#asahi-dev · 2022-04-06 17:36
marcanDCP is definitely recent#asahi-dev · 2022-04-06 17:35
marcanbut that's to be expected#asahi-dev · 2022-04-06 17:35
marcanI do know the UART driver works all the way back at least, someone already tried that one :')#asahi-dev · 2022-04-06 17:35
marcanand going back from that would get progressively harder#asahi-dev · 2022-04-06 17:34
marcanI mean I expect it to just work on A14 if you have a way to run the code#asahi-dev · 2022-04-06 17:34
marcanyeah, pretty much#asahi-dev · 2022-04-06 17:34
marcanalso AIC, even if similar, would be wired completely differently to the FIQ stuff#asahi-dev · 2022-04-06 17:33
marcanI wonder when they even introduced i2c, I wouldn't be surprised if that pasemi block is recent-ish and they were using samsung before#asahi-dev · 2022-04-06 17:33
marcansome jellybean drivers like UART/SPI might work, maybe PMGR too? definitely not cpufreq, maaybe some of the RTKit stuff with a lot of porting work if it's similar?#asahi-dev · 2022-04-06 16:19
marcanand yes, it will have an update/reinstall mode, not the least because apparently Apple sometimes breaks m1n1 with their updates...#asahi-dev · 2022-04-06 16:19
marcanthe main issue is going to be the firmware copy script and dealing with multi ESP#asahi-dev · 2022-04-05 08:05
marcanonly thing is I think there's still a bug in vUART input, it drops characters if you paste too many at once#asahi-dev · 2022-04-05 08:04
marcan_jannau_: I almost *always* run it under the HV while testing, I would call that extensively tested enough#asahi-dev · 2022-04-05 08:03
marcanand yeah, the HV was made for this (also)#asahi-dev · 2022-04-05 08:03
marcanarnd: I might poke around in a bit, just need to get a chore taken care of#asahi-dev · 2022-04-05 06:00
marcanin PMGR it's actually called something like DISP0_CPU or so IIRC#asahi-dev · 2022-04-05 05:59
marcandcp is just the CPU#asahi-dev · 2022-04-05 05:59
marcanthat's the actual display hardware#asahi-dev · 2022-04-04 20:39
marcansven: remember we both hail from fail0verflow, it's in our genes#asahi-dev · 2022-04-04 19:44
marcanthe IRQ should be cleared at the NVMe side#asahi-dev · 2022-04-04 19:44
marcankettenis: no, that's for fake software IRQs#asahi-dev · 2022-04-04 18:07
marcansven: leftover IRQ flag bits from iBoot?#asahi-dev · 2022-04-04 17:29
marcanit exists in AICv2 too#asahi-dev · 2022-04-04 17:29
marcansven: yes, that is what you want#asahi-dev · 2022-04-03 16:14
marcanbasically all the python stuff is for development and we're pretty loose with what goes in, standalone scripts to achieve something useful for someone are fine, that's kind of the whole point#asahi-dev · 2022-04-03 16:00
marcanwe actually need to make sure that u-boot+linux thing still works / is a bit saner in the near future as we start depending on u-boot to provide PSCI over EFI services#asahi-dev · 2022-04-03 15:59
marcanfor production of course I'd rather everyone use UEFI but it sounds like you're already doing that#asahi-dev · 2022-04-03 15:58
marcanI expect different OSes to have different loader requirements for development#asahi-dev · 2022-04-03 15:58
marcanyeah, for development feel free to put together a fbsd-specific loader script and submit that for upstreaming; I'm happy to carry whatever as long as it works for you#asahi-dev · 2022-04-03 13:04
marcanit's 3 power domains for all of them, it's just that one power domain can be turned off on <4T but I don't think apple does at the OS level and neither should we#asahi-dev · 2022-04-03 11:51
marcanyou can disable it and it'll still run though, if need be, but it slows down a lot and some things won't work in that state#asahi-dev · 2022-04-03 10:35
marcanalso IIRC rcombs said that the h264/hevc blocks work better parallelizing, so you need to have multiple streams going on simultaneously to get the full throughput, but that might be mostly decode?#asahi-dev · 2022-04-03 10:34
marcanbut also encoding HEVC/h264 is, afaik, *much* more compute-intensive than prores (especially HEVC)#asahi-dev · 2022-04-03 10:33
marcanI suspect rcombs might be interested in this conversation#asahi-dev · 2022-03-31 13:36
marcansorry, I'm tired today#asahi-dev · 2022-03-31 13:36
marcanah :)#asahi-dev · 2022-03-31 13:35
marcanI got the whole thing CCed to me, right?#asahi-dev · 2022-03-31 13:34
marcanI saw your mail fly by as I was hitting send :)#asahi-dev · 2022-03-28 17:23
marcanwayland compositors can do the same thing#asahi-dev · 2022-03-28 17:23
marcanAndroid has a whole thing for offloading part of the screen compositing to hardware#asahi-dev · 2022-03-28 17:23
marcannicolas17: that's normal for any modern display controller#asahi-dev · 2022-03-28 09:39
marcanI don't think we've looked much at them. I have some dumps of a compressed framebuffer and it looked quite complicated.#asahi-dev · 2022-03-28 09:39
marcancompressed image formats are designed to reduce memory bandwidth#asahi-dev · 2022-03-28 09:38
marcanrqou_: AGX has several compressed image formats, some are quite complicated#asahi-dev · 2022-03-27 21:20
marcanjannau: repeated chainloads work for me#asahi-dev · 2022-03-27 09:49
marcanyeah, sounds right#asahi-dev · 2022-03-27 09:48
marcan(it'll just uniformly suck more on the TR because Apple's interconnect is better, probably)#asahi-dev · 2022-03-27 09:48
marcanthe same is true for any other system configured like that, e.g. I can do it on my Threadripper (and then Linux sees it as UMA)#asahi-dev · 2022-03-27 09:48
marcanif memory is interleaved then by definition any large copy will average out regardless of where you do it from#asahi-dev · 2022-03-27 09:47
marcanby the way, did m1n1 pick up all the AMCCs? since it has to enable things there