The Extend version of the samplers is used (except for Bilinear which will use Wrap when requested because it was doing it correctly). Clip is done by here so it can be enabled per-axis and it…
These are just code cleanup from an aborted attempt to make the same changes to the bicubic sampler. I gave up on that as being too intrusive, in particular because the bicubic one always does…
This removed a double-invert of the matricies and most likely made the result a lot more accurate.
Sure, but that doesn't belong to this pull request.
I'll look into splitting it if…
I do not think the current GPU implementation needs as many changes as this proposal claims.
The Domain has a transform and a size that define a rectangle (actually a quad) which I believe…
This removed a double-invert of the matricies and most likely made the result a lot more accurate. But it's primary advantage is that it is a lot easier to figure out what is happening. I had to…
I have merged this change into the main patch blender/blender#124381 as that results in the minimal change of behavior.
Patch has been updated with something that fixes the unexpected blur when odd images are centered in even ones, without altering normal transformations, I have been testing this for some time…
Patch has been updated. It includes what seems to be a working solution for the "unexpected blur" when an odd-sized domain is realized on an even-sized one. This is a .5 shift, just like before,…
The reported bug is caused by incorrect crops in MemoryBuffer::read(). Fixing or removing these however revealed that interpolation of pixels at the edge is done incorrectly. Nearest and…
I have been using it like this for a while and it seems to be working well.
I now believe that blender/blender#124841 is a better fix for unexpected blur.
Note that I am convinced any solution must actually do a .5 translation of one…
This does not work for non-integer translations. Cancelling this pull request.