#panfrost on 2020-08-05 — irc logs at freenode.irclog.whitequark.org

2019-09-06 11:20 alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature

00:23 rhyskidd has quit [Quit: rhyskidd]

01:04 vstehle has quit [Ping timeout: 256 seconds]

02:21 phh has quit [Remote host closed the connection]

02:22 phh has joined #panfrost

03:03 buzzmarshall has quit [Remote host closed the connection]

03:39 davidlt has joined #panfrost

04:16 davidlt has quit [Ping timeout: 256 seconds]

05:00 vstehle has joined #panfrost

05:02 kaspter has joined #panfrost

05:22 icecream95 has joined #panfrost

05:51 camus1 has joined #panfrost

05:53 kaspter has quit [Ping timeout: 240 seconds]

05:53 camus1 is now known as kaspter

06:40 kaspter has quit [Ping timeout: 240 seconds]

06:40 kaspter has joined #panfrost

06:57 camus1 has joined #panfrost

06:58 kaspter has quit [Ping timeout: 240 seconds]

06:58 camus1 is now known as kaspter

07:25 davidlt has joined #panfrost

07:49 camus1 has joined #panfrost

07:49 kaspter has quit [Ping timeout: 240 seconds]

07:49 camus1 is now known as kaspter

07:50 <icecream95> `glmark2-es2 -b terrain` shows rendering artefacts when PAN_MESA_DEBUG=nofp16 isn't used

08:45 stikonas has joined #panfrost

08:54 raster has joined #panfrost

09:27 <tomeu> wonder what happened here: https://tracie.freedesktop.org/dashboard/imagediff/tomeu/mesa/3950673/traces/799ce7b2d8050d6ba2e3ae595cf8729e/gputest/plot3d.trace/

09:27 <tomeu> the reference image was rendered by radeonsi

09:29 <tomeu> this one is also interesting: https://tracie.freedesktop.org/dashboard/imagediff/tomeu/mesa/3950673/traces/98e742377cef049e98014300a34d7bba/gputest/gimark.trace/

09:29 <tomeu> some blue rings are missing on the top left corner

09:30 raster has quit [Quit: Gettin' stinky!]

09:35 raster has joined #panfrost

09:41 <icecream95> tomeu: plot3d uses glPolygonMode, which isn't implemented. This also affects moebius from xscreensaver.

09:42 <tomeu> cool

09:42 <tomeu> icecream95: btw, regarding terrain: https://tracie.freedesktop.org/dashboard/imagediff/tomeu/mesa/3950673/traces/52f19a65bd77ae03b5f51ee0038a2982/glmark2/terrain.rdc/

09:43 <tomeu> guess the main difference with radeonsi is that whatever random function is used to generate the terrain, is different

09:43 <tomeu> but I don't see any artifacts on the panfrost rendering

09:45 <tomeu> all the fragment ones seem to display the same problem: https://tracie.freedesktop.org/dashboard/imagediff/tomeu/mesa/3950673/traces/716f1ebf84d9ab30a8c4261ecbed492b/glmark2/loop-fragment-steps=5:fragment-uniform=false:vertex-steps=5.rdc/

09:56 _whitelogger has joined #panfrost

09:58 <tomeu> oh, freedreno has the same problem there: https://tracie.freedesktop.org/dashboard/imagediff/mesa/mesa/3937596/traces/ec5ea3eab5b54524748a0c883cc929d6/glmark2/loop-fragment-steps=5:fragment-uniform=true:vertex-steps=5.rdc/

10:05 raster has quit [Quit: Gettin' stinky!]

10:05 raster has joined #panfrost

10:27 <icecream95> It seems to be a floating-point precision error - adding brackets to make gl_FragCoord.y be multiplied with 0.0001 before gl_FragCoord.x fixes the glmark loop tests

10:27 raster has quit [Quit: Gettin' stinky!]

10:32 nlhowell has quit [Quit: WeeChat 2.9]

10:34 nlhowell has joined #panfrost

10:44 raster has joined #panfrost

10:54 kaspter has quit [Ping timeout: 265 seconds]

10:54 kaspter has joined #panfrost

10:56 <icecream95> tomeu: The job log for gimark has OUT_OF_MEMORY faults

10:56 <tomeu> yeah, I see those in other jobs as well

10:56 <tomeu> not sure what this is about

10:57 <tomeu> maybe we aren't sizing some BO that the GPU renders to

10:57 <tomeu> the polygon list, or the scratch

11:54 warpme_ has joined #panfrost

11:58 icecream95 has quit [Quit: leaving]

12:08 camus1 has joined #panfrost

12:08 kaspter has quit [Ping timeout: 260 seconds]

12:08 camus1 is now known as kaspter

12:32 camus1 has joined #panfrost

12:33 kaspter has quit [Ping timeout: 256 seconds]

12:33 camus1 is now known as kaspter

12:41 nlhowell has quit [Ping timeout: 256 seconds]

13:13 camus1 has joined #panfrost

13:13 kaspter has quit [Ping timeout: 246 seconds]

13:13 camus1 is now known as kaspter

13:32 <alyssa> icecream95: glmark2 has precision bugs affecting terrain and the fragcoord ones, freedreno has the same issues, etc

13:32 <alyssa> oh tomeu brought that up

13:32 <alyssa> FragCoord I think you need to copy into a highp var before multiplying

13:33 <alyssa> the terrain one I once bisected which line needed a highp qualifier

13:35 <alyssa> tomeu: OUT_OF_MEMORY refers to the (GROWABLE) tiler_heap being too small, I believe.

13:35 <alyssa> but usually those faults are a symptom of something else going wrong I think

13:45 camus1 has joined #panfrost

13:46 kaspter has quit [Ping timeout: 240 seconds]

13:47 camus1 is now known as kaspter

13:48 <tomeu> it's interesting that the rendering seems to be correct

13:48 <tomeu> wonder if we are lucky and the polygons are hidden, or something like that

17:05 davidlt has quit [Ping timeout: 265 seconds]

17:40 gcl_ has joined #panfrost

17:43 gcl has quit [Ping timeout: 265 seconds]

18:06 mifritscher has quit [Quit: Quit]

18:22 davidlt has joined #panfrost

19:27 davidlt has quit [Ping timeout: 260 seconds]

19:35 raster has quit [Quit: Gettin' stinky!]

19:51 raster has joined #panfrost

19:58 buzzmarshall has joined #panfrost

20:17 mifritscher has joined #panfrost

20:18 mifritscher has quit [Client Quit]

20:20 mifritscher has joined #panfrost

21:40 raster has quit [Quit: Gettin' stinky!]

23:03 stikonas has quit [Remote host closed the connection]

23:05 warpme_ has quit [Quit: Connection closed for inactivity]

23:12 robmur01_ has joined #panfrost

23:14 robmur01 has quit [Ping timeout: 240 seconds]