2026-11-14
00:01chaos_princessthe point is that it turned out that if you hook up the gpu's pmp report, it starts actively managing the fabric speed, and with pmp on, you get better gpu memory bandwidth than with it off
00:02Mary has quit [Quit: .]
00:02Mary has joined #asahi-dev
00:10chadmedhmmm i wonder if turning it on for avd will explode anything given we have custom firmware...
00:11chaos_princessavd is extra special in pmp messages
00:12amarioguy has joined #asahi-dev
00:36amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
00:44zaprit has joined #asahi-dev
00:48amarioguy has joined #asahi-dev
00:52zaprit has quit [Ping timeout: 480 seconds]
01:05maettu102 has quit [Quit: brb]
01:06maettu102 has joined #asahi-dev
01:29zaprit has joined #asahi-dev
01:33barusu has quit [Ping timeout: 480 seconds]
01:37zaprit has quit [Ping timeout: 480 seconds]
01:59chadmed has quit [Quit: Konversation terminated!]
02:00DragoonAethis has quit [Quit: hej-hej!]
02:00DragoonAethis has joined #asahi-dev
02:05chadmed has joined #asahi-dev
02:05chadmedtheres barely any difference in star wars battlefront between pmp and no pmp, both sit in the ~160-210 fps range on bespin platforms
02:06chadmedand of course direct scanout but we need to talk to kwin folks about that because it still doesnt work
02:07chadmed has quit []
02:11chadmed has joined #asahi-dev
02:37ipatch has quit [Read error: Connection reset by peer]
02:40amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
02:54linkmauve has quit [Remote host closed the connection]
03:00yurix has quit [Remote host closed the connection]
03:00yurix has joined #asahi-dev
03:01linkmauve has joined #asahi-dev
03:10chodaQ[m] has joined #asahi-dev
03:16MichaelReeves[m]iirc /usr/share/firmware/wifi/?
03:17amarioguy has joined #asahi-dev
03:24bit_runner80 has joined #asahi-dev
03:29al3xtjames2 has quit [Read error: Connection reset by peer]
03:29al3xtjames2 has joined #asahi-dev
03:35al3xtjames2 has quit [Read error: Connection reset by peer]
03:35al3xtjames2 has joined #asahi-dev
03:56tobhe_ has joined #asahi-dev
04:03tobhe has quit [Ping timeout: 480 seconds]
04:10amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
04:11nora has joined #asahi-dev
04:12bit_runner80 has quit [Remote host closed the connection]
04:12nora_ has quit [Ping timeout: 480 seconds]
04:44salva has joined #asahi-dev
04:45MichaelReeves[m]what makes you think that?
05:14nicolas17the contents of /usr/share/firmware/bluetooth :P and apparently someone else made the same discovery from the devicetree days ago
05:24myria has quit [Ping timeout: 480 seconds]
05:34jcbm has quit [Remote host closed the connection]
05:35jcbm has joined #asahi-dev
05:50myria has joined #asahi-dev
06:12chadmed has quit [Quit: Konversation terminated!]
06:12chadmed has joined #asahi-dev
07:22N_Tempus_04 has joined #asahi-dev
07:58MichaelReeves[m]oh cool. maybe after bcm upstreaming is done I'll have to get my hands on a neo and try writing (and hopefully upstreaming) the mediatek driver, sounds fun but probably too expensive for me right now.
07:58MichaelReeves[m]they're $1000 AUD right?
07:59MichaelReeves[m]anyway, working on continuing the installer refactor now (unless chadmed you'd like to keep doing it and prefer I leave it for you?)
