#
✕ nick:marcan
#asahi · 2021-12-08 18:00
marcanI think that's a GPU thing
#asahi · 2021-12-08 17:37
marcanthere's little point
#asahi · 2021-12-08 17:37
marcanm1n1 will probably never support USB3.0
#asahi · 2021-12-08 16:42
marcanconsidering USB2 has 8 microframes per millisecond, that looks like m1n1 proxyclient is already pretty much as fast as possible
#asahi · 2021-12-08 16:42
marcanalso this makes me realize I'm pushing ~4 write32s per millisecond, blocking
#asahi · 2021-12-08 16:27
marcanalso added register sub-field support to the vcd dumper part ( https://twitter.com/marcan42/status/1468617255147941888/photo/1 )
#asahi · 2021-12-08 16:26
marcanpushed the gpiola/spi experiment stuff
#asahi · 2021-12-07 03:36
marcanjmr2: pushed the proper fix to the asahi branch for now
#asahi · 2021-12-07 03:21
marcanjmr2: looks like DCP needs a special min-ps PMGR setting to disable power gating; it seems only DCP uses this. I'll submit something for this later, for now I'll send the pulls without it since I don't want to block the core driver (and DCP isn't anywhere near upstreaming yet)
#asahi · 2021-12-06 18:37
marcananyway, good night for now
#asahi · 2021-12-06 18:37
marcanbut yeah, if you have a branch I can mess around with that'd be great :)
#asahi · 2021-12-06 18:37
marcanI see macos enabling auto-pm on that one... :/
#asahi · 2021-12-06 18:36
marcanps_rmx was definitely the root cause, leaving it with everything below it commented out didn't fix it?
#asahi · 2021-12-06 18:35
marcanwell that might throw a wrench into my idea of submitting PMGR tomorrow... though I can always send a fix later. thanks for finding that out!
#asahi · 2021-12-06 18:31
marcanoff to sleep
#asahi · 2021-12-06 18:20
marcan(but really I doubt anyone cares enough to try to apply it themselves for this kind of thing... it should be obvious)
#asahi · 2021-12-06 18:19
marcanyes, just base it on that branch. I'll be the one applying it, so just note in the comments section (outside the commit description) that this is based on that branch, for whoever else wants to apply it.
#asahi · 2021-12-06 18:16
marcankettenis: I mean, since I'm going to send those pulls tomorrow, if you do it now based on that branch I can just include your patch too :)
#asahi · 2021-12-06 18:16
marcanremove that part too
#asahi · 2021-12-06 18:16
marcan/* Turn on auto-PM if the domain is already on */
#asahi · 2021-12-06 18:15
marcanthere's also a thing that sets the auto bit in the init function
#asahi · 2021-12-06 18:12
marcanassuming the patch is fine that can go in with the pull too then, we don't need to wait on anyone for that
#asahi · 2021-12-06 18:10
marcankettenis: I just pushed asahi-soc/dt; it's WIP and not what I will submit tomorrow, but it has those patches which should make applying it easy
#asahi · 2021-12-06 18:06
marcanI don't mind fixing conflicts though, it's kind of expected at this point
#asahi · 2021-12-06 18:05
marcankettenis: just do it against upstream (the latest RC or whatever), unless you want to wait for my pull tags tomorrow
#asahi · 2021-12-06 18:05
marcanjmr2: in the code, not the DTS
#asahi · 2021-12-06 18:02
marcanyou mean for the DT? or the wifi driver?
#asahi · 2021-12-06 17:59
marcanmy assumption so far was that it's safe to enable auto-pm for everything; that might be incorrect
#asahi · 2021-12-06 17:59
marcanif those break DCP then that's rather important information
#asahi · 2021-12-06 17:59
marcanjmr2: can you try commenting out the auto-pm related bits?
#asahi · 2021-12-06 17:54
marcan(take that apple, I can roll my own on-chip LA whether you want me to or not :p)
#asahi · 2021-12-06 17:53
marcan(this is doubly funny because I'm pretty sure these chips have on-chip logic analyzer features, via the debug stuff, that would of course be fused off/locked out)
#asahi · 2021-12-06 17:51
marcanm1n1 scripts dumping vcds for gtkwave ;)
#asahi · 2021-12-06 17:50
marcanso yeah, expect that on spi stream day :p
#asahi · 2021-12-06 17:49
marcanjust a little asm busyloop
#asahi · 2021-12-06 17:49
marcanthat wouldn't even require IRQs
#asahi · 2021-12-06 17:49
marcanindeed
#asahi · 2021-12-06 17:49
marcanlol
#asahi · 2021-12-06 17:49
marcanso just a little busyloop or IRQ that dumps out samples
#asahi · 2021-12-06 17:49
marcanI definitely saw that with i2c
#asahi · 2021-12-06 17:49
marcanthe input buffer is always active AIUI
#asahi · 2021-12-06 17:49
marcanyup
#asahi · 2021-12-06 17:48
marcanI bet we can do a software scope.... exactly
#asahi · 2021-12-06 17:48
marcanfor figuring out what the SPI configs do... we probably don't need an actual scope
#asahi · 2021-12-06 17:48
marcanalso I had an idea yesterday
#asahi · 2021-12-06 17:48
marcanawesome, thanks!
#asahi · 2021-12-06 17:47
marcanis this with the PMGR version in the asahi branch?
#asahi · 2021-12-06 17:42
marcan(the disp0/dcp stuff)
#asahi · 2021-12-06 17:42
marcanI think I set those domains to always-on anyway since I don't even know if we can safely power them down
#asahi · 2021-12-06 17:42
marcanI didn't but I don't see why it would make a difference

plus ›