mirror of
https://github.com/deater/dos33fsprogs.git
synced 2024-11-16 23:20:43 +00:00
41 lines
1.2 KiB
Plaintext
41 lines
1.2 KiB
Plaintext
frame50 as benchmark:
|
|
+ raw binary is 1024 bytes (2048 pixels)
|
|
|
|
hand-crafted:
|
|
+ 25e0 - 24d8 = 264 bytes
|
|
|
|
todo, compressed:
|
|
+ ???
|
|
|
|
|
|
7680 bytes: try1: all plots, even zeros (4 bytes each pixel)
|
|
4537 bytes: try2: all plots, don't write black pixels
|
|
3701 bytes: try3: only do color if it changes
|
|
2568 bytes: try4: use compact form (only change type if changed)
|
|
1194 bytes: try5: use hlin (essentially RLE)
|
|
906 bytes: try6: sort by color first
|
|
628 bytes: try7: initial use boxes
|
|
526 bytes: try8: boxes that optimize to HLIN/VLIN/PLOT
|
|
522 bytes: try9: optimize HLIN to HLIN_ADD variants
|
|
506 bytes: try10: sort types per color
|
|
498 bytes: try11: sort HLIN by Y1
|
|
431 bytes: try12: with don't cares
|
|
336 bytes: try13: sort colors by popularity
|
|
299 bytes: try14: shrink the window for finding colors to only include colors
|
|
299 bytes: try15: sort BOX (not expected to be smaller)
|
|
293 bytes: try16: add BOX_ADD types
|
|
293 bytes: try17: permutate the top 4 colors to see if it improves
|
|
|
|
TODO:
|
|
Only use don't care result if not wider than max X or longer than
|
|
max Y
|
|
sort BOX and do optimizations like HLIN
|
|
|
|
|
|
Try all color combos?
|
|
K-permutations. For K=16 it's literally billions+?
|
|
Maybe try just top 4? Only 24 combos then?
|
|
|
|
MOVE PLOTS one line before HLIN?
|
|
|