#asahi-dev · 2021-02-11 18:32
modwizcodeyeahhhh I guess that would make sense#asahi-dev · 2021-02-11 18:32
modwizcodealthough for all I know it might#asahi-dev · 2021-02-11 18:31
modwizcodeI meant fake as in it probably isn't actually ever sending real PCIe transactions#asahi-dev · 2021-02-11 18:30
modwizcodeA lot of platform devices do seem exposed as fake PCI devices. Which if it's enough faked to respond properly to the PCI driver stack then whatever it probably saves code.#asahi-dev · 2021-02-11 18:29
modwizcodeSo it's his fault :p#asahi-dev · 2021-02-11 18:28
modwizcodeThey probably did that the same reason I wrote hacky bindings in qemu. It worked#asahi-dev · 2021-02-11 18:26
modwizcodeThat's going to be a seperate patch though right?#asahi-dev · 2021-02-11 18:21
modwizcodeif you read virt it shouldn't need to do that either way. But you have to intentionally be needing virt#asahi-dev · 2021-02-11 18:21
modwizcodeBut like from what I can tell if you read phys you either get a trap (without extension) or virt (maybe?) (with extension?)#asahi-dev · 2021-02-11 18:20
modwizcodeRight#asahi-dev · 2021-02-11 18:20
modwizcodeIt seems like they can work that way if you have a specific enhanced virtualization flag (which I expect everything has), but it doesn't seem that's how they're intended to be used? Or rather they don't redirect.#asahi-dev · 2021-02-11 18:19
modwizcodeSo on ARM there's the VIRT timers, which at first I thought were like "if (virtualized) { return VIRT_TIMER_REG; } else { return PHYS_TIMER_REG; }" but that's not at all how that seems to work?#asahi-dev · 2021-02-08 04:27
modwizcodeThat makefile for xnu is insane though, they seem to specify everything via linker arguments instead of using a linker script for anything#asahi-dev · 2021-02-08 04:26
modwizcodeYeah I'm not suer if it honestly matters only marcan would know. As far as I know once iBoot kicks execution over to you, you're executing from a physical address and the virtual load address is nothing. I'm assuming that iBoot would relocate things using the virtual base address if you had any mach-o relocations, but that's now how M1N1 works (for good reason)#asahi-dev · 2021-02-08 04:17
modwizcodeHeh I didn't see that when I first looked at this file awhile back#asahi-dev · 2021-02-08 04:08
modwizcodeSpecifically line 342#asahi-dev · 2021-02-08 04:08
modwizcodehttps://github.com/apple/darwin-xnu/blob/main/makedefs/MakeInc.def#asahi-dev · 2021-02-08 03:53
modwizcodeI googled it and it turns up one result as an XNU thing#asahi-dev · 2021-02-05 16:43
modwizcodeAnyway go to sleep and maybe when you wake up I'll have answers :)#asahi-dev · 2021-02-05 16:43
modwizcode*faster than Rx#asahi-dev · 2021-02-05 16:43
modwizcodeTX is "slow" but I think has literally always worked slower than RX#asahi-dev · 2021-02-05 16:42
modwizcodeQemu doesn't implement anything for TX but that's not an issue#asahi-dev · 2021-02-05 16:42
modwizcodeNeed to implement timing traces that don't write one line per char to qemu's trace output which involves taking locks at least sometimes#asahi-dev · 2021-02-05 16:42
modwizcodeBut I haven't done the tracing yet :)#asahi-dev · 2021-02-05 16:42
modwizcodeIt should mostly be doing full fifo writes if my last effort on the fifo code was anything to go by#asahi-dev · 2021-02-05 16:41
modwizcodeI hope not#asahi-dev · 2021-02-05 16:41
modwizcodegood luck#asahi-dev · 2021-02-05 16:41
modwizcodehah#asahi-dev · 2021-02-05 16:41
modwizcodeI'm also going to have to figure out why when loading linux I time transfers to qemu at being like 0.3MB/s when testing in shell I can copy at like 0.5MB/s and both of those are slow#asahi-dev · 2021-02-05 16:40
modwizcodeMy current config is a lot bigger than i expect I'm kind of wondering where it's picking up the bloat#asahi-dev · 2021-02-05 16:39
modwizcodeI wasn't using your config because I lost it when I installed natively to do further work outside a VM#asahi-dev · 2021-02-05 16:39
modwizcodeLOL#asahi-dev · 2021-02-05 16:30
modwizcodeI'm surprised by now there's not like a late binding syscall number assignment.#asahi-dev · 2021-02-05 16:24
modwizcodeI think there's an extension to ignore that issue#asahi-dev · 2021-02-05 16:24
modwizcodeYeah that's weird, maybe your CFLAGS are weird?#asahi-dev · 2021-02-05 16:23
modwizcodeYeah I was trying to pull it up for you again but I couldn't find it when I tried to replicate it locally, I was going to clean and rebuild at some point but it seems you didn't need it#asahi-dev · 2021-02-05 16:14
modwizcodeYeah that's basically what I figured. Correct me if I'm wrong but I think that x86 in contrast since x32 isn't a totally different instruction set and mode, can easily benefit in multiple ways, whereas for arm64 you already need a new build and x32 is just another build to maintain with hard to explain tradeoffs for users and limitations for developers.#asahi-dev · 2021-02-05 14:08
modwizcodeI'll get you a printout. j`ey linked a mailing list reply from the build test bot that had the same error tho#asahi-dev · 2021-02-05 14:06
modwizcodemarcan: can you specify the error in question?#asahi-dev · 2021-02-05 14:06
modwizcodepower is way better with hardware like that#asahi-dev · 2021-02-05 14:05
modwizcodei believe the whole trick of the M1 is their accelerators and their platform integration making using them to do something like decode a videk seemless#asahi-dev · 2021-02-05 05:00
modwizcodeI don't think memory is *that* cheap#asahi-dev · 2021-02-05 05:00
modwizcodeprobably too much effort and not enough understanding#asahi-dev · 2021-02-05 04:59
modwizcodeI'm a little surprised it didn't#asahi-dev · 2021-02-05 04:57
modwizcodeYes ILP32 would indeed save memory#asahi-dev · 2021-02-05 04:57
modwizcodeI misread lol I thought you said ILP64#asahi-dev · 2021-02-05 04:57
modwizcodeI'm not exactly sure how that would save memory#asahi-dev · 2021-02-05 04:56
modwizcodeweird#asahi-dev · 2021-02-05 04:55
modwizcodeI meant to say LP64#asahi-dev · 2021-02-05 04:55
modwizcodeYeah I put an extra L oops