00:47 gfxstrand[d]: Oh, what I meant was that the N in `GOB_N` was `1 << y_log2`. The actual computation to get the enum value is `y_log2 - 1`, I think.
01:13 dwlsalmeida[d]: Yes I got that, problem is, there’s two values of y_log2
01:14 dwlsalmeida[d]: If you use the one for plane[0], chroma is broken, if you use the one for plane[1], luma is broken
01:26 gfxstrand[d]: Do they need to match?
01:26 gfxstrand[d]: I could believe that they do
01:28 gfxstrand[d]: If there's only one gob_height field in the decoder, I could easily believe they have to match.
01:29 gfxstrand[d]: Which means the video code needs to assert they match and we need to figure out how to teach the image code to make them match.
01:45 dwlsalmeida[d]: That’s what I think too
01:45 dwlsalmeida[d]: To change the nil stuff so they match
01:46 dwlsalmeida[d]: Is it possible to just change it? What more do I have to update in theory?
01:59 gfxstrand[d]: The tricky part is that they're separate `nil::Image`s and they're constructed independently.
01:59 gfxstrand[d]: We could add an entry point that constructed a pair of images and somehow wire that up when video is requested in the usage.
02:02 gfxstrand[d]: We could also construct the UV plane first on the assumption that it'll always have the smaller tiling and then add a way to force the tiling when we create the Y plane.
03:01 airlied: we have the same problem on radv, we ended up going linear but I do want to fix it
03:01 airlied: anv also has a problem with arrays for video as it does all the Y planes then all the UV planes and the decoder can't deal with the gap
23:30 althaser: Hi Folks, I have an old graphical card and currently I am facing the following issue:
23:30 althaser: https://pastebin.com/Ba3ZZsLn it seems I am missing the firmware, can you help me getting these files ? any hint ? thank you