use std::arch::aarch64::*;
#[inline(always)]
fn load_u16x4_i32(p: &[u16]) -> int32x4_t {
unsafe { vreinterpretq_s32_u32(vmovl_u16(vld1_u16(p.as_ptr()))) }
}
#[inline]
#[target_feature(enable = "neon")]
fn round_s32(v: int32x4_t, rnd: i32, shift: i32) -> int32x4_t {
vshlq_s32(vaddq_s32(v, vdupq_n_s32(rnd)), vdupq_n_s32(-shift))
}
#[inline]
#[target_feature(enable = "neon")]
fn store_clip_u16x4(dst: &mut [u16], v: int32x4_t, rnd: i32, shift: i32, max: uint16x4_t) {
let v = round_s32(v, rnd, shift);
let p = vmin_u16(vqmovun_s32(v), max);
unsafe {
vst1_u16(dst.as_mut_ptr(), p);
}
}
#[inline(always)]
fn store_i16x4(dst: &mut [i16], v: int32x4_t, rnd: i32, shift: i32, bias: i32) {
unsafe {
let v = vsubq_s32(round_s32(v, rnd, shift), vdupq_n_s32(bias));
vst1_s16(dst.as_mut_ptr(), vqmovn_s32(v));
}
}
#[inline(always)]
fn load_u16x4(p: &[u16]) -> uint16x4_t {
unsafe { vld1_u16(p.as_ptr()) }
}
#[inline(always)]
fn load_i16x4(p: &[i16]) -> int16x4_t {
unsafe { vld1_s16(p.as_ptr()) }
}
#[inline(always)]
fn load_u16x8_i32(p: &[u16]) -> (int32x4_t, int32x4_t) {
unsafe {
let v = vld1q_u16(p.as_ptr());
(
vreinterpretq_s32_u32(vmovl_u16(vget_low_u16(v))),
vreinterpretq_s32_u32(vmovl_u16(vget_high_u16(v))),
)
}
}
#[inline(always)]
fn load_u16x8(p: &[u16]) -> uint16x8_t {
unsafe { vld1q_u16(p.as_ptr()) }
}
#[inline(always)]
fn load_i16x8(p: &[i16]) -> int16x8_t {
unsafe { vld1q_s16(p.as_ptr()) }
}
#[inline]
#[target_feature(enable = "neon")]
fn store_clip_u16x8(
dst: &mut [u16],
lo: int32x4_t,
hi: int32x4_t,
rnd: i32,
shift: i32,
max: uint16x8_t,
) {
let lo = round_s32(lo, rnd, shift);
let hi = round_s32(hi, rnd, shift);
let p = vminq_u16(vcombine_u16(vqmovun_s32(lo), vqmovun_s32(hi)), max);
unsafe { vst1q_u16(dst.as_mut_ptr(), p) };
}
#[inline]
#[target_feature(enable = "neon")]
fn store_i16x8_from_i32(
dst: &mut [i16],
lo: int32x4_t,
hi: int32x4_t,
rnd: i32,
shift: i32,
bias: i32,
) {
let bias = vdupq_n_s32(bias);
let lo = vqmovn_s32(vsubq_s32(round_s32(lo, rnd, shift), bias));
let hi = vqmovn_s32(vsubq_s32(round_s32(hi, rnd, shift), bias));
unsafe { vst1q_s16(dst.as_mut_ptr(), vcombine_s16(lo, hi)) };
}
#[inline]
#[target_feature(enable = "neon")]
fn madd_i16x4_pair_s32(acc: int32x4_t, a: int16x4_t, b: int16x4_t, c0: i16, c1: i16) -> int32x4_t {
vmlal_n_s16(vmlal_n_s16(acc, a, c0), b, c1)
}
#[inline]
#[target_feature(enable = "neon")]
fn filter_u16x4(src: &[u16], base: usize, stride: isize, f: &[i8; 8]) -> int32x4_t {
static OFFSETS: [isize; 8] = [-3isize, -2, -1, 0, 1, 2, 3, 4];
let mut sum = vdupq_n_s32(0);
for k in (0..8).step_by(2) {
let c0 = f[k];
let c1 = f[k + 1];
if c0 == 0 && c1 == 0 {
continue;
}
let idx0 = (base as isize + OFFSETS[k] * stride) as usize;
let idx1 = (base as isize + OFFSETS[k + 1] * stride) as usize;
let a = vreinterpret_s16_u16(load_u16x4(unsafe { src.get_unchecked(idx0..) }));
let b = vreinterpret_s16_u16(load_u16x4(unsafe { src.get_unchecked(idx1..) }));
sum = madd_i16x4_pair_s32(sum, a, b, c0 as i16, c1 as i16);
}
sum
}
#[inline]
#[target_feature(enable = "neon")]
fn filter_i16x4(src: &[i16], base: usize, stride: isize, f: &[i8; 8]) -> int32x4_t {
static OFFSETS: [isize; 8] = [-3isize, -2, -1, 0, 1, 2, 3, 4];
let mut sum = vdupq_n_s32(0);
for k in (0..8).step_by(2) {
let c0 = f[k];
let c1 = f[k + 1];
if c0 == 0 && c1 == 0 {
continue;
}
let idx0 = (base as isize + OFFSETS[k] * stride) as usize;
let idx1 = (base as isize + OFFSETS[k + 1] * stride) as usize;
let a = load_i16x4(unsafe { src.get_unchecked(idx0..) });
let b = load_i16x4(unsafe { src.get_unchecked(idx1..) });
sum = madd_i16x4_pair_s32(sum, a, b, c0 as i16, c1 as i16);
}
sum
}
#[inline]
#[target_feature(enable = "neon")]
fn filter_u16x8(src: &[u16], base: usize, stride: isize, f: &[i8; 8]) -> (int32x4_t, int32x4_t) {
static OFFSETS: [isize; 8] = [-3isize, -2, -1, 0, 1, 2, 3, 4];
let mut lo = vdupq_n_s32(0);
let mut hi = vdupq_n_s32(0);
