rotate-me-fast/TODO.md

[-] Draw rotated pixels in src order -> cache write miss
[X] Use atan2 at beginning and end of line.
    Interpolation in-between values
[X] Test pixel perfect 0, 90

[X] Optimization for square images
[X] Fixed point computation
[-] -funroll-loops -> no gain
[-] restrict qualifier -> unavailable in C++
[ ] All positions as simple integer

# Cache
[-] Rotate per channel -> no gain
[X] Cut image in tiles
 [X] Overlap
 [-] Rotate in one temp tile then copy/move it
 [X] Align tiles in memory
[ ] Touch beginning of tile

## Alignement
[X] RGBX format (create pixel structure) on 8 bytes (can do computation in-place)
[X] Load pixels in 64-bit variable
 [X] Directly load in SIMD 128-bit variable
[ ] Align memory on 16 bytes (would require padding)
[X] RGBX tiles

## Layout
[ ] Pack 4 neighbors in 16B structure (aligned)
    Each point is followed by the point below
[ ] Spiral layout?

# Quality
[X] Interpolate using SIMD, SSE
[ ] Fix out-of-bounds pixel set
  [X] Use padding
  [ ] Image borders
Implement pixels as RGBX structure. - Add ‘make debug’ target - Add links in README 2014-07-12 22:37:15 +02:00			`[-] Draw rotated pixels in src order -> cache write miss`
Add check_90(). - Fix border. 2014-06-26 21:28:34 +02:00			`[X] Use atan2 at beginning and end of line.`
			`Interpolation in-between values`
Remove boundary check in rotate_pixel(). This check makes the last line invalid. 2014-08-06 22:33:41 +02:00			`[X] Test pixel perfect 0, 90`
FIll new interpolated pixels. - Take input image in argument. - Allow set_pixel() out of bounds, noop. 2014-06-26 23:45:02 +02:00
Update TODO. 2014-08-06 21:44:00 +02:00			`[X] Optimization for square images`
Implement pixels as RGBX structure. - Add ‘make debug’ target - Add links in README 2014-07-12 22:37:15 +02:00			`[X] Fixed point computation`
			`[-] -funroll-loops -> no gain`
			`[-] restrict qualifier -> unavailable in C++`
Do one memset for left and right padding. 2014-08-08 19:29:46 +02:00			`[ ] All positions as simple integer`
Use a unique buffer for pixels (scanline). - Add rotate_pixel(). - Remove APoint, use Point instead. 2014-06-28 13:52:24 +02:00
			`# Cache`
Update TODO. 2014-07-04 00:23:21 +02:00			`[-] Rotate per channel -> no gain`
			`[X] Cut image in tiles`
Create small overlap in tiles. Tiles overlap on the bottom line and the right column. Having an overlap allows us to get neighbor points for each point in a tile, regardless of its position. 2014-07-10 19:40:51 +02:00			`[X] Overlap`
Remove temp tile. 2014-07-08 08:48:17 +02:00			`[-] Rotate in one temp tile then copy/move it`
Allocate tiles as a unified buffer. 2014-07-09 07:50:09 +02:00			`[X] Align tiles in memory`
Implement RGBX for tiled images. Interpolation is now done with SIMD with -DSIMD for tiled images as well. 2014-07-16 20:44:32 +02:00			`[ ] Touch beginning of tile`
Implement pixels as RGBX structure. - Add ‘make debug’ target - Add links in README 2014-07-12 22:37:15 +02:00
			`## Alignement`
			`[X] RGBX format (create pixel structure) on 8 bytes (can do computation in-place)`
Remove RGBX structure. The pixels are still packed as RGBX in memory but no structure is created, it’s just a contiguous buffer. Interpolation is now done with SIMD on integer values. - Add SIMD define. 2014-07-16 02:08:32 +02:00			`[X] Load pixels in 64-bit variable`
			`[X] Directly load in SIMD 128-bit variable`
			`[ ] Align memory on 16 bytes (would require padding)`
Implement RGBX for tiled images. Interpolation is now done with SIMD with -DSIMD for tiled images as well. 2014-07-16 20:44:32 +02:00			`[X] RGBX tiles`
Implement pixels as RGBX structure. - Add ‘make debug’ target - Add links in README 2014-07-12 22:37:15 +02:00
			`## Layout`
			`[ ] Pack 4 neighbors in 16B structure (aligned)`
			`Each point is followed by the point below`
Create small overlap in tiles. Tiles overlap on the bottom line and the right column. Having an overlap allows us to get neighbor points for each point in a tile, regardless of its position. 2014-07-10 19:40:51 +02:00			`[ ] Spiral layout?`
Use a unique buffer for pixels (scanline). - Add rotate_pixel(). - Remove APoint, use Point instead. 2014-06-28 13:52:24 +02:00
			`# Quality`
Implement RGBX for tiled images. Interpolation is now done with SIMD with -DSIMD for tiled images as well. 2014-07-16 20:44:32 +02:00			`[X] Interpolate using SIMD, SSE`
Update TODO. 2014-08-06 21:44:00 +02:00			`[ ] Fix out-of-bounds pixel set`
			`[X] Use padding`
			`[ ] Image borders`