00:43 <icecream95> alyssa: That actually doesn't appear to be the case - maybe the fragment shader needing to be in the same block as well is Bifrost only

00:44 <icecream95> So my fix is reliable, but maybe for Midgard only

05:24 <icecream95> Here's a kernel error I haven't seen before...: error powering up gpu {L2,shader,tiler}

09:18 <icecream95> The crc_{read,write}_enable bits did what I expected: FRAG_TRANS_ELIM: 43034 :)

09:40 <icecream95> Mali T760 ... glmark2 Score: 742. Almost there, I'll retry with a release build of Mesa...

10:15 <icecream95> glmark2 Score: 789. Now to find what optimisations to disable to get the score matching the model nubmer :P

10:19 <urjaman> lmao

11:45 <alyssa> icecream95: diuhagueragyraeihugrhk :D

11:46 <alyssa> vsync to 760fps ;P

12:24 <alyssa> For CRC, the first frame should be write but not read

12:25 <alyssa> and then if the surface gets dirtied from a CPU-side access (i.e. the checksum is no longer valid), you do write-but-not-read for one frame to update

12:29 <alyssa> That might have messy interactions with a Bifrost specific opt but we'll cross that when we get there

15:20 <macc24> HdkR: i can rotate screen on odroid go advance back to correct orientation :3