for k in (0..8).step_by(2) {
let c0 = f[k];
let c1 = f[k + 1];
if c0 == 0 && c1 == 0 {
continue;
}
let idx0 = (base as isize + OFFSETS[k] * stride) as usize;
let idx1 = (base as isize + OFFSETS[k + 1] * stride) as usize;
let a = vreinterpretq_s16_u16(load_u16x8(unsafe { src.get_unchecked(idx0..) }));
let b = vreinterpretq_s16_u16(load_u16x8(unsafe { src.get_unchecked(idx1..) }));
lo = madd_i16x4_pair_s32(lo, vget_low_s16(a), vget_low_s16(b), c0 as i16, c1 as i16);
hi = madd_i16x4_pair_s32(hi, vget_high_s16(a), vget_high_s16(b), c0 as i16, c1 as i16);
}
(lo, hi)
}
#[inline]
#[target_feature(enable = "neon")]
fn filter_i16x8(src: &[i16], base: usize, stride: isize, f: &[i8; 8]) -> (int32x4_t, int32x4_t) {
static OFFSETS: [isize; 8] = [-3isize, -2, -1, 0, 1, 2, 3, 4];
let mut lo = vdupq_n_s32(0);
let mut hi = vdupq_n_s32(0);
for k in (0..8).step_by(2) {
let c0 = f[k];
let c1 = f[k + 1];
if c0 == 0 && c1 == 0 {
continue;
}
let idx0 = (base as isize + OFFSETS[k] * stride) as usize;
let idx1 = (base as isize + OFFSETS[k + 1] * stride) as usize;
let a = load_i16x8(unsafe { src.get_unchecked(idx0..) });
let b = load_i16x8(unsafe { src.get_unchecked(idx1..) });
lo = madd_i16x4_pair_s32(lo, vget_low_s16(a), vget_low_s16(b), c0 as i16, c1 as i16);
hi = madd_i16x4_pair_s32(hi, vget_high_s16(a), vget_high_s16(b), c0 as i16, c1 as i16);
}
(lo, hi)
}
#[inline(always)]
fn filter_u16_scalar(src: &[u16], base: usize, stride: isize, f: &[i8; 8]) -> i32 {
let c = base as isize;
f[0] as i32 * src[(c - 3 * stride) as usize] as i32
+ f[1] as i32 * src[(c - 2 * stride) as usize] as i32
+ f[2] as i32 * src[(c - stride) as usize] as i32
+ f[3] as i32 * src[base] as i32
+ f[4] as i32 * src[(c + stride) as usize] as i32
+ f[5] as i32 * src[(c + 2 * stride) as usize] as i32
+ f[6] as i32 * src[(c + 3 * stride) as usize] as i32
+ f[7] as i32 * src[(c + 4 * stride) as usize] as i32
}
#[inline(always)]
fn filter_i16_scalar(src: &[i16], base: usize, stride: isize, f: &[i8; 8]) -> i32 {
let c = base as isize;
f[0] as i32 * src[(c - 3 * stride) as usize] as i32
+ f[1] as i32 * src[(c - 2 * stride) as usize] as i32
+ f[2] as i32 * src[(c - stride) as usize] as i32
+ f[3] as i32 * src[base] as i32
+ f[4] as i32 * src[(c + stride) as usize] as i32
+ f[5] as i32 * src[(c + 2 * stride) as usize] as i32
+ f[6] as i32 * src[(c + 3 * stride) as usize] as i32
+ f[7] as i32 * src[(c + 4 * stride) as usize] as i32
}
#[inline(always)]
fn clip(v: i32, bitdepth: u8) -> u16 {
v.clamp(0, (1 << bitdepth) - 1) as u16
}
#[inline(always)]
fn round_scalar(v: i32, rnd: i32, shift: i32) -> i32 {
if shift == 0 {
v + rnd
} else {
(v + rnd) >> shift
}
}
#[inline]
#[target_feature(enable = "neon")]
fn bilin_u16x4(src: &[u16], base: usize, stride: usize, mxy: i32) -> int32x4_t {
unsafe {
let a16 = load_u16x4(src.get_unchecked(base..));
let b16 = load_u16x4(src.get_unchecked(base + stride..));
let a = vreinterpretq_s32_u32(vmovl_u16(a16));
let diff = vsub_s16(vreinterpret_s16_u16(b16), vreinterpret_s16_u16(a16));
vmlal_n_s16(vshlq_n_s32::<4>(a), diff, mxy as i16)
}
}
#[inline]
#[target_feature(enable = "neon")]
fn bilin_i16x4(a16: int16x4_t, b16: int16x4_t, mxy: i32) -> int32x4_t {
let a = vmovl_s16(a16);
let diff = vsub_s16(b16, a16);
vmlal_n_s16(vshlq_n_s32::<4>(a), diff, mxy as i16)
}
#[inline]
#[target_feature(enable = "neon")]
fn bilin_u16x8(src: &[u16], base: usize, stride: usize, mxy: i32) -> (int32x4_t, int32x4_t) {
unsafe {
let a16 = load_u16x8(src.get_unchecked(base..));
let b16 = load_u16x8(src.get_unchecked(base + stride..));
let alo = vreinterpretq_s32_u32(vmovl_u16(vget_low_u16(a16)));
let ahi = vreinterpretq_s32_u32(vmovl_u16(vget_high_u16(a16)));
let dlo = vsub_s16(
vreinterpret_s16_u16(vget_low_u16(b16)),
vreinterpret_s16_u16(vget_low_u16(a16)),
);
let dhi = vsub_s16(
vreinterpret_s16_u16(vget_high_u16(b16)),
vreinterpret_s16_u16(vget_high_u16(a16)),
);
(
