bgra_to_rgb benchmark

benches.rs

use bgra_to_rgb::*;
use criterion::*;
use rand::{Rng, SeedableRng};
use rand_chacha::ChaChaRng;

pub fn benchmark_move(c: &mut Criterion) {
    // random bytes from random.org
    let rand = ChaChaRng::from_seed([
        0x1c, 0xd7, 0x75, 0x38, 0x10, 0x43, 0xf8, 0x9a, 0x98, 0xe5, 0x53, 0x56, 0xfa, 0xe5, 0x07,
        0x96, 0x02, 0x84, 0xcf, 0xc8, 0x3f, 0x63, 0xac, 0x55, 0xee, 0x6d, 0x24, 0xb5, 0x5b, 0xff,
        0x08, 0xb7,
    ]);

    let mut group = c.benchmark_group("bgra×4_to_rgb×4 (by move)");
    // NB: macro for monomorphization
    macro_rules! bench {
        ($name:expr, $f:ident) => {
            group.bench_function($name, |b| {
                let mut rand = rand.clone();
                b.iter_batched(|| rand.gen(), |bgra| $f(bgra), BatchSize::SmallInput);
            });
        };
    }

    bench!("loop", convert_loop);
    bench!("chunks", convert_chunks);
    bench!("raw", convert_raw);
    bench!("scatter", convert_scatter);
    bench!("swizzle", convert_swizzle);
    bench!("shift", convert_shift);
    bench!("overwrite", convert_overwrite);
}

pub fn benchmark_ref(c: &mut Criterion) {
    // random bytes from random.org
    let rand = ChaChaRng::from_seed([
        0x1c, 0xd7, 0x75, 0x38, 0x10, 0x43, 0xf8, 0x9a, 0x98, 0xe5, 0x53, 0x56, 0xfa, 0xe5, 0x07,
        0x96, 0x02, 0x84, 0xcf, 0xc8, 0x3f, 0x63, 0xac, 0x55, 0xee, 0x6d, 0x24, 0xb5, 0x5b, 0xff,
        0x08, 0xb7,
    ]);

    let mut group = c.benchmark_group("bgra×4_to_rgb×4 (by ref)");
    // NB: macro for monomorphization
    macro_rules! bench {
        ($name:expr, $f:ident) => {
            group.bench_function($name, |b| {
                let mut rand = rand.clone();
                b.iter_batched_ref(|| rand.gen(), |bgra| $f(bgra), BatchSize::SmallInput);
            });
        };
    }

    bench!("loop", convert_loop_r);
    bench!("chunks", convert_chunks_r);
    bench!("raw", convert_raw_r);
    bench!("scatter", convert_scatter_r);
    bench!("swizzle", convert_swizzle_r);
    bench!("shift", convert_shift_r);
    bench!("overwrite", convert_overwrite_r);
}

criterion_group!(
    name = benches;
    config = Criterion::default();
    targets = benchmark_move, benchmark_ref
);

criterion_main!(benches);

lib.rs

#![feature(portable_simd, slice_as_chunks)]

// https://users.rust-lang.org/t/converting-a-bgra-u8-to-rgb-u8-n-for-images/67938/7?u=cad97
pub fn convert_loop(bgra: [u8; 16]) -> [u8; 12] {
    let mut rgb = [0u8; 12];
    for i in (0..12).step_by(3) {
        rgb[i] = bgra[i / 3 * 4 + 2];
        rgb[i + 1] = bgra[i / 3 * 4 + 1];
        rgb[i + 2] = bgra[i / 3 * 4];
    }
    rgb
}

pub fn convert_loop_r(bgra: &[u8; 16]) -> [u8; 12] {
    let mut rgb = [0u8; 12];
    for i in (0..12).step_by(3) {
        rgb[i] = bgra[i / 3 * 4 + 2];
        rgb[i + 1] = bgra[i / 3 * 4 + 1];
        rgb[i + 2] = bgra[i / 3 * 4];
    }
    rgb
}

// https://users.rust-lang.org/t/converting-a-bgra-u8-to-rgb-u8-n-for-images/67938/9?u=cad97
pub fn convert_chunks(bgra: [u8; 16]) -> [u8; 12] {
    let mut rgb: [u8; 12] = [0; 12];
    for (src, dst) in bgra.chunks_exact(4).zip(rgb.chunks_exact_mut(3)) {
        dst[0] = src[2];
        dst[1] = src[1];
        dst[2] = src[0];
    }
    rgb
}

pub fn convert_chunks_r(bgra: &[u8; 16]) -> [u8; 12] {
    let mut rgb: [u8; 12] = [0; 12];
    for (src, dst) in bgra.chunks_exact(4).zip(rgb.chunks_exact_mut(3)) {
        dst[0] = src[2];
        dst[1] = src[1];
        dst[2] = src[0];
    }
    rgb
}