08:00weirdtreething has quit [Remote host closed the connection]
08:00weirdtreething has joined #asahi-dev
08:03MichaelReeves[m]yes I saw you have a core-refactor branch. I've got some free time now so that's why I was wondering, but there's no rush
08:03myria has quit [Remote host closed the connection]
08:04chadmedi think my local branch is ahead of what ive pushed. ill push whatever's sitting locally after dinner
08:04barusu has joined #asahi-dev
08:04myria has joined #asahi-dev
08:16Bhyve has joined #asahi-dev
09:16JaminMartin[m] has joined #asahi-dev
09:35Bhyve has quit []
09:37Bhyve has joined #asahi-dev
09:40chadmedif you need a hand just shout, im just really really unwell at the moment and trying to close out 5 projects at work so dont really have a lot of time or energy atm :P
09:43MichaelReeves[m]thank you! hope you feel better soon :)
09:52yurix_ has joined #asahi-dev
09:52yurix has quit [Read error: Connection reset by peer]
09:52yurix_ is now known as yurix
10:11zaprit has joined #asahi-dev
10:19zaprit has quit [Ping timeout: 480 seconds]
10:20linuxgemini is now known as Guest882
10:20linuxgemini has joined #asahi-dev
10:23Guest882 has quit [Ping timeout: 480 seconds]
10:30N_Tempus_04 has quit [Ping timeout: 480 seconds]
10:38chaos_princessthe new touchscreen macs are a 'k', not a 'j', and the touchscreen is on mtp, called 'kiwi', firmware upload is most likely similar to trackpad
10:56s0ullight[m]chadmed: yeah it's gonna depend on where you're bottlenecked
10:57JamesCalligeros[m]Its a 22 year old d3d9 game :p
10:57chaos_princessit could also be a pro/max exclusive thing, as they have a more complex memory system
10:57s0ullight[m]But gpu write bandwidth was affected by hooking pmp up to the gpu on my machine
10:57JamesCalligeros[m]The mac studio gets about double the performance which is what i expect
10:58JamesCalligeros[m]This morning’s test was on j415
10:58s0ullight[m]My machine == mbp m1 pro
11:03s0ullight[m]<JamesCalligeros[m]> "Its a 22 year old d3d9 game :p" <- virtio implicit sync? Try with HK_PERFTEST=batch
11:04s0ullight[m]And yeah p3s might be interesting as well
11:05JamesCalligeros[m]I may look into it further tomorrow but im not too fussed
11:06JamesCalligeros[m]I dont think the performance is pathological or anything, it was just interesting that pmp had little effect
11:06N_Tempus_04 has joined #asahi-dev
11:17___nick___ has joined #asahi-dev
11:18zaprit has joined #asahi-dev
11:51xdbob has quit [Remote host closed the connection]
11:52xdbob has joined #asahi-dev
12:01linuxgemini is now known as Guest885
12:01linuxgemini has joined #asahi-dev
12:05Guest885 has quit [Ping timeout: 480 seconds]
12:09Bhyve has quit []
12:25zaprit has quit [Quit: leaving]
12:30zaprit has joined #asahi-dev
12:32elisaadohm reading https://github.com/corellium/linux-m1/blob/master/drivers/hwmon/apple-m1-smc.c I think I might have better luck doing it with ebpf :3
12:32MichaelReeves[m]T2 peeps have a patchset to enable that, while we're not going to put be putting it in default builds, you can add it if thats what u want https://lore.kernel.org/asahi/20261104195317.8284-1-atharvatiwarilinuxdev@gmail.com/T/#m4ad842d09cd741541cdea52dd84c3bc1633bb2cc
12:35elisaadodoes that require a lot of setup? My end goal is to be able to still be booted in my install where I can change the state of my system and then observe changed in the read SMC values
12:36N_Tempus_04 has quit [Ping timeout: 480 seconds]
12:36MichaelReeves[m]yes it is but from your described use case its not ideal for that
12:37MichaelReeves[m]you could make it be a hypervisor while linux is in EL1 but thats overkill when its just a patch to get it in sysfs, IMO
12:38elisaadoyeah I think I'll go that route, or create a small kernel module that uses the apple_smc_get_key* functions
13:27zaprit has quit [Remote host closed the connection]
13:30Bhyve has joined #asahi-dev
13:31N_Tempus_04 has joined #asahi-dev
13:32al3xtjames2 has quit [Read error: Connection reset by peer]
13:32al3xtjames2 has joined #asahi-dev
13:34zaprit has joined #asahi-dev
13:40jannaufinally working experiments/dcp.py on t6030, that doesn't have issues displaying XRGB (with X/alpha set to 0) though
13:45zaprit has quit [Remote host closed the connection]
13:51AnEagle[m]Enabling PMP report on avd in the devicetree doesn't crash anything, does it just do nothing because of the custom firmware ?
13:51chaos_princessyou also need to do some special memory pokes for avd
14:06al3xtjames25 has joined #asahi-dev
14:06al3xtjames2 has quit [Read error: Connection reset by peer]
14:06al3xtjames25 is now known as al3xtjames2
14:18sofusi dont think the cm3 can do power managment, i have a hunch that it would not require any driver/fw changes
14:34N_Tempus_04 has quit [Quit: My Mac went to sleep]
15:27hardy94 has joined #asahi-dev
15:29jannausigh, XRGB works when the framebuffer covers the whole display, i.e. without notch avoidance
15:37Bhyve has quit []
15:49linuxgemini is now known as Guest897
15:49linuxgemini has joined #asahi-dev
15:53Guest897 has quit [Ping timeout: 480 seconds]
16:23alessaba has quit [Ping timeout: 480 seconds]
16:28WindowPain8 has joined #asahi-dev