vmlal_n_s16(vshlq_n_s32::<4>(alo), dlo, mxy as i16),
vmlal_n_s16(vshlq_n_s32::<4>(ahi), dhi, mxy as i16),
)
}
}
#[inline]
#[target_feature(enable = "neon")]
fn bilin_i16x8(a16: int16x8_t, b16: int16x8_t, mxy: i32) -> (int32x4_t, int32x4_t) {
let alo = vmovl_s16(vget_low_s16(a16));
let ahi = vmovl_s16(vget_high_s16(a16));
let dlo = vsub_s16(vget_low_s16(b16), vget_low_s16(a16));
let dhi = vsub_s16(vget_high_s16(b16), vget_high_s16(a16));
(
vmlal_n_s16(vshlq_n_s32::<4>(alo), dlo, mxy as i16),
vmlal_n_s16(vshlq_n_s32::<4>(ahi), dhi, mxy as i16),
)
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn prep_hbd_neon(
tmp: &mut [i16],
tmp_stride: usize,
src: &[u16],
src_stride: usize,
w: usize,
h: usize,
bitdepth: u8,
) {
let ib = 14 - bitdepth as i32;
let bias = 8192i32;
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = load_u16x8_i32(unsafe { src_row.get_unchecked(x..) });
store_i16x8_from_i32(
tmp_chunk,
vshlq_s32(lo, vdupq_n_s32(ib)),
vshlq_s32(hi, vdupq_n_s32(ib)),
0,
0,
bias,
);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
unsafe {
let s = load_u16x4_i32(src_row.get_unchecked(x..));
let v = vsubq_s32(vshlq_s32(s, vdupq_n_s32(ib)), vdupq_n_s32(bias));
vst1_s16(tmp_chunk.as_mut_ptr(), vqmovn_s32(v));
}
}
let processed = x8_done + tmp_chunks.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
*tmp_px = (((src_row[x] as i32) << ib) - bias) as i16;
}
}
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn put_bilin_hbd_neon(
dst: &mut [u16],
dst_stride: usize,
src: &[u16],
src_stride: usize,
w: usize,
h: usize,
mx: i32,
my: i32,
bitdepth: u8,
mid_scratch: &mut [i16],
) {
let ib = 14 - bitdepth as i32;
let maxv8 = vdupq_n_u16(((1 << bitdepth) - 1) as u16);
let maxv = vget_low_u16(maxv8);
let intermediate_rnd = (1 << ib) >> 1;
if mx != 0 && my != 0 {
let mid_stride = w.next_multiple_of(16).max(64);
let mid = &mut mid_scratch[..mid_stride * (h + 1)];
let sh0 = 4 - ib;
let rnd0 = if sh0 == 0 { 0 } else { 1 << (sh0 - 1) };
for (y, mid_row) in mid.chunks_mut(mid_stride).take(h + 1).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let (mid_chunks8, mid_rem8) = mid_row[..w].as_chunks_mut::<8>();
for (chunk_idx, mid_chunk) in mid_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = bilin_u16x8(src_row, x, 1, mx);
store_i16x8_from_i32(mid_chunk, lo, hi, rnd0, sh0, 0);
}
let x8_done = mid_chunks8.len() * 8;
let (mid_chunks, mid_rem) = mid_rem8.as_chunks_mut::<4>();
for (chunk_idx, mid_chunk) in mid_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(mid_chunk, bilin_u16x4(src_row, x, 1, mx), rnd0, sh0, 0);
}
let processed = x8_done + mid_chunks.len() * 4;
for (x, mid_px) in (processed..w).zip(mid_rem.iter_mut()) {
let a = src_row[x] as i32;
let b = src_row[x + 1] as i32;
*mid_px = round_scalar(16 * a + mx * (b - a), rnd0, sh0) as i16;
}
}
for (y, dst_row) in dst.chunks_mut(dst_stride).take(h).enumerate() {
let mid_row = unsafe { mid.get_unchecked(y * mid_stride..) };
let mid_next_row = unsafe { mid.get_unchecked((y + 1) * mid_stride..) };
let (dst_chunks8, dst_rem8) = dst_row[..w].as_chunks_mut::<8>();
for (chunk_idx, dst_chunk) in dst_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let a = load_i16x8(unsafe { mid_row.get_unchecked(x..) });
let b = load_i16x8(unsafe { mid_next_row.get_unchecked(x..) });
let (lo, hi) = bilin_i16x8(a, b, my);
store_clip_u16x8(dst_chunk, lo, hi, 1 << (3 + ib), 4 + ib, maxv8);
}
let x8_done = dst_chunks8.len() * 8;
let (dst_chunks, dst_rem) = dst_rem8.as_chunks_mut::<4>();
for (chunk_idx, dst_chunk) in dst_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
let a = load_i16x4(unsafe { mid_row.get_unchecked(x..) });
let b = load_i16x4(unsafe { mid_next_row.get_unchecked(x..) });
let v = bilin_i16x4(a, b, my);
store_clip_u16x4(dst_chunk, v, 1 << (3 + ib), 4 + ib, maxv);
}
let processed = x8_done + dst_chunks.len() * 4;
for (x, dst_px) in (processed..w).zip(dst_rem.iter_mut()) {