// https://users.rust-lang.org/t/converting-a-bgra-u8-to-rgb-u8-n-for-images/67938/12?u=cad97
#[rustfmt::skip]
pub fn convert_raw(bgra: [u8; 16]) -> [u8; 12] {
    let [
        b0, g0, r0, _a0,
        b1, g1, r1, _a1,
        b2, g2, r2, _a2,
        b3, g3, r3, _a3,
    ] = bgra;
    [
        r0, g0, b0,
        r1, g1, b1,
        r2, g2, b2,
        r3, g3, b3,
    ]
}

#[rustfmt::skip]
pub fn convert_raw_r(bgra: &[u8; 16]) -> [u8; 12] {
    let [
        b0, g0, r0, _a0,
        b1, g1, r1, _a1,
        b2, g2, r2, _a2,
        b3, g3, r3, _a3,
    ] = *bgra;
    [
        r0, g0, b0,
        r1, g1, b1,
        r2, g2, b2,
        r3, g3, b3,
    ]
}

pub fn convert_scatter(bgra: [u8; 16]) -> [u8; 12] {
    use std::simd::Simd; //u8x16;
    let vector = Simd::from_array(bgra);
    let mut rgb = [0; 12];
    #[rustfmt::skip]
    let idxs = [
        2, 1, 0, !0,
        5, 4, 3, !0,
        8, 7, 6, !0,
        11, 10, 9, !0,
    ];
    vector.scatter(&mut rgb, Simd::from_array(idxs));
    rgb
}

pub fn convert_scatter_r(bgra: &[u8; 16]) -> [u8; 12] {
    use std::simd::Simd; //u8x16;
    let vector = Simd::from_array(*bgra);
    let mut rgb = [0; 12];
    #[rustfmt::skip]
    let idxs = [
        2, 1, 0, !0,
        5, 4, 3, !0,
        8, 7, 6, !0,
        11, 10, 9, !0,
    ];
    vector.scatter(&mut rgb, Simd::from_array(idxs));
    rgb
}

pub fn convert_swizzle(bgra: [u8; 16]) -> [u8; 12] {
    use std::simd::{simd_swizzle, Simd};
    let bgra = Simd::from_array(bgra);
    #[rustfmt::skip]
    const IDXS: [usize; 16] = [
        2, 1, 0,
        6, 5, 4,
        10, 9, 8,
        14, 13, 12,
        3, 7, 11, 15,
    ];
    let rgb = simd_swizzle!(bgra, IDXS);
    rgb.to_array().as_chunks().0[0]
}

pub fn convert_swizzle_r(bgra: &[u8; 16]) -> [u8; 12] {
    use std::simd::{simd_swizzle, Simd};
    let bgra = Simd::from_array(*bgra);
    #[rustfmt::skip]
    const IDXS: [usize; 16] = [
        2, 1, 0,
        6, 5, 4,
        10, 9, 8,
        14, 13, 12,
        3, 7, 11, 15,
    ];
    let rgb = simd_swizzle!(bgra, IDXS);
    rgb.to_array().as_chunks().0[0]
}

pub fn convert_shift(bgra: [u8; 16]) -> [u8; 12] {
    let bgra = bgra.as_chunks().0;
    let mut buffer = 0_u128;
    for bgra in bgra.iter().cloned().rev() {
        buffer <<= 24;
        buffer |= u128::from(u32::from_be_bytes(bgra) >> 8);
    }
    buffer.to_le_bytes().as_chunks().0[0]
}

pub fn convert_shift_r(bgra: &[u8; 16]) -> [u8; 12] {
    let bgra = bgra.as_chunks().0;
    let mut buffer = 0_u128;
    for bgra in bgra.iter().cloned().rev() {
        buffer <<= 24;
        buffer |= u128::from(u32::from_be_bytes(bgra) >> 8);
    }
    buffer.to_le_bytes().as_chunks().0[0]
}

pub fn convert_overwrite(bgra: [u8; 16]) -> [u8; 12] {
    let bgra = bgra.as_chunks().0;
    let mut padded_rgb = [0; 13];
    for i in (0..4).rev() {
        let pixel = u32::from_be_bytes(bgra[i]);
        padded_rgb[i * 3..][..4].copy_from_slice(&pixel.to_le_bytes());
    }

    padded_rgb[1..].as_chunks().0[0]
}

pub fn convert_overwrite_r(bgra: &[u8; 16]) -> [u8; 12] {
    let bgra = bgra.as_chunks().0;
    let mut padded_rgb = [0; 13];
    for i in (0..4).rev() {
        let pixel = u32::from_be_bytes(bgra[i]);
        padded_rgb[i * 3..][..4].copy_from_slice(&pixel.to_le_bytes());
    }

    padded_rgb[1..].as_chunks().0[0]
}