|
01:49
ilbot3 joined
02:45
MasterDuke joined
03:53
geekosaur joined
06:43
domidumont joined
06:49
domidumont joined
|
|||
| samcv | oh jnthn interestingly those linker scripts, i think the ruby implementation of libffi actually wrote some ruby code to parse those. so it seems this problem is not unique | 09:42 | |
| would have been nice if *cough* the library loading framework did this library loading functionality | |||
| but alas | |||
| looks like gentoo and freebsd at least use linker scripts | 09:43 | ||
|
10:05
AlexDaniel joined
|
|||
| jnthn | samcv: Oh, interesting | 10:41 | |
| That at least means somebody else was crazy enough to try and make stuff Just Work :) | |||
| samcv | :) | 10:43 | |
| see this issue here: github.com/ffi/ffi/issues/308 | 10:44 | ||
| links to part of their codebase where they parse it in case it fails | |||
| in the 2nd post | |||
| jnthn | Hm, not much code to try and do that | 10:47 | |
| samcv | but yeah essentialy we are mostly looking for 'GROUP ( /usr/lib/real.so )' | 10:51 | |
| and that's it | |||
| so just have to extract that part from the linker script | 10:52 | ||
|
11:03
lizmat joined
12:05
geekosaur joined
13:10
zakharyas joined
|
|||
| Geth | MoarVM/even-moar-jit: c24d1cfbee | (Bart Wiegmans)++ | 2 files Remove point spilling for calls Point spilling was supposed to be an optimization for rare function call paths, but it interacted really poorly with full-spilling and judging from the generated code was just as often a pessimization. So I've decided to remove it, simplifying the ARGLIST preparing code, but also unocvering a bug in handling spilled values for ARGLIST, which I've also fixed. I intend to optimize spilling separately, making sure a value is not redundantly 'redefined' if the earlier definition is still in the same basic block. |
16:03 | |
|
16:43
domidumont joined
17:08
geekosaur joined
17:44
zakharyas joined
21:43
AlexDaniel joined
|
|||