let a = mid_row[x] as i32;
let b = mid_next_row[x] as i32;
*dst_px = clip(
round_scalar(16 * a + my * (b - a), 1 << (3 + ib), 4 + ib),
bitdepth,
);
}
}
} else if mx != 0 {
let sh0 = 4 - ib;
let rnd0 = if sh0 == 0 { 0 } else { 1 << (sh0 - 1) };
for (y, dst_row) in dst.chunks_mut(dst_stride).take(h).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let (dst_chunks8, dst_rem8) = dst_row[..w].as_chunks_mut::<8>();
for (chunk_idx, dst_chunk) in dst_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = bilin_u16x8(src_row, x, 1, mx);
store_clip_u16x8(
dst_chunk,
round_s32(lo, rnd0, sh0),
round_s32(hi, rnd0, sh0),
intermediate_rnd,
ib,
maxv8,
);
}
let x8_done = dst_chunks8.len() * 8;
let (dst_chunks, dst_rem) = dst_rem8.as_chunks_mut::<4>();
for (chunk_idx, dst_chunk) in dst_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
let px = round_s32(bilin_u16x4(src_row, x, 1, mx), rnd0, sh0);
store_clip_u16x4(dst_chunk, px, intermediate_rnd, ib, maxv);
}
let processed = x8_done + dst_chunks.len() * 4;
for (x, dst_px) in (processed..w).zip(dst_rem.iter_mut()) {
let a = src_row[x] as i32;
let b = src_row[x + 1] as i32;
let px = round_scalar(16 * a + mx * (b - a), rnd0, sh0);
*dst_px = clip(round_scalar(px, intermediate_rnd, ib), bitdepth);
}
}
} else if my != 0 {
for (y, dst_row) in dst.chunks_mut(dst_stride).take(h).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let src_next_row = unsafe { src.get_unchecked((y + 1) * src_stride..) };
let (dst_chunks8, dst_rem8) = dst_row[..w].as_chunks_mut::<8>();
for (chunk_idx, dst_chunk) in dst_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = bilin_u16x8(src_row, x, src_stride, my);
store_clip_u16x8(dst_chunk, lo, hi, 8, 4, maxv8);
}
let x8_done = dst_chunks8.len() * 8;
let (dst_chunks, dst_rem) = dst_rem8.as_chunks_mut::<4>();
for (chunk_idx, dst_chunk) in dst_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_clip_u16x4(
dst_chunk,
bilin_u16x4(src_row, x, src_stride, my),
8,
4,
maxv,
);
}
let processed = x8_done + dst_chunks.len() * 4;
for (x, dst_px) in (processed..w).zip(dst_rem.iter_mut()) {
let a = src_row[x] as i32;
let b = src_next_row[x] as i32;
*dst_px = clip(round_scalar(16 * a + my * (b - a), 8, 4), bitdepth);
}
}
} else {
for (src_row, dst_row) in src
.chunks(src_stride)
.zip(dst.chunks_mut(dst_stride))
.take(h)
{
dst_row[..w].copy_from_slice(&src_row[..w]);
}
}
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn prep_bilin_hbd_neon(
tmp: &mut [i16],
tmp_stride: usize,
src: &[u16],
src_stride: usize,
w: usize,
h: usize,
mx: i32,
my: i32,
bitdepth: u8,
mid_scratch: &mut [i16],
) {
let ib = 14 - bitdepth as i32;
let bias = 8192i32;
if mx != 0 && my != 0 {
let mid_stride = w.next_multiple_of(16).max(64);
let mid = &mut mid_scratch[..mid_stride * (h + 1)];
let sh0 = 4 - ib;
let rnd0 = if sh0 == 0 { 0 } else { 1 << (sh0 - 1) };
for (y, mid_row) in mid.chunks_mut(mid_stride).take(h + 1).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let (mid_chunks8, mid_rem8) = mid_row[..w].as_chunks_mut::<8>();
for (chunk_idx, mid_chunk) in mid_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = bilin_u16x8(src_row, x, 1, mx);
store_i16x8_from_i32(mid_chunk, lo, hi, rnd0, sh0, 0);
}
let x8_done = mid_chunks8.len() * 8;
let (mid_chunks, mid_rem) = mid_rem8.as_chunks_mut::<4>();
for (chunk_idx, mid_chunk) in mid_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(mid_chunk, bilin_u16x4(src_row, x, 1, mx), rnd0, sh0, 0);
}
let processed = x8_done + mid_chunks.len() * 4;
for (x, mid_px) in (processed..w).zip(mid_rem.iter_mut()) {
let a = src_row[x] as i32;
let b = src_row[x + 1] as i32;
*mid_px = round_scalar(16 * a + mx * (b - a), rnd0, sh0) as i16;
}
}
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let mid_row = unsafe { mid.get_unchecked(y * mid_stride..) };
let mid_next_row = unsafe { mid.get_unchecked((y + 1) * mid_stride..) };
