#
…

2022-04-07

00:14al3xtjames has quit [Read error: Connection reset by peer]
00:14al3xtjames has joined #asahi-re
00:49amw has joined #asahi-re
01:20chadmed has quit [Remote host closed the connection]
01:26yuyichao has quit [Ping timeout: 480 seconds]
01:28chadmed has joined #asahi-re
02:34yuyichao has joined #asahi-re
02:56PhilippvK has joined #asahi-re
03:00phiologe has quit [Ping timeout: 480 seconds]
03:26rqou_anybody have any guesses what these prores parameters might control? https://github.com/rqou/m1n1/blob/prores/proxyclient/experiments/prores.py#L621
03:26rqou_in other news, i have an experiment that's able to drive the prores encoder usefully
03:29nicolas17 has quit [Quit: Konversation terminated!]
03:42rqou_in addition to the magic quality parameters, i do still have a concern of "how the heck does the hardware know what size of descriptor it should be reading?"
03:44rqou_i asuume _that_ might be related to some of the magic flags that are also not understood
06:06rqou_does anybody know what IOSurface::getTileHeaderRegionOffsetOfPlane does?
06:06rqou_is this related to AGX compression or is this a different way to make data be tiled?
06:51MajorBiscuit has joined #asahi-re
06:53Major_Biscuit has joined #asahi-re
07:00MajorBiscuit has quit [Ping timeout: 480 seconds]
07:21jaalsa has joined #asahi-re
08:15svenamarioguy: finally found the slightly better sep trace i worked on a while ago. it's probably broken though https://f.svpe.de/d48b3719f56e413e94ca40a6e9cbc8636577695c66054ef44093270998999449_trace_sep.py
08:22the_lanetly_052___ has quit [Remote host closed the connection]
08:23the_lanetly_052___ has joined #asahi-re
10:34chadmed has quit [Read error: Connection reset by peer]
10:57the_lanetly_052___ has quit [Ping timeout: 480 seconds]
11:24chadmed has joined #asahi-re
12:29caef^ has quit [Remote host closed the connection]
13:19bisko has joined #asahi-re
13:42yuyichao has quit [Ping timeout: 480 seconds]
13:53atsalyuk has joined #asahi-re
14:00yuyichao has joined #asahi-re
14:30atsalyuk has quit [Ping timeout: 480 seconds]
14:42ave3 has quit [Remote host closed the connection]
14:42linuxgemini9 has quit [Remote host closed the connection]
14:42linuxgemini9 has joined #asahi-re
14:42ave3 has joined #asahi-re
14:48atsalyuk has joined #asahi-re
14:52doggkruse has joined #asahi-re
14:53linuxgemini95 has joined #asahi-re
14:55ave36 has joined #asahi-re
14:57linuxgemini9 has quit [Ping timeout: 480 seconds]
14:57ave3 has quit [Ping timeout: 480 seconds]
15:01nicolas17 has joined #asahi-re
15:41doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
15:58doggkruse has joined #asahi-re
16:12atsalyuk has quit [Remote host closed the connection]
16:12atsalyuk has joined #asahi-re
16:47sikkileo[m] is now known as sikkiladho[m]
17:35Major_Biscuit has quit [Ping timeout: 480 seconds]
18:09rqou_sven: re DART "uncacheable": i'm going off of the DWARF symbols in the KDK KernelSupport object files which contain a DW_TAG_typedef for "tXXXXdart_l3tte_t"
18:10rqou_which for t6000 and t8110 indeed look the same, with bits 0-3 being valid, uncacheable, wrprot, rdprot
18:10rqou_i'm not sure how i would go about testing whether device accesses are bypassing cache though
18:11rqou_t8020 also appears to have wrprot at bit7 and rdprot at bit8 according to the symbols
18:11svenhrm, so write and read protection also moved around compared to t8103
18:12svendo you have some device for which you can easily trigger a read from a page and time it?
18:12svenif it bypasses the cache that should be slower
18:13rqou_i do know of such devices (jpeg, prores), but i don't know how to get any useful timings
18:13rqou_i'm driving them with m1n1 python scripts and polling for the done bit
18:13svenhrm, let me check if I ever pushed my code to drive that AES engine
18:14svenit should be behind a t6000 dart at least
18:15rqou_also, from the symbols, t6000 doesn't have the error disable bits that t8020 has in the TCR (none of which are in m1n1)
18:16sveni wasn't aware there are any error disable bits
18:16svenI figured out most of DART by poking registers and observing what happens, we didn't have the HV back then :D
18:17rqou_the typedef in the KDK for "dart_blk_sid_config_v8_t" has a bunch of disable bits, e.g. "disable_ttbr_invalid_err" "disable_ste_invalid_error" etc.
18:17svennever looked at the KDK
18:17rqou_there's also bypass_addr_39_32 (and bypass_addr_41_32 for t6000) in the TCR
18:18rqou_but i don't quite understand how bypass mode works (and i'm not sure t8110 has it)
18:18svenif you set that bypass enable bit in TCR it just passes through all DMA transactions
18:18svenI guess you can also give them an offset then? dunno
18:19rqou_isn't the iova space smaller than the physical address space?
