#[inline(always)]
fn load_u32<const BE: bool>(v: u32) -> u32 {
if BE { u32::from_be(v) } else { u32::from_le(v) }
}
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbr32_to_rgb_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
rgb_out: &mut [u8],
width: usize,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(rgb_out.len() >= width * 3, "rgb_out row too short");
for x in 0..width {
let dst = x * 3;
rgb_out[dst] = (load_u32::<BE>(r[x]) >> 24) as u8;
rgb_out[dst + 1] = (load_u32::<BE>(g[x]) >> 24) as u8;
rgb_out[dst + 2] = (load_u32::<BE>(b[x]) >> 24) as u8;
}
}
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbr32_to_rgb_u16_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
rgb_u16_out: &mut [u16],
width: usize,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(rgb_u16_out.len() >= width * 3, "rgb_u16_out row too short");
for x in 0..width {
let dst = x * 3;
rgb_u16_out[dst] = (load_u32::<BE>(r[x]) >> 16) as u16;
rgb_u16_out[dst + 1] = (load_u32::<BE>(g[x]) >> 16) as u16;
rgb_u16_out[dst + 2] = (load_u32::<BE>(b[x]) >> 16) as u16;
}
}
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbra32_to_rgba_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
a: &[u32],
rgba_out: &mut [u8],
width: usize,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(a.len() >= width, "a row too short");
debug_assert!(rgba_out.len() >= width * 4, "rgba_out row too short");
for x in 0..width {
let dst = x * 4;
rgba_out[dst] = (load_u32::<BE>(r[x]) >> 24) as u8;
rgba_out[dst + 1] = (load_u32::<BE>(g[x]) >> 24) as u8;
rgba_out[dst + 2] = (load_u32::<BE>(b[x]) >> 24) as u8;
rgba_out[dst + 3] = (load_u32::<BE>(a[x]) >> 24) as u8;
}
}
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbra32_to_rgba_u16_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
a: &[u32],
rgba_u16_out: &mut [u16],
width: usize,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(a.len() >= width, "a row too short");
debug_assert!(
rgba_u16_out.len() >= width * 4,
"rgba_u16_out row too short"
);
for x in 0..width {
let dst = x * 4;
rgba_u16_out[dst] = (load_u32::<BE>(r[x]) >> 16) as u16;
rgba_u16_out[dst + 1] = (load_u32::<BE>(g[x]) >> 16) as u16;
rgba_u16_out[dst + 2] = (load_u32::<BE>(b[x]) >> 16) as u16;
rgba_u16_out[dst + 3] = (load_u32::<BE>(a[x]) >> 16) as u16;
}
}
#[cfg(any(feature = "std", feature = "alloc"))]
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbr32_to_rgb_u32_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
rgb_u32_out: &mut [u32],
width: usize,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(rgb_u32_out.len() >= width * 3, "rgb_u32_out row too short");
for x in 0..width {
let dst = x * 3;
rgb_u32_out[dst] = load_u32::<BE>(r[x]);
rgb_u32_out[dst + 1] = load_u32::<BE>(g[x]);
rgb_u32_out[dst + 2] = load_u32::<BE>(b[x]);
}
}
#[cfg(any(feature = "std", feature = "alloc"))]
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbra32_to_rgba_u32_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
a: &[u32],
rgba_u32_out: &mut [u32],
width: usize,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(a.len() >= width, "a row too short");
debug_assert!(
rgba_u32_out.len() >= width * 4,
"rgba_u32_out row too short"
);
for x in 0..width {
let dst = x * 4;
rgba_u32_out[dst] = load_u32::<BE>(r[x]);
rgba_u32_out[dst + 1] = load_u32::<BE>(g[x]);
rgba_u32_out[dst + 2] = load_u32::<BE>(b[x]);
rgba_u32_out[dst + 3] = load_u32::<BE>(a[x]);
}
}
#[cfg_attr(not(tarpaulin), inline(always))]
pub(crate) fn gbr32_to_luma_u16_row<const BE: bool>(
g: &[u32],
b: &[u32],
r: &[u32],
luma_out: &mut [u16],
width: usize,
matrix: crate::KernelMatrix,
full_range: bool,
) {
debug_assert!(g.len() >= width, "g row too short");
debug_assert!(b.len() >= width, "b row too short");
debug_assert!(r.len() >= width, "r row too short");
debug_assert!(luma_out.len() >= width, "luma row too short");
let (k_r, k_g, k_b) = super::luma_coefficients_q15(matrix);
let k_r = k_r as i64;
let k_g = k_g as i64;
let k_b = k_b as i64;
const RND: i64 = 1 << 14;
const NATIVE_MAX: i64 = 65535;
if full_range {
for x in 0..width {
let rv = (load_u32::<BE>(r[x]) >> 16) as i64;
let gv = (load_u32::<BE>(g[x]) >> 16) as i64;
let bv = (load_u32::<BE>(b[x]) >> 16) as i64;
let y = (k_r * rv + k_g * gv + k_b * bv + RND) >> 15;
luma_out[x] = y.clamp(0, NATIVE_MAX) as u16;
}
} else {
const Y_OFF: i64 = 16 << 8;
const RANGE: i64 = 219 << 8;
const Y_MAX: i64 = 235 << 8;
const Y_MIN: i64 = Y_OFF;
for x in 0..width {
let rv = (load_u32::<BE>(r[x]) >> 16) as i64;
let gv = (load_u32::<BE>(g[x]) >> 16) as i64;
let bv = (load_u32::<BE>(b[x]) >> 16) as i64;
let y_full = (k_r * rv + k_g * gv + k_b * bv + RND) >> 15;
let y_full_clamped = y_full.clamp(0, NATIVE_MAX);
let y_lim = Y_OFF + (y_full_clamped * RANGE + NATIVE_MAX / 2) / NATIVE_MAX;
luma_out[x] = y_lim.clamp(Y_MIN, Y_MAX) as u16;
}
}
}
#[cfg(all(test, feature = "std"))]
mod tests;