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let a = load_i16x8(unsafe { mid_row.get_unchecked(x..) });
let b = load_i16x8(unsafe { mid_next_row.get_unchecked(x..) });
let (lo, hi) = bilin_i16x8(a, b, my);
store_i16x8_from_i32(tmp_chunk, lo, hi, 8, 4, bias);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
let a = load_i16x4(unsafe { mid_row.get_unchecked(x..) });
let b = load_i16x4(unsafe { mid_next_row.get_unchecked(x..) });
let v = bilin_i16x4(a, b, my);
store_i16x4(tmp_chunk, v, 8, 4, bias);
}
let processed = x8_done + tmp_chunks.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
let a = mid_row[x] as i32;
let b = mid_next_row[x] as i32;
*tmp_px = (round_scalar(16 * a + my * (b - a), 8, 4) - bias) as i16;
}
}
} else if mx != 0 {
let sh0 = 4 - ib;
let rnd0 = if sh0 == 0 { 0 } else { 1 << (sh0 - 1) };
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = bilin_u16x8(src_row, x, 1, mx);
store_i16x8_from_i32(tmp_chunk, lo, hi, rnd0, sh0, bias);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(tmp_chunk, bilin_u16x4(src_row, x, 1, mx), rnd0, sh0, bias);
}
let processed = x8_done + tmp_chunks.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
let a = src_row[x] as i32;
let b = src_row[x + 1] as i32;
*tmp_px = (round_scalar(16 * a + mx * (b - a), rnd0, sh0) - bias) as i16;
}
}
} else if my != 0 {
let sh0 = 4 - ib;
let rnd0 = if sh0 == 0 { 0 } else { 1 << (sh0 - 1) };
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let src_row = unsafe { src.get_unchecked(y * src_stride..) };
let src_next_row = unsafe { src.get_unchecked((y + 1) * src_stride..) };
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = bilin_u16x8(src_row, x, src_stride, my);
store_i16x8_from_i32(tmp_chunk, lo, hi, rnd0, sh0, bias);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(
tmp_chunk,
bilin_u16x4(src_row, x, src_stride, my),
rnd0,
sh0,
bias,
);
}
let processed = x8_done + tmp_chunks.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
let a = src_row[x] as i32;
let b = src_next_row[x] as i32;
*tmp_px = (round_scalar(16 * a + my * (b - a), rnd0, sh0) - bias) as i16;
}
}
} else {
prep_hbd_neon(tmp, tmp_stride, src, src_stride, w, h, bitdepth);
}
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn put_8tap_hbd_neon(
dst: &mut [u16],
dst_stride: usize,
src: &[u16],
src_off: usize,
src_stride: usize,
w: usize,
h: usize,
mx: i32,
my: i32,
filter_type: i32,
bitdepth: u8,
mid_scratch: &mut [i16],
) {
let bits = 6 + (filter_type < 0) as i32;
let ib = 14 - bitdepth as i32;
let intermediate_rnd = ((1 << bits) >> 1) + ((1 << (bits - ib)) >> 1);
let fh = crate::mc::get_h_filter(mx, filter_type, w);
let fv = crate::mc::get_v_filter(my, filter_type, h);
let maxv8 = vdupq_n_u16(((1 << bitdepth) - 1) as u16);
let maxv = vget_low_u16(maxv8);
match (fh, fv) {
(Some(fh), Some(fv)) => {
let tmp_h = h + 7;
let mid_stride = w.next_multiple_of(16).max(64);
let mid = &mut mid_scratch[..mid_stride * tmp_h];
let sh0 = bits - ib;
let rnd0 = (1 << sh0) >> 1;
for (y, mid_row) in mid.chunks_mut(mid_stride).take(tmp_h).enumerate() {
let base = (src_off as isize + (y as isize - 3) * src_stride as isize) as usize;
let (mid_chunks8, mid_rem8) = mid_row[..w].as_chunks_mut::<8>();
for (chunk_idx, mid_chunk) in mid_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = filter_u16x8(src, base + x, 1, &fh);
store_i16x8_from_i32(mid_chunk, lo, hi, rnd0, sh0, 0);
}
let x8_done = mid_chunks8.len() * 8;
let (mid_chunks4, mid_rem) = mid_rem8.as_chunks_mut::<4>();
for (chunk_idx, mid_chunk) in mid_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(mid_chunk, filter_u16x4(src, base + x, 1, &fh), rnd0, sh0, 0);
}
let processed = x8_done + mid_chunks4.len() * 4;
for (x, mid_px) in (processed..w).zip(mid_rem.iter_mut()) {
*mid_px =
round_scalar(filter_u16_scalar(src, base + x, 1, &fh), rnd0, sh0) as i16;
}
}
let sh1 = bits + ib;
let rnd1 = (1 << sh1) >> 1;
for (y, dst_row) in dst.chunks_mut(dst_stride).take(h).enumerate() {
