Zen master-of-zen

let mut sum = 0;

for x in 1..=scale {
  let src_row = &data_origin[(src.cfg.stride * 2 + x)..];
  for y in 1..=scale {
    sum += u32::cast_from(src_row[col * 2 + y]) as usize
  }
}

let mut sum = 0;

for x in 1..=scale {
  let src_row = &data_origin[(src.cfg.stride * 2 + x)..];
  sum += (1..scale)
    .into_par_iter()
    .for_each(|y| u32::cast_from(src_row[col * 2 + y]) as usize)
    .sum();
 }

  for x in 1..=scale {
    let src_row = &data_origin[(src.cfg.stride * 2 + x)..];
    let local_sum: usize = (1..scale)
      .into_par_iter()
      .map(|y| u32::cast_from(src_row[col * 2 + y]) as usize)
      .sum();
    sum += local_sum;
  }

Goals and motivation

Current fast scene detection in rav1e relatively slow to compared methods of scene detection, and have proclivity to show false results or don't detect scene changes where they appear

This pr reworks fast scene detection algorithm, making it faster, better, and more accurate

Achieved goals are:

Faster decision making ( Both less and more efficient computations )
More accurate scene detection, by adjusting threshold based on previous frames
Frame downscale for faster decisions.

Total wall time difference on speed 9

Version	Time
old:	`47,783.666`
new:	`44,202.432`

-8% wall time diffrence

1080p is ~7 times faster

new | old

Install python
Install rustup
Install rust nightly

git clone https://github.com/master-of-zen/Av1an
cd Av1an
python3 -m venv env
source env/bin/activate
pip install maturin

---- api::test::output_frameno_no_scene_change_at_multiple_flashes stdout ----
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false
[src/scenechange/mod.rs:186] scenecut = false

Slow

>  Using y4m decoder: 1920x1080p @ 25/1 fps, 4:2:0, 8-bit
>  Using 16 tiles (4x4)
>  encoded 250 frames, 5.239 fps, 2482.74 Kb/s, elap. time: 47s
>  ----------
>  Key frame:             3 | avg QP:  79.00 | avg size:   38309 B
>  Inter frame:         247 | avg QP: 138.00 | avg size:   12099 B
>  Intra only frame:      0 | avg QP:   0.00 | avg size:       0 B
>  Switching frame:       0 | avg QP:   0.00 | avg size:       0 B

	# Fields meanings: <source root>/av1/encoder/firstpass.h
	fields = [
	"frame",
	"weight",
	"intra_error",
	"frame_avg_wavelet_energy",
	"coded_error",
	"sr_coded_error",
	"tr_coded_error",
	"pcnt_inter",

	filmgrn1
	E 0 834166 1 7391 1
	p 3 8 0 11 0 1 128 192 256 128 192 256
	sY 8 0 43 13 43 27 51 40 68 54 82 67 90 81 93 255 93
	sCb 8 0 41 13 41 40 51 54 58 67 63 81 65 94 65 255 65
	sCr 4 0 26 13 26 54 32 255 35
	cY 1 0 -1 8 3 0 1 0 3 -1 -21 -12 2 1 1 0 -21 83 19 -10 2 17 -42 108
	cCb 3 2 0 8 2 2 2 3 1 2 -16 -5 5 3 0 6 -16 73 8 -3 5 20 -37 90 -2
	cCr 2 1 1 6 3 1 2 1 5 -2 -6 -8 4 1 2 -3 -14 85 15 -1 3 11 -12 87 -1
	E 834166 1251250 1 0 1