16:27 <steev> bamse: been testing that branch - aside from the thing i noticed in #freedreno last night, seems good(ish) - wifi issues not withstanding :)

16:38 <bamse> steev: what was the #freedreno-reported thing?

16:39 <bamse> steev: i did speak to sibi about further debugging the wifi issue, so i have some more work to acquire the right ramdumps etc so that i can get some assitance from qualcomm

16:39 <steev> the screen is all "corrupted" with mesa 21.3 rc1, because it uses uwbc not linear

16:40 <steev> bamse: https://pastebin.com/xTxjbc6r

16:40 <bamse> cool effect

16:42 <bamse> okay, so we need to double check the ubwc configuration on the dpu side of sc8180x then i presume

16:42 <steev> likely, and that's way above my skillset/paygrade - i'm gonna double check tonight if reverting the commit rob mentioned "fixes" it

16:47 <bamse> i have no idea what the ubwc configuration would entail, the only thing i know of was to say that sc8180x has DPU_HW_UBWC_VER_30

16:47 <bamse> perhaps that's wrong, or perhaps there's something missing elsewhere

16:52 <robclark> bamse: for dpu side, I guess 544d8b96150dea3d439dcaf66ac01eca58f4016c might give an idea where to look.. and d0bac4e9cd66fe0c0f5d7fbbf2d0f9956a3d8916 for gpu side..

16:54 <robclark> looking at downstream, it does look like there are some gpu side differences.. (but I guess that makes sense because double # of ddr interfaces compared to 640)

16:55 <bamse> robclark: thanks for the pointers

16:55 <bamse> yeah, that looks most likely

16:55 <robclark> this looks plausibly important:

16:55 <robclark> https://www.irccloud.com/pastebin/rVhAWisW/

16:56 <robclark> also I think we want amsbc to be 1

16:57 <bamse> amsbc should be 1 already, thanks to a680 being in the a640 familiy

16:57 <bamse> we do however not set BIT(0) of NC_MODE_CNTL

16:59 <robclark> ahh, right