let (dst_chunks8, dst_rem8) = dst_row[..w].as_chunks_mut::<8>();
for (chunk_idx, dst_chunk) in dst_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) =
filter_i16x8(&mid, (y + 3) * mid_stride + x, mid_stride as isize, &fv);
store_clip_u16x8(dst_chunk, lo, hi, rnd1, sh1, maxv8);
}
let x8_done = dst_chunks8.len() * 8;
let (dst_chunks4, dst_rem) = dst_rem8.as_chunks_mut::<4>();
for (chunk_idx, dst_chunk) in dst_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_clip_u16x4(
dst_chunk,
filter_i16x4(&mid, (y + 3) * mid_stride + x, mid_stride as isize, &fv),
rnd1,
sh1,
maxv,
);
}
let processed = x8_done + dst_chunks4.len() * 4;
for (x, dst_px) in (processed..w).zip(dst_rem.iter_mut()) {
*dst_px = clip(
round_scalar(
filter_i16_scalar(
&mid,
(y + 3) * mid_stride + x,
mid_stride as isize,
&fv,
),
rnd1,
sh1,
),
bitdepth,
);
}
}
}
(Some(fh), None) => {
for (y, dst_row) in dst.chunks_mut(dst_stride).take(h).enumerate() {
let base = src_off + y * src_stride;
let (dst_chunks8, dst_rem8) = dst_row[..w].as_chunks_mut::<8>();
for (chunk_idx, dst_chunk) in dst_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = filter_u16x8(src, base + x, 1, &fh);
store_clip_u16x8(dst_chunk, lo, hi, intermediate_rnd, bits, maxv8);
}
let x8_done = dst_chunks8.len() * 8;
let (dst_chunks4, dst_rem) = dst_rem8.as_chunks_mut::<4>();
for (chunk_idx, dst_chunk) in dst_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_clip_u16x4(
dst_chunk,
filter_u16x4(src, base + x, 1, &fh),
intermediate_rnd,
bits,
maxv,
);
}
let processed = x8_done + dst_chunks4.len() * 4;
for (x, dst_px) in (processed..w).zip(dst_rem.iter_mut()) {
*dst_px = clip(
round_scalar(
filter_u16_scalar(src, base + x, 1, &fh),
intermediate_rnd,
bits,
),
bitdepth,
);
}
}
}
(None, Some(fv)) => {
let ss = src_stride as isize;
for (y, dst_row) in dst.chunks_mut(dst_stride).take(h).enumerate() {
let base = src_off + y * src_stride;
let (dst_chunks8, dst_rem8) = dst_row[..w].as_chunks_mut::<8>();
for (chunk_idx, dst_chunk) in dst_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = filter_u16x8(src, base + x, ss, &fv);
store_clip_u16x8(dst_chunk, lo, hi, (1 << bits) >> 1, bits, maxv8);
}
let x8_done = dst_chunks8.len() * 8;
let (dst_chunks4, dst_rem) = dst_rem8.as_chunks_mut::<4>();
for (chunk_idx, dst_chunk) in dst_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_clip_u16x4(
dst_chunk,
filter_u16x4(src, base + x, ss, &fv),
(1 << bits) >> 1,
bits,
maxv,
);
}
let processed = x8_done + dst_chunks4.len() * 4;
for (x, dst_px) in (processed..w).zip(dst_rem.iter_mut()) {
*dst_px = clip(
round_scalar(
filter_u16_scalar(src, base + x, ss, &fv),
(1 << bits) >> 1,
bits,
),
bitdepth,
);
}
}
}
(None, None) => {
for (src_row, dst_row) in src[src_off..]
.chunks(src_stride)
.zip(dst.chunks_mut(dst_stride))
.take(h)
{
dst_row[..w].copy_from_slice(&src_row[..w]);
}
}
}
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn prep_8tap_hbd_neon(
tmp: &mut [i16],
tmp_stride: usize,
src: &[u16],
src_off: usize,
src_stride: usize,
w: usize,
h: usize,
mx: i32,
my: i32,
filter_type: i32,
bitdepth: u8,
mid_scratch: &mut [i16],
) {
let bits = 6 + (filter_type < 0) as i32;
let ib = 14 - bitdepth as i32;
let bias = 8192i32;
let fh = crate::mc::get_h_filter(mx, filter_type, w);
let fv = crate::mc::get_v_filter(my, filter_type, h);
match (fh, fv) {
(Some(fh), Some(fv)) => {
let tmp_h = h + 7;
let mid_stride = w.next_multiple_of(16).max(64);
let mid = &mut mid_scratch[..mid_stride * tmp_h];
let sh0 = bits - ib;
let rnd0 = (1 << sh0) >> 1;
for (y, mid_row) in mid.chunks_mut(mid_stride).take(tmp_h).enumerate() {
let base = (src_off as isize + (y as isize - 3) * src_stride as isize) as usize;
let (mid_chunks8, mid_rem8) = mid_row[..w].as_chunks_mut::<8>();
for (chunk_idx, mid_chunk) in mid_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = filter_u16x8(src, base + x, 1, &fh);
