Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
dds:mipmaps is slow #558
When enabling mipmaps while writing a DDS file, I expected convert's CPU time to double, since it is encoding about twice as many pixels; however, I'm seeing much larger increases in time when running:
For example, the first command takes 32s while the second takes only 7s for an 8192x8192 image. My hunch is that this is due to the fact that in
This likely results in the highest quality mips, but if each iteration could re-use the previous iteration's scaled image, the function could likely be sped up considerably. Perhaps it would make sense to add an option for this? For example:
Hi @dlemstra, sorry for the slow response! Unfortunately, I can't share the particular image that I was using as a test case, though I would guess that almost any large image would exhibit similar behaviour.
If I can find some time to submit a patch, I'll do so :) Thanks.