#asahi-gpu · 2023-09-03 14:09
karolherbstand at first I thought that was my fault :')#asahi-gpu · 2023-09-03 13:38
marcanit's less broken at least :p#asahi-gpu · 2023-09-03 13:17
alyssastill broken, another fix coming#asahi-gpu · 2023-09-03 13:03
jannaufix confirmed on asahi alarm#asahi-gpu · 2023-09-03 12:41
marcanpushed a fix#asahi-gpu · 2023-09-03 09:03
rosefromthedeaddid drm_asahi_get_time land? i can't see it anywhere#asahi-gpu · 2023-09-03 09:00
jannausame as reported with supertuxkart in https://github.com/AsahiLinux/linux/issues/72#issuecomment-1703891618#asahi-gpu · 2023-09-03 08:48
jannauseen with prusa-slicer#asahi-gpu · 2023-09-03 08:48
jannaufixed by reverting "bf4d2a40bada asahi: Force staging copy for LINEAR buffer reads from WC BOs"#asahi-gpu · 2023-09-03 08:45
jannauresult in a non-debug build is a segfault in disk_cache_compute_key()#asahi-gpu · 2023-09-03 08:44
jannauthe agx_linear_allowed one in agx_select_best_modifier()#asahi-gpu · 2023-09-03 08:43
jannauasahi-20230902 hits an assert while allocating a staging resource for a compressed texture (DXT5_RGBA)#asahi-gpu · 2023-09-01 11:07
jannaulina: segfault https://paste.debian.net/1290653/ but I'll try first the branch from #184#asahi-gpu · 2023-09-01 10:03
_jannau__lina: sysprof works and is alyssa approved#asahi-gpu · 2023-09-01 10:03
_jannau__lina: it end up with _mesa_GetTexSubImage_sw/util_format_b10g10r10x2_unorm_unpack_rgba_8unorm#asahi-gpu · 2023-09-01 10:00
_jannau__lina: latest analyse regarding the screen capture is in https://oftc.irclog.whitequark.org/asahi-gpu/2023-08-20#asahi-gpu · 2023-08-31 15:01
rosefromthedeadnoted#asahi-gpu · 2023-08-31 15:01
karolherbstrosefromthedead: if you want to do that timer stuff feel free, I have way too many other things to work on#asahi-gpu · 2023-08-31 14:57
rosefromthedeadgood to know though :eyes:#asahi-gpu · 2023-08-31 14:57
i509vcbAh ok nvm#asahi-gpu · 2023-08-31 14:56
rosefromthedeadmb i meant wrt the timer stuff, because i said i'd do that but i haven't got round to it#asahi-gpu · 2023-08-31 14:56
i509vcbVK_EXT_image_drm_format_modifier is one of the big extensions for getting wlroots vulkan renderer to work, but there are a few dependent extensions needed there#asahi-gpu · 2023-08-31 14:54
i509vcbThere's definitely stuff in agxv if you want to burn time there#asahi-gpu · 2023-08-31 14:49
rosefromthedeadhmm am i needed#asahi-gpu · 2023-08-31 12:02
karolherbstanyway.. what rusticl is kinda doing is to insert `PIPE_QUERY_TIMESTAMP` around gallium API calls, that happens in batches and at some point that all gets flushed out. And reading the result happens all on the CPU side. #asahi-gpu · 2023-08-31 11:53
karolherbst_but_ I guess `drm_asahi_result_compute` is good enough here, I just have to figure out how to pipe that all through#asahi-gpu · 2023-08-31 11:53
alyssa:+1:#asahi-gpu · 2023-08-31 11:50
karolherbstI don't need anything inside the shader afaik#asahi-gpu · 2023-08-31 11:50
karolherbstanyway.. I need proper support for PIPE_QUERY_TIMESTAMP and pipe_screen::get_timestamp (which I guess should use the drm_asahi_get_time ioctl)#asahi-gpu · 2023-08-31 11:48
karolherbstlike I kinda need to know when the GPU reaches a certain point in the command stream#asahi-gpu · 2023-08-31 11:47
karolherbstyeah.. so I kinda need it as part of command submission and as a host query, so probably 1 and 2#asahi-gpu · 2023-08-31 11:47
alyssait does NOT map to gl timestamps which are bonkers for tilers#asahi-gpu · 2023-08-31 11:46
alyssahf#asahi-gpu · 2023-08-31 11:46
alyssaand also os_time_get_nano()#asahi-gpu · 2023-08-31 11:46
alyssaand drm_asahi_result_* has timestamps in it at the end of the job#asahi-gpu · 2023-08-31 11:46
alyssadrm_asahi_get_time ioctl#asahi-gpu · 2023-08-31 11:46
alyssatimestamp you have two options#asahi-gpu · 2023-08-31 11:46
alyssascratch is also nightmare level yes#asahi-gpu · 2023-08-31 11:43
karolherbstunless it's already REed#asahi-gpu · 2023-08-31 11:43
karolherbstI guess#asahi-gpu · 2023-08-31 11:43
karolherbstbut anyway, I can look into TIMESTAMP, I just don't have any of the RE setup done which might be needed here#asahi-gpu · 2023-08-31 11:42
karolherbstI need scratch for indirect access to temporaries at least#asahi-gpu · 2023-08-31 11:41
karolherbstlina told me that the uapi is at least done. I can see that spilling is a nightmare to implement, but is scratch on its own as a thread private memory buffer all that complex? lina told me about some of the details though, so I can also kinda see why that's annoying to implement#asahi-gpu · 2023-08-31 11:40
alyssascratch/spilling are nightmare-level difficulty and yes it's in the cards for me & lina#asahi-gpu · 2023-08-31 11:38
karolherbstuhm... maybe? are you planning to finish scratch memory + spilling? But I kinda prefer to figure out the linear image situation first, but if timestamp queries are easy enough I can also look into that#asahi-gpu · 2023-08-31 11:37
alyssakarolherbst: Any chance you would do TIMESTAMP queries?#asahi-gpu · 2023-08-31 10:28
Marynice#asahi-gpu · 2023-08-31 00:01
alyssa1% improvement? i'll take it lol#asahi-gpu · 2023-08-31 00:01
alyssaforce caching sources gets me up to 10.1fps ... need to figure out a heuristic for that ...#asahi-gpu · 2023-08-30 22:53
i509vcbalyssa: oops in patch, here's another