store_i16x8_from_i32(mid_chunk, lo, hi, rnd0, sh0, 0);
}
let x8_done = mid_chunks8.len() * 8;
let (mid_chunks4, mid_rem) = mid_rem8.as_chunks_mut::<4>();
for (chunk_idx, mid_chunk) in mid_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(mid_chunk, filter_u16x4(src, base + x, 1, &fh), rnd0, sh0, 0);
}
let processed = x8_done + mid_chunks4.len() * 4;
for (x, mid_px) in (processed..w).zip(mid_rem.iter_mut()) {
*mid_px =
round_scalar(filter_u16_scalar(src, base + x, 1, &fh), rnd0, sh0) as i16;
}
}
let rnd1 = (1 << bits) >> 1;
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) =
filter_i16x8(&mid, (y + 3) * mid_stride + x, mid_stride as isize, &fv);
store_i16x8_from_i32(tmp_chunk, lo, hi, rnd1, bits, bias);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks4, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(
tmp_chunk,
filter_i16x4(&mid, (y + 3) * mid_stride + x, mid_stride as isize, &fv),
rnd1,
bits,
bias,
);
}
let processed = x8_done + tmp_chunks4.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
*tmp_px = (round_scalar(
filter_i16_scalar(&mid, (y + 3) * mid_stride + x, mid_stride as isize, &fv),
rnd1,
bits,
) - bias) as i16;
}
}
}
(Some(fh), None) => {
let sh0 = bits - ib;
let rnd0 = (1 << sh0) >> 1;
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let base = src_off + y * src_stride;
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = filter_u16x8(src, base + x, 1, &fh);
store_i16x8_from_i32(tmp_chunk, lo, hi, rnd0, sh0, bias);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks4, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(
tmp_chunk,
filter_u16x4(src, base + x, 1, &fh),
rnd0,
sh0,
bias,
);
}
let processed = x8_done + tmp_chunks4.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
*tmp_px = (round_scalar(filter_u16_scalar(src, base + x, 1, &fh), rnd0, sh0)
- bias) as i16;
}
}
}
(None, Some(fv)) => {
let ss = src_stride as isize;
let sh0 = bits - ib;
let rnd0 = (1 << sh0) >> 1;
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(h).enumerate() {
let base = src_off + y * src_stride;
let (tmp_chunks8, tmp_rem8) = tmp_row[..w].as_chunks_mut::<8>();
for (chunk_idx, tmp_chunk) in tmp_chunks8.iter_mut().enumerate() {
let x = chunk_idx * 8;
let (lo, hi) = filter_u16x8(src, base + x, ss, &fv);
store_i16x8_from_i32(tmp_chunk, lo, hi, rnd0, sh0, bias);
}
let x8_done = tmp_chunks8.len() * 8;
let (tmp_chunks4, tmp_rem) = tmp_rem8.as_chunks_mut::<4>();
for (chunk_idx, tmp_chunk) in tmp_chunks4.iter_mut().enumerate() {
let x = x8_done + chunk_idx * 4;
store_i16x4(
tmp_chunk,
filter_u16x4(src, base + x, ss, &fv),
rnd0,
sh0,
bias,
);
}
let processed = x8_done + tmp_chunks4.len() * 4;
for (x, tmp_px) in (processed..w).zip(tmp_rem.iter_mut()) {
*tmp_px = (round_scalar(filter_u16_scalar(src, base + x, ss, &fv), rnd0, sh0)
- bias) as i16;
}
}
}
(None, None) => prep_hbd_neon(tmp, tmp_stride, &src[src_off..], src_stride, w, h, bitdepth),
}
}
#[inline]
#[target_feature(enable = "neon")]
fn load_u16x8_i32_warp(src: &[u16]) -> (int32x4_t, int32x4_t) {
let v = unsafe { vld1q_u16(src.as_ptr()) };
(
vreinterpretq_s32_u32(vmovl_u16(vget_low_u16(v))),
vreinterpretq_s32_u32(vmovl_u16(vget_high_u16(v))),
)
}
#[inline]
#[target_feature(enable = "neon")]
fn load_i16x8_i32_warp(src: &[i16]) -> (int32x4_t, int32x4_t) {
let v = unsafe { vld1q_s16(src.as_ptr()) };
(vmovl_s16(vget_low_s16(v)), vmovl_s16(vget_high_s16(v)))
}
#[inline]
#[target_feature(enable = "neon")]
fn warp_coeff_i32x4(pos: i32, step: i32, tap: usize, lane_base: i32) -> int32x4_t {
let f0 =
&crate::tables::MC_WARP_FILTER[(192 + ((pos + step * lane_base + 512) >> 10)) as usize];
let f1 = &crate::tables::MC_WARP_FILTER
[(192 + ((pos + step * (lane_base + 1) + 512) >> 10)) as usize];
let f2 = &crate::tables::MC_WARP_FILTER
[(192 + ((pos + step * (lane_base + 2) + 512) >> 10)) as usize];
let f3 = &crate::tables::MC_WARP_FILTER
