joined
15:20
apogee_ntv left
15:21
apogee_ntv joined
|
|||
| timo | something i randomly stumbled upon the other day is the ops arrays of our dispatcher programs. we do generate a boatload of these during setting compilation and it turns out (probably thanks to the constants and gc constants being put into different arrays) there are only relatively few distinct ones that are duplicated up to a few thousand times. they are individually relatively small, however | 16:18 | |
| of course if you aggressively intern these, you now have the very possible risk of leaking these programs if the frames they were generated for get garbage collected and any of the interned programs becomes actually unused. now you need to extend garbage collection to these tiny blobs of memory that were previously free of any reference counting or usage tracking | 16:36 | ||
| and of course core setting compilation is a relative outlier, and that's the only place i've analysed for this thing so far | 16:37 | ||
|
17:35
Sario left
17:40
sario528 joined
17:58
sario528 is now known as Sario
|
|||
| Geth | MoarVM/tail_call_interpreter: 6e1c5a60d3 | (Timo Paulssen)++ | build/probe.pm Fix probing for musttail and preserve_none |
20:50 | |
| MoarVM/tail_call_interpreter: 2489c46808 | (Timo Paulssen)++ | 7 files Tail-Call style for interp.c |
20:53 | ||
| MoarVM/tail_call_interpreter: f5bd647755 | (Timo Paulssen)++ | 4 files Split MVM_CGOTO into "has feature" and "use for interp" so that we can have interp.c as musttail and program.c as Computed Goto |
|||
| MoarVM/tail_call_interpreter: 0c56bf54b0 | (Timo Paulssen)++ | build/probe.pm Fix probing for musttail and preserve_none |
|||
| timo | and a rebase | ||
|
22:04
apogee_ntv left
22:05
apogee_ntv joined
22:27
harrow left
22:42
harrow joined
|
|||
| Geth | MoarVM: MasterDuke17++ created pull request #2001: Bump mimalloc to v2.3.2 |
23:56 | |