On Fri, 2 Feb 2024 10:26:01 +0100 Miquel Raynal <miquel.raynal@xxxxxxxxxxx> wrote: > Hi Pekka, > > pekka.paalanen@xxxxxxxxxxxxx wrote on Fri, 2 Feb 2024 10:55:22 +0200: > > > On Thu, 01 Feb 2024 18:31:32 +0100 > > Louis Chauvet <louis.chauvet@xxxxxxxxxxx> wrote: > > > > > Change the composition algorithm to iterate over pixels instead of lines. > > > It allows a simpler management of rotation and pixel access for complex formats. > > > > > > This new algorithm allows read_pixel function to have access to x/y > > > coordinates and make it possible to read the correct thing in a block > > > when block_w and block_h are not 1. > > > The iteration pixel-by-pixel in the same method also allows a simpler > > > management of rotation with drm_rect_* helpers. This way it's not needed > > > anymore to have misterious switch-case distributed in multiple places. > > > > Hi, > > > > there was a very good reason to write this code using lines: > > performance. Before lines, it was indeed operating on individual pixels. > > > > Please, include performance measurements before and after this series > > to quantify the impact on the previously already supported pixel > > formats, particularly the 32-bit-per-pixel RGB variants. > > > > VKMS will be used more and more in CI for userspace projects, and > > performance actually matters there. > > > > I'm worrying that this performance degradation here is significant. I > > believe it is possible to keep blending with lines, if you add new line > > getters for reading from rotated, sub-sampled etc. images. That way you > > don't have to regress the most common formats' performance. > > While I understand performance is important and should be taken into > account seriously, I cannot understand how broken testing could be > considered better. Fast but inaccurate will always be significantly > less attractive to my eyes. > > I am in favor of making this working first, and then improving the code > for faster results. Maybe the line-driven approach can be dedicated to > "simpler" formats where more complex corner cases do not happen. But > for now I don't see the point in comparing performances between broken > and (hopefully) non broken implementations. Hi, usually that is a good idea, but in this case you are changing the fundamental design of the algorithm. That makes it easy to add new things (e.g. YUV support) that will be even harder to make faster later. If you then later improve performance, that is another re-design. You would be making maintainers review two rewrites instead of one or less. I suspect it might be less effort for the author as well to not ditch the line-based based algorithm as the first step. Thanks, pq
Attachment:
pgp0JzAiIX8E6.pgp
Description: OpenPGP digital signature
