torfmaster
torfmaster
Wow, good catch! I think this is a general issue coming from `image::imageops::sample::horizontal_sample` which basically affects all filters defined by convolution with 2d functions. I think your suggested fix (using...
Thank you for your comments. I added tasks for your (and my points) so we can track progress more easily.
I have pushed a rather coarse fix for the private dependency checks. However, I don't know how to deal with these checks. I found no easy way of systematically fix...
I did some initial benchmarking: on my intel i9 machine the bench (1024x768 rgb) takes 680ms with the true gaussian blur whereas `fast_blur` takes 57 ms. I checked in both...
I invested some time into performance optimization: * simplify leftmost value by `r * f(0)` * get rid of bound checks by asserting the length of the samples * get...
Furthermore, I see no need for kani/quickcheck tests. First the bounds are obviously correct. On the other hand there is no point in using unsafe so securing the code using...
@etemesi254: I now implemented the performance improvements I am able to understand (that is I used transposing to improve usage of CPU caches). The algorithmic tweaks were already in place,...
My work here is done for now and I would be happy about a code review
@ripytide sorry to bother you (but you already commented on the issue): Do you know what would be the next steps to merge this PR (and get it released)? Thank...
Thank you for your feedback. I created a PR in `imageproc` as well (https://github.com/image-rs/imageproc/pull/683).