16:33linuxgemini is now known as Guest902
16:33linuxgemini has joined #asahi-dev
16:33WindowPain has quit [Ping timeout: 480 seconds]
16:33WindowPain8 is now known as WindowPain
16:37Guest902 has quit [Ping timeout: 480 seconds]
16:44Bhyve has joined #asahi-dev
16:48balrog has quit [Ping timeout: 480 seconds]
16:50hardy1994 has joined #asahi-dev
16:50hardy94 has quit [Read error: Connection reset by peer]
16:50hardy1994 is now known as hardy94
16:52N_Tempus_04 has joined #asahi-dev
16:54balrog has joined #asahi-dev
16:57salva has quit [Read error: Connection reset by peer]
17:03rivendell has quit [Ping timeout: 480 seconds]
17:04rivendell has joined #asahi-dev
17:08amarioguy has joined #asahi-dev
17:12balrog has quit [Ping timeout: 480 seconds]
17:15balrog has joined #asahi-dev
17:27amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
17:48N_Tempus_04 has quit [Ping timeout: 480 seconds]
18:01Bhyve has quit []
18:07yeezuz2030[m] has joined #asahi-dev
18:20balrog has quit [Ping timeout: 480 seconds]
18:21___nick___ has quit [Ping timeout: 480 seconds]
18:31alessaba has joined #asahi-dev
18:31alessaba_ has joined #asahi-dev
18:33amarioguy has joined #asahi-dev
18:44jannauthis is strange. it is still somewhat pre-multiplied alpha compositing but somewhat weird
18:47jannaualpha values smaller than 0x28 result in a less bright green and it vanished with 0x00 completly
19:00bgb has quit [Remote host closed the connection]
19:00bgb has joined #asahi-dev
19:11balrog has joined #asahi-dev
19:38chaos_princesselisaado: if you want something that is even easier to access from userspace - here https://github.com/WhatAmISupposedToPutHere/linux/commit/0eb7513c53831bfbd9331860ecfff2d2f8d3da24
19:38chaos_princessbut you will need to port that to modern kernel api most likely
19:54___nick___ has joined #asahi-dev
19:55___nick___ has quit []
19:57___nick___ has joined #asahi-dev
20:06amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
20:19amarioguy has joined #asahi-dev
20:22ipatch has joined #asahi-dev
20:32Lyude has quit [Quit: Bouncer restarting]
20:33Lyude has joined #asahi-dev
20:34alessaba_ has quit [Remote host closed the connection]
20:34alessaba has quit [Remote host closed the connection]
20:36alessaba has joined #asahi-dev
20:40__nick__ has joined #asahi-dev
20:40alessaba_ has joined #asahi-dev
20:43___nick___ has quit [Ping timeout: 480 seconds]
20:49alessaba has quit [Remote host closed the connection]
20:49alessaba_ has quit [Remote host closed the connection]
20:51alessaba has joined #asahi-dev
20:53amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
20:55alessaba_ has joined #asahi-dev
20:56JamesCalligeros[m]jannau: are you trying to figure out if we can do coverage?
20:57JamesCalligeros[m]I have a fix for the busted premultiply blending locally that seems to be about as good as it’ll get
20:58cow123 has quit [Remote host closed the connection]
21:01jannauI have a few fixes for DCP 14.8.3. the padding you added for swap_submit_dcp is an additional surfaces2. still no idea what those do
21:01JamesCalligeros[m]Lol they cant even make it work properly with one set of surfaces let alone 3
21:03amarioguy has joined #asahi-dev
21:04jannauis fixing premult alpha blending more than `.is_premultiplied |= format->has_alpha`? but the strange pre-multiplied alpha compositing results might throw a wrench through that too
21:04amarioguy has quit []
21:05JamesCalligeros[m]No, i use the actual blending mode of the plane now that the property is wired up and that works better
21:05amarioguy has joined #asahi-dev
21:05JamesCalligeros[m]I didnt notice any inaccurate colours on m2 or m1
21:05__nick__ has quit [Remote host closed the connection]
21:09JamesCalligeros[m]It is the default but clients can specify that theyre not doing it
21:10JamesCalligeros[m]So we cant assume that every alpha surface will always be premultiplied
21:17jannauI don't think clients can at the moment. we would need to add the blend property with `BIT(DRM_MODE_BLEND_PREMULTI) | BIT(DRM_MODE_BLEND_PIXEL_NONE)`
21:18JamesCalligeros[m]Yeah, i did
21:18JamesCalligeros[m]Let me just push what i did
21:25jannauthe checks for XRGB8888 and XBGR8888 need to stay at least for Pro/Max (it looked like DCP used coverage for ARGB/BGRA on M1 or M2)
21:29cow123 has joined #asahi-dev
21:51jannauDCP firmware reports now key='IOMFBNumLayers': {'MaxNumLayers': 3, 'FullFrameRequired': 0}
21:52N_Tempus_04 has joined #asahi-dev
21:53JamesCalligeros[m]Lol thanks apple
21:53JamesCalligeros[m]Does that trigger a property callback before we’ve set up our planes? Maybe we can use that instead of our crazy heuristic thing
21:56yurix has quit [Remote host closed the connection]
21:58___nick___ has joined #asahi-dev
22:01JamesCalligeros[m]Mmm if we were willing to drop 13.x id say we should use it but we’re not going to do that :p
22:02JamesCalligeros[m]So we may as well just keep the approach we’re doing now
22:03yurix has joined #asahi-dev
22:19N_Tempus_04 has quit [Ping timeout: 480 seconds]
22:24nst has quit [Ping timeout: 480 seconds]
22:33nst has joined #asahi-dev
23:03amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
23:48halloy2736 has joined #asahi-dev