#asahi · 2021-01-16 17:02
davidrysk[m]noc0lour's responses appear adequate to me.#asahi · 2021-01-16 17:01
davidrysk[m]asmon: this is an english language channel#asahi · 2021-01-16 16:56
davidrysk[m]I didn't have a chance to test this time#asahi · 2021-01-16 16:56
davidrysk[m]youtube's automatic resolution doesn't always work as well as it could#asahi · 2021-01-16 16:56
davidrysk[m]the other day it was, after I set youtube to 1080p#asahi · 2021-01-16 16:54
davidrysk[m]even when you set youtube to 1080p?#asahi · 2021-01-16 16:20
davidrysk[m]when upstreaming is the plan it's not "copyright concern troll[ing]"#asahi · 2021-01-16 16:20
davidrysk[m]yeah, and subtweets#asahi · 2021-01-16 16:15
davidrysk[m]but if you want me to ask whitequark to import my logs (and marcan is okay with that) I can...#asahi · 2021-01-16 16:14
davidrysk[m]there was not really any useful discussion that early#asahi · 2021-01-16 16:00
davidrysk[m]asmon: more specifics?#asahi · 2021-01-16 16:00
davidrysk[m]sven: basically if we can't tell that code was developed cleanly, it could become a risk to the project if it is used or in some cases even if knowledge from it is used (e.g. if the development of said code involved illegally obtained leaked code -- note that I am not saying that pongoOS involved that, I am stating the hypothetical)#asahi · 2021-01-16 15:57
davidrysk[m]sven: I don't mean upstreaming pongoos. I mean the lineage of pongoos code, to ensure that it was cleanly RE'd and whatnot#asahi · 2021-01-16 15:56
davidrysk[m]But the standards for upstreaming code are higher.#asahi · 2021-01-16 15:56
davidrysk[m]as for RHEL tarballs: those are downstream and not upstream distributions. And afaik Red Hat / CentOS do include a file that lists which files were changed and whose changes went into the files, which means they comply with GPLv2 2.a#asahi · 2021-01-16 15:54
davidrysk[m]I did a `git log` of pongoOS and there's all this about syncing with an internal repo. There is zero ability to have an audit trail with that#asahi · 2021-01-16 15:54
davidrysk[m]this licensing stuff (and seemingly fundamental misunderstanding, or intentional ignorance, of how upstreaming stuff works) is frustrating#asahi · 2021-01-16 05:32
davidrysk[m]pongoOS has license issues, and corellium's linux-sandcastle couldn't be upstreamed as-is#asahi · 2021-01-15 22:56
davidrysk[m]implr: mcsema is from trailofbits#asahi · 2021-01-15 19:26
davidrysk[m]yeah, which makes things kinda annoying#asahi · 2021-01-15 19:23
davidrysk[m]I wouldn't necessarily recommend that design in 2020, but it made sense at the time. :)#asahi · 2021-01-15 19:23
davidrysk[m]If you're interested in how MAME's intermediate language works, you can find documentation here. https://wiki.mamedev.org/index.php/UML_Architecture#asahi · 2021-01-15 19:20
davidrysk[m]of course, now with ARM/aarch64 that will probably end up being an advantage#asahi · 2021-01-15 19:20
davidrysk[m]the intermediate language was designed when both PPC and x86 were common and it wasn't clear that Apple would dump PPC / PPC would become less relevant#asahi · 2021-01-15 19:19
davidrysk[m]yeah. MAME's use of asmjit is only for the macro assembler library -- it does its own register allocation (which is a bit less than ideal because the MAME intermediate language has more registers than typical Intel machines)#asahi · 2021-01-15 19:15
davidrysk[m]it only has x86 (32/64) and aarch64 backends at this time#asahi · 2021-01-15 19:14
davidrysk[m]one of its features is just a macro assembler library, but it also has higher level features including register allocation#asahi · 2021-01-15 19:12
davidrysk[m]It's not a dynamic recompiler as much as a library to make it easier to write a dynamic recompiler... MAME takes some advantage of it#asahi · 2021-01-15 19:12
davidrysk[m]awordnot: are you familiar with asmjit? A bunch of software takes advantage of it to simplify the complexity of the internal backend#asahi · 2021-01-15 19:07
davidrysk[m]awordnot: hmmm :)#asahi · 2021-01-15 17:36
davidrysk[m]I'm not aware of very many FOSS implementations of AOT recompilation. The RPCS3 emulator uses it for Cell SPU and PPU code; mcsema can do it but it's more geared towards reverse-engineering and not optimized for performance; there's also https://github.com/MerryMage/dynarmic but it's for the other way around (ARM on x86)#asahi · 2021-01-15 17:33
davidrysk[m]The TSO bit helps, but ahead-of-time helps more.#asahi · 2021-01-15 17:33
davidrysk[m]clover: the major thing that makes Rosetta work as well as it does is ahead-of-time recompilation.#asahi · 2021-01-15 16:58
davidrysk[m]Yep, it's not :)#asahi · 2021-01-15 16:57
davidrysk[m]I don't think there's anything FOSS that can do this efficiently#asahi · 2021-01-15 16:57
davidrysk[m]yeah, though rosetta 2 uses ahead-of-time static recompilation and only falls back to interpreting / dynarec when needed#asahi · 2021-01-15 05:43
davidrysk[m](And the host mode driver might be fine for the otg core in host mode. Haven’t dug deep enough)#asahi · 2021-01-15 05:42
davidrysk[m]marcan: I don’t believe upstream Linux has support for the dwc otg core. It does have support for the dwc host mode cores#asahi · 2021-01-15 05:28
davidrysk[m]though USBC is more useful than FL1100, but the latter might be simpler#asahi · 2021-01-15 05:28
davidrysk[m]still, gonna need USB sooner rather than later#asahi · 2021-01-15 05:27
davidrysk[m]1.5M wouldn't be TOO unbearable#asahi · 2021-01-15 05:27
davidrysk[m]marcan: do we have any idea of how complex accessing the internal NAND will be?#asahi · 2021-01-15 05:09
davidrysk[m]re apfs: apparently apfs has additional fields to make it easy to locate the boot object without having to parse/mount the standard filesystem structures. I wish we had the Apple Silicon version of the APFS documentation; Apple drags their feet on releasing it#asahi · 2021-01-15 05:08
davidrysk[m]Necrosporus: marcan spent a lot of time working on device tree handling#asahi · 2021-01-15 05:01
davidrysk[m]Necrosporus: what do you mean by "work"?#asahi · 2021-01-15 05:01
davidrysk[m]if it's the dwc_otg core, well, raspberry pi and android use a core from synopsys that has some weird license text but appears permissive#asahi · 2021-01-15 04:55
davidrysk[m]if this is the dwc3 core, there are stacks in Linux, FreeBSD, and probably elsewhere.#asahi · 2021-01-15 04:54
davidrysk[m]unfortunately I am not surprised by this#asahi · 2021-01-14 16:35
davidrysk[m]if I get no sleep I'm completely unproductive the next day#asahi · 2021-01-14 16:34
davidrysk[m]winocm: I feel you :(