[(192 + ((pos + step * (lane_base + 3) + 512) >> 10)) as usize];
unsafe {
vld1q_s32(
[
f0[tap] as i32,
f1[tap] as i32,
f2[tap] as i32,
f3[tap] as i32,
]
.as_ptr(),
)
}
}
#[inline]
#[target_feature(enable = "neon")]
fn warp_horz_u16x8(
src: &[u16],
row_base: usize,
mx: i32,
alpha: i32,
rnd: i32,
shift: i32,
) -> int16x8_t {
let mut lo = vdupq_n_s32(0);
let mut hi = vdupq_n_s32(0);
for tap in 0..8 {
let (px_lo, px_hi) = load_u16x8_i32_warp(unsafe { src.get_unchecked(row_base + tap..) });
lo = vmlaq_s32(lo, px_lo, warp_coeff_i32x4(mx, alpha, tap, 0));
hi = vmlaq_s32(hi, px_hi, warp_coeff_i32x4(mx, alpha, tap, 4));
}
let lo = round_s32(lo, rnd, shift);
let hi = round_s32(hi, rnd, shift);
vcombine_s16(vqmovn_s32(lo), vqmovn_s32(hi))
}
#[inline]
#[target_feature(enable = "neon")]
fn warp_vert_i16x8(
mid: &[i16],
base: usize,
stride: usize,
my: i32,
gamma: i32,
) -> (int32x4_t, int32x4_t) {
let mut lo = vdupq_n_s32(0);
let mut hi = vdupq_n_s32(0);
for tap in 0..8 {
let (px_lo, px_hi) =
load_i16x8_i32_warp(unsafe { mid.get_unchecked(base + tap * stride..) });
lo = vmlaq_s32(lo, px_lo, warp_coeff_i32x4(my, gamma, tap, 0));
hi = vmlaq_s32(hi, px_hi, warp_coeff_i32x4(my, gamma, tap, 4));
}
(lo, hi)
}
#[inline]
#[target_feature(enable = "neon")]
fn store_clip_u16x8_warp(
dst: &mut [u16],
lo: int32x4_t,
hi: int32x4_t,
rnd: i32,
shift: i32,
max: uint16x8_t,
) {
let lo = round_s32(lo, rnd, shift);
let hi = round_s32(hi, rnd, shift);
let p = vminq_u16(vcombine_u16(vqmovun_s32(lo), vqmovun_s32(hi)), max);
unsafe { vst1q_u16(dst.as_mut_ptr(), p) };
}
#[inline]
#[target_feature(enable = "neon")]
fn store_i16x8_warp(
dst: &mut [i16],
lo: int32x4_t,
hi: int32x4_t,
rnd: i32,
shift: i32,
bias: i32,
) {
let lo = vsubq_s32(round_s32(lo, rnd, shift), vdupq_n_s32(bias));
let hi = vsubq_s32(round_s32(hi, rnd, shift), vdupq_n_s32(bias));
unsafe {
vst1q_s16(
dst.as_mut_ptr(),
vcombine_s16(vqmovn_s32(lo), vqmovn_s32(hi)),
)
};
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn warp_affine_8x8_hbd_neon(
dst: &mut [u16],
dst_stride: usize,
src: &[u16],
src_stride: usize,
src_off: usize,
abcd: &[i16; 4],
mut mx: i32,
mut my: i32,
bitdepth: u8,
) {
let ib = 14 - bitdepth as i32;
let h_shift = 7 - ib;
let h_rnd = (1 << h_shift) >> 1;
let v_shift = 7 + ib;
let v_rnd = (1 << v_shift) >> 1;
let max = vdupq_n_u16(((1 << bitdepth) - 1) as u16);
let alpha = abcd[0] as i32;
let beta = abcd[1] as i32;
let gamma = abcd[2] as i32;
let delta = abcd[3] as i32;
let mut mid = [0i16; 15 * 8];
let mut row_base = src_off.wrapping_sub(3 * src_stride + 3);
for mid_row in mid.as_chunks_mut::<8>().0.iter_mut() {
unsafe {
vst1q_s16(
mid_row.as_mut_ptr(),
warp_horz_u16x8(src, row_base, mx, alpha, h_rnd, h_shift),
)
};
row_base += src_stride;
mx += beta;
}
for (y, dst_row) in dst.chunks_mut(dst_stride).take(8).enumerate() {
let (lo, hi) = warp_vert_i16x8(&mid, y * 8, 8, my, gamma);
store_clip_u16x8_warp(&mut dst_row[..8], lo, hi, v_rnd, v_shift, max);
my += delta;
}
}
#[allow(clippy::too_many_arguments)]
#[target_feature(enable = "neon")]
pub(crate) fn warp_affine_8x8t_hbd_neon(
tmp: &mut [i16],
tmp_stride: usize,
src: &[u16],
src_stride: usize,
src_off: usize,
abcd: &[i16; 4],
mut mx: i32,
mut my: i32,
bitdepth: u8,
) {
let ib = 14 - bitdepth as i32;
let h_shift = 7 - ib;
let h_rnd = (1 << h_shift) >> 1;
let alpha = abcd[0] as i32;
let beta = abcd[1] as i32;
let gamma = abcd[2] as i32;
let delta = abcd[3] as i32;
let mut mid = [0i16; 15 * 8];
let mut row_base = src_off.wrapping_sub(3 * src_stride + 3);
for mid_row in mid.as_chunks_mut::<8>().0.iter_mut() {
unsafe {
vst1q_s16(
mid_row.as_mut_ptr(),
warp_horz_u16x8(src, row_base, mx, alpha, h_rnd, h_shift),
)
};
row_base += src_stride;
mx += beta;
}
for (y, tmp_row) in tmp.chunks_mut(tmp_stride).take(8).enumerate() {
let (lo, hi) = warp_vert_i16x8(&mid, y * 8, 8, my, gamma);
store_i16x8_warp(&mut tmp_row[..8], lo, hi, 64, 7, 8192);
my += delta;
}
}