ImBuf: no need to use double precision inside IMB_transform #117160

Aras Pranckevicius · 2024-01-16T10:24:52+01:00

2024-01-16 10:24:52 +01:00

Double precision pixel coordinate interpolation was added in 3a65d2f591 to fix an issue of "wobbly" resulting image at very high scale factors. But the root cause of that was the fact that the scanline loop was repeatedly adding the floating point step for each pixel, instead of doing multiplication by pixel index (repeated floating point additions can "drift" due to imprecision, whereas multiplications are much more accurate).

Change all that math back to use single precision floats. I checked the original issue the commit was fixing, it is still fine. Also added a gtest to cover this situation: imbuf_transform, nearest_very_large_scale

This makes IMB_transform a tiny bit faster, on Windows/VS2022/Ryzen5950X scaling an image at 4K resolution:

Bilinear: 5.92 -> 5.83 ms
Subsampled3x3: 53.6 -> 52.7 ms

Double precision pixel coordinate interpolation was added in 3a65d2f5911 to fix an issue of "wobbly" resulting image at very high scale factors. But the root cause of that was the fact that the scanline loop was repeatedly adding the floating point step for each pixel, instead of doing multiplication by pixel index (repeated floating point additions can "drift" due to imprecision, whereas multiplications are much more accurate). Change all that math back to use single precision floats. I checked the original issue the commit was fixing, it is still fine. Also added a gtest to cover this situation: `imbuf_transform, nearest_very_large_scale` This makes `IMB_transform` a tiny bit faster, on Windows/VS2022/Ryzen5950X scaling an image at 4K resolution: - Bilinear: 5.92 -> 5.83 ms - Subsampled3x3: 53.6 -> 52.7 ms

Aras Pranckevicius changed title from ~~ImBuf: no need to use double precision inside IMB_transform~~ to WIP: ImBuf: no need to use double precision inside IMB_transform

2024-01-16 11:33:00 +01:00

Aras Pranckevicius force-pushed imbuf-no-doubles from f929c15500 to cd2d407061

2024-01-16 11:39:14 +01:00

Compare

Aras Pranckevicius commented

2024-01-16 11:39:48 +01:00

@blender-bot build

Aras Pranckevicius changed title from ~~WIP: ImBuf: no need to use double precision inside IMB_transform~~ to ImBuf: no need to use double precision inside IMB_transform

2024-01-16 12:43:50 +01:00

Aras Pranckevicius requested review from Richard Antalik 2024-01-16 12:44:01 +01:00

Richard Antalik approved these changes 2024-01-16 12:55:58 +01:00

Richard Antalik left a comment

That's quite nice catch. Probably wouldn't cross my mind to solve it this way.

Aras Pranckevicius merged commit f16dbd0360 into main

2024-01-16 13:32:56 +01:00

Aras Pranckevicius referenced this issue from a commit

2024-01-16 13:32:57 +01:00

ImBuf: no need to use double precision inside IMB_transform

Aras Pranckevicius deleted branch imbuf-no-doubles

2024-01-16 13:32:58 +01:00