#asahi-dev · 2025-12-26 14:55
oliverLet's see how fast we are if need to transfer 30mb of frame buffer over serial each swap ;p#asahi-dev · 2025-12-26 14:21
chadmedwhich is why i suggested forcing software decode. we at least have _some_ documentation for the agx formats#asahi-dev · 2025-12-26 14:21
chadmeddownside is the framebuffer formats are all totally fucked and youll have to reverse engineer them :P #asahi-dev · 2025-12-26 14:20
chadmedso then you know which swap request, which iosurface, and the fb that corresponded to both #asahi-dev · 2025-12-26 14:20
chadmedyou could read out the framebuffer and store it with the corresponding surface and swap request and then decode that all later#asahi-dev · 2025-12-26 14:18
oliverThanks. Ive disabled most printing actually, I am using the parse script later. But I want to record the screen and some timestamp in the future so can match an event to what's actually happening.#asahi-dev · 2025-12-26 14:12
chadmedjust ucomment the block i commented out if you still want/need to see the other messages#asahi-dev · 2025-12-26 14:12
chadmeddo something like this to make life a little easier#asahi-dev · 2025-12-26 14:12
chadmedalso if you havent already: https://github.com/chadmed/m1n1/commit/b000ada977896a265e5ee972aaace1a7297b4adb#asahi-dev · 2025-12-26 14:11
chadmeddcp supports both HLG and PQ, which both will use the BT.2020 colour primaries#asahi-dev · 2025-12-26 14:11
chadmedlikewise the transfer functions are going to matter a lot for correct hdr#asahi-dev · 2025-12-26 14:10
chadmedcolourspace is going to be important for _correct_ looking video on overlay surfaces#asahi-dev · 2025-12-26 14:10
oliverBut that's just experimenting, I need to check what gnome is actually doing, I guess #asahi-dev · 2025-12-26 14:09
oliverI set xfer to hdr and keep colorspace in native. That at least gives me a good looking result. #asahi-dev · 2025-12-26 14:08
oliverThanks#asahi-dev · 2025-12-26 14:08
chadmedyeah break into the hv shell and try to read out the framebuffer at the iova in the swap request i guess#asahi-dev · 2025-12-26 14:07
oliverOkay, then I'll try to verify that by just looking at the data that Is scanned out. I should be able to read the pixel data from m1n1 right? #asahi-dev · 2025-12-26 14:07
chadmedare you setting the colourspace and framebuffer format appropriately too#asahi-dev · 2025-12-26 14:06
chadmedits probably just software tone mapping and local dimming then i guess#asahi-dev · 2025-12-26 14:05
oliverIf I force gnome into hdr rendering and tell dcp to scan out with xfer hdr, it actually looks quite good already. If I then force the brightness as high as mac does, it looks pretty good already. #asahi-dev · 2025-12-26 14:05
oliveryes exactly. And the primary surface format changes to 16bit float #asahi-dev · 2025-12-26 14:04
chadmeddo the video and media controls go back on to the primary surface? thats not how i interpreted your message sorry#asahi-dev · 2025-12-26 14:03
oliverUnless the media controls are overlaying the video #asahi-dev · 2025-12-26 14:02
chadmedquicktime also ends up with windowed content on its own surface#asahi-dev · 2025-12-26 14:02
oliver I am going for non fullscreen hdr content right now. You're right with full screen content, that's probably going onto its own surface. #asahi-dev · 2025-12-26 13:59
chadmedtry using ffmpeg to force sw decode and bypass all the compressed/tiled avd format bs #asahi-dev · 2025-12-26 13:57
chadmedif a video is playing and nothing else on the screen is changing, then yes it will only scan out the iosurface with the video on it. thats the whole point, we dont need to be constantly scanning out and refreshing surfaces where nothing is going on #asahi-dev · 2025-12-26 13:56
chadmedim not 100% sure, just a guess from what ive seen playing back hdr video on a machine being traced#asahi-dev · 2025-12-26 13:52
oliverYou're sure? I saw macos switching from a 10bit rgb format to 16bit half floats, but it is just scanning out that one surface while the hdr media is playing (well, as along as something is overlaying something on top of the hdr content at least). And interestingly it is also scanning that one out as sdr with the same colorspace. #asahi-dev · 2025-12-26 13:19
chadmedkwin really does not want to play ball with the overlays though so ive mostly been testing from a tty, which is not really a fair test since most of the primary surface is black lol#asahi-dev · 2025-12-26 13:18
chadmedive been playing with the per-surface colour space/transfer function stuff in the dcp driver and it seems to also mostly Just Work on linux#asahi-dev · 2025-12-26 13:17
chadmedthe sdr content staying the same relative brightness is probably a mix of that, local dimming, and perhaps some fancy tone mapping/compensation on the sdr surface itself#asahi-dev · 2025-12-26 13:16
chadmedthe hdr content will be displayed on a different iosurface with its own colour primaries, xfer function, etc. #asahi-dev · 2025-12-26 13:04
oliverLooking for educated guesses: when you display some hdr content on a Macbook, it takes half a second for the hdr content to reach full brightness. During tracing I see, that the display brightness is slowly increased above the normal sdr maximum during this time. Question is now: how does the sdr content stay at the same brightness during that time? Is macos just rendering it darker in software or might it just adjust some gamma curve or #asahi-dev · 2025-12-26 13:04
oliveranything? #asahi-dev · 2025-12-26 12:30
chadmedi will test j416 tomorrow. still kind of crazy that page flips take so long...#asahi-dev · 2025-12-26 12:29
chadmedseems a little smoother but by and large no changes. 120hz is still dodgy on my j314.#asahi-dev · 2025-12-26 12:28
chadmedall atomic_flush does now is get a reference to the old state framebuffer, 1:1 put the plane state into the swap request then send it off#asahi-dev · 2025-12-26 12:27
chadmeda couple of experimental things are turned on and it needs cleaning up/authorship of commits fixed up but seems mostly sane and doesnt explode#asahi-dev · 2025-12-26 12:27
chadmedhttps://github.com/chadmed/linux/tree/dcp/scanout works on all laptops and tested with 144hz monitor#asahi-dev · 2025-12-26 10:57
chadmednot a big deal, ive just done a loooot of rebooting today and its a bit annoying #asahi-dev · 2025-12-26 10:56
chadmedyeah not sure whats up then. all my t602x machines take 20-30 seconds to decompress kernel/initramfs and u-boots idea of "1 second" does not match reality :P #asahi-dev · 2025-12-26 10:48
jannaum1n1 intends to initializes the t602x e-cores to the same 2GHz as all other e-cores. I haven't noticed a difference in m1n1 pre-kboot on t6020 or t6022 either#asahi-dev · 2025-12-26 10:22
chadmedpre-linux environment is very slow and always has been#asahi-dev · 2025-12-26 10:22
chadmedare we spinning up the t602x ecores in a different pstate to every other soc (in m1n1) #asahi-dev · 2025-12-26 10:05
integralpilotthough the HDD is SMART "pre-fail" so i probably need to pick-up a new one#asahi-dev · 2025-12-26 10:05
integralpiloti've just gotten out an old x86 linux tower i had, and it's working now#asahi-dev · 2025-12-26 10:03
chaos_princessimo, just don't bother trying to compile linux from macos and use a linux vm, it works, but not worth the extra hassle#asahi-dev · 2025-12-26 10:01
integralpilotoh and also enjoy 39c3 for everyone that's going!! good luck with the demo :)#asahi-dev · 2025-12-26 09:55
integralpiloti've updated the gist with all this info now