18:19svenbut that limitation only applies in translated mode on t8103 at least
18:20svenif you e.g. enable bypass for dwc3 you can just pass the phys addrs that are above that limit to the HW
18:20rqou_but that won't work for devices like jpeg that only seem to itself be able to pass 32-bit physaddrs?
18:21svenunless there's a second register for the upper half of the address or something like that
18:21sventhe DART also needs to support bypass mode
18:21rqou_that's probably what the "bypass_addr_39_32" does then
18:21sventhe pcie ones e.g. don't
18:21rqou_it's in the register that m1n1 seems to be calling TCR
18:22rqou_at bit [23:16]
18:22svenwhat i meant was that to program a 64bit address into dwc3 you write the halves to two separate 32 bit registers
18:22svenmaybe jpeg has something similar or maybe it really only does 32bit addressing
18:22svenif it's the latter you're probably right
18:22rqou_afaict it only does 32bit, but i've only been testing in m1n1 with DART enabled
18:23rqou_fwiw t8110 doesn't seem to have any equivalent setting, so either i haven't found it or it cannot do that
18:24svenmaybe they just got rid of the bypass mode completely
18:24svenis t8110 still limited to 32bit iovas?
18:25rqou_no, 36 bits on everything except the apciec ones, which i haven't managed to check
18:26svenokay, nice
18:26svenon t8103 everything above 32bit got truncated even though technically the pagetables supported more
18:26rqou_i should buy a thunderbolt widget just so i can do what you suggested (read register 0 once macos has set everything up)
18:27rqou_fwiw i haven't _tested_ the full extent of the iova space
18:27rqou_that's just from static analysis of macos drivers
18:27svenokay, then it might still be limited to 32bit
18:27rqou_the hardware config register says 36 bits
18:27rqou_i can test it with the prores block later
18:28svenwould be interesting
18:28rqou_the prores block itself is limited to 42 bit iova space
18:28sventhe t8103 limitation feels like legacy from the old iphones when they had 4kb DARTs
18:29rqou_of the t8110 darts i've looked at, they all support 16 streams, 42-bit physaddrs, and 36-bit iova according to the parameter registers
18:29rqou_the register interface seems to be built for up to 256 streams, but the first 16 seem to still be special with more features
18:29sveninteresting
18:30sveni'd expect the thunderbolt ones to support at least 64
18:30rqou_ok i really need to go out and buy a thunderbolt widget and take a look
18:30svenI wonder if there's something cheap, I just used my tb dock that I had anyway
18:31svenoh, you could also intercept the i2c reads from that tps6598x chip under the HV and pretend that you connected a thunderbolt cable
18:33rqou_the t8110 dart in the t600x feels like a weird transitionary implementation with most of the features missing
18:33rqou_the m2 might be much more "exciting"
18:36jannaunot sure about that, t8110 is used in a15 so it will probably be in m2 (if that doesn't skip a Axy generation)
18:36rqou_yes, what i mean is that the m2 might have t8110 darts with version 2 instead of version 1
18:36rqou_version 2 supports a lot more features
18:38rqou_e.g. invalidating caches with an address range, extra page table levels, and a bunch of other acronym soup i don't understand
18:39rqou_e.g. "STT" "CTC" "STC"
18:41rqou_t8110 also seems to have settings related to controlling or otherwise affecting the number of outstanding transactions, but i don't understand anything about that either
18:46jannauversion is unfortunately not annotated in the ADT
18:46rqou_it's found at register +0x8
18:47rqou_bits [15:8] and [7:0] for major/minor
18:51svensounds like they intend to keep that version for a while then
18:52rqou_feels like it
18:53rqou_btw for reference: not applying the t8110 DART tunables still seems to work (consequences not yet determined), but not applying the prores tunables seems to make the block completely stop working
18:56svenyeah, there’s no need for any dart tunables on t8103 or the either
18:57sven*or t6000
18:57rqou_the tunables code has a debug error message mentioning something about "client partitions"
18:57rqou_but i don't know what that is either lol
18:57svenas long as ignoring it works 🤷🏻‍♂️
18:58rqou_that's why my current guess is that it might be something about caching or something about transaction queue limits
19:00rqou_the scaler and prores blocks have a _huge_ pile of dart tunables
19:31doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
19:55atsalyuk has quit [Ping timeout: 480 seconds]
19:56atsalyuk has joined #asahi-re
20:08atsalyuk has quit [Ping timeout: 480 seconds]
21:06atsalyuk has joined #asahi-re
21:09f14h has joined #asahi-re
21:16atsalyuk has quit [Ping timeout: 480 seconds]
21:17f14h has quit [Quit: f14h]
21:20f14h has joined #asahi-re
21:27f14h has quit [Remote host closed the connection]
22:05alexsv has quit [Ping timeout: 480 seconds]
22:06f14h has joined #asahi-re
22:06f14h has quit [Remote host closed the connection]
22:06f14h has joined #asahi-re
22:12f14h has quit [Remote host closed the connection]
22:48jaalsa has quit [Remote host closed the connection]
…