| 1 | use nihav_codec_support::codecs::blockdsp::*; |
| 2 | |
| 3 | use super::clip_u8; |
| 4 | |
| 5 | const TMP_BUF_STRIDE: usize = 32; |
| 6 | |
| 7 | fn interp_block1(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool, avg0: bool) { |
| 8 | |
| 9 | let step = if hor { 1 } else { sstride }; |
| 10 | let mut idx = 0; |
| 11 | let avgidx = if avg0 { step * 2 } else { step * 3 }; |
| 12 | |
| 13 | for dline in dst.chunks_mut(dstride).take(h) { |
| 14 | for (x, pix) in dline.iter_mut().take(w).enumerate() { |
| 15 | let t = clip_u8(( i16::from(src[idx + x]) |
| 16 | - 5 * i16::from(src[idx + x + step]) |
| 17 | + 20 * i16::from(src[idx + x + step * 2]) |
| 18 | + 20 * i16::from(src[idx + x + step * 3]) |
| 19 | - 5 * i16::from(src[idx + x + step * 4]) |
| 20 | + i16::from(src[idx + x + step * 5]) |
| 21 | + 16) >> 5); |
| 22 | *pix = ((u16::from(t) + u16::from(src[idx + x + avgidx]) + 1) >> 1) as u8; |
| 23 | } |
| 24 | idx += sstride; |
| 25 | } |
| 26 | } |
| 27 | |
| 28 | fn interp_block2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool) { |
| 29 | let step = if hor { 1 } else { sstride }; |
| 30 | let mut idx = 0; |
| 31 | for dline in dst.chunks_mut(dstride).take(h) { |
| 32 | for (x, pix) in dline.iter_mut().take(w).enumerate() { |
| 33 | *pix = clip_u8(( i16::from(src[idx + x]) |
| 34 | - 5 * i16::from(src[idx + x + step]) |
| 35 | + 20 * i16::from(src[idx + x + step * 2]) |
| 36 | + 20 * i16::from(src[idx + x + step * 3]) |
| 37 | - 5 * i16::from(src[idx + x + step * 4]) |
| 38 | + i16::from(src[idx + x + step * 5]) |
| 39 | + 16) >> 5); |
| 40 | } |
| 41 | idx += sstride; |
| 42 | } |
| 43 | } |
| 44 | |
| 45 | fn mc_avg_tmp(dst: &mut [u8], dstride: usize, w: usize, h: usize, tmp: &[u8], tmp2: &[u8]) { |
| 46 | for (dline, (sline0, sline1)) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE).zip(tmp2.chunks(TMP_BUF_STRIDE))).take(h) { |
| 47 | for (pix, (&a, &b)) in dline.iter_mut().zip(sline0.iter().zip(sline1.iter())).take(w) { |
| 48 | *pix = ((u16::from(a) + u16::from(b) + 1) >> 1) as u8; |
| 49 | } |
| 50 | } |
| 51 | } |
| 52 | |
| 53 | fn h264_mc00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 54 | for (dline, sline) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) { |
| 55 | dline[..w].copy_from_slice(&sline[..w]); |
| 56 | } |
| 57 | } |
| 58 | |
| 59 | fn h264_mc01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 60 | interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, true); |
| 61 | } |
| 62 | |
| 63 | fn h264_mc02(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 64 | interp_block2(dst, dstride, &src[sstride * 2..], sstride, w, h, true); |
| 65 | } |
| 66 | |
| 67 | fn h264_mc03(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 68 | interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, false); |
| 69 | } |
| 70 | |
| 71 | fn h264_mc10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 72 | interp_block1(dst, dstride, &src[2..], sstride, w, h, false, true); |
| 73 | } |
| 74 | |
| 75 | fn h264_mc11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 76 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 77 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 78 | h264_mc02(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 79 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h); |
| 80 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 81 | } |
| 82 | |
| 83 | fn h264_mc12(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 84 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 85 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 86 | h264_mc02(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 87 | h264_mc22(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h); |
| 88 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 89 | } |
| 90 | |
| 91 | fn h264_mc13(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 92 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 93 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 94 | h264_mc02(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 95 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h); |
| 96 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 97 | } |
| 98 | |
| 99 | fn h264_mc20(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 100 | interp_block2(dst, dstride, &src[2..], sstride, w, h, false); |
| 101 | } |
| 102 | |
| 103 | fn h264_mc21(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 104 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 105 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 106 | h264_mc22(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 107 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h); |
| 108 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 109 | } |
| 110 | |
| 111 | fn h264_mc22(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 112 | let mut tmp = [0i32; TMP_BUF_STRIDE * 16]; |
| 113 | let mut idx = 0; |
| 114 | for dline in tmp.chunks_mut(TMP_BUF_STRIDE).take(h) { |
| 115 | for (x, pix) in dline.iter_mut().take(w + 5).enumerate() { |
| 116 | *pix = i32::from(src[idx + x]) |
| 117 | - 5 * i32::from(src[idx + x + sstride]) |
| 118 | + 20 * i32::from(src[idx + x + sstride * 2]) |
| 119 | + 20 * i32::from(src[idx + x + sstride * 3]) |
| 120 | - 5 * i32::from(src[idx + x + sstride * 4]) |
| 121 | + i32::from(src[idx + x + sstride * 5]); |
| 122 | } |
| 123 | idx += sstride; |
| 124 | } |
| 125 | for (dline, sline) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE)).take(h) { |
| 126 | for (x, pix) in dline.iter_mut().take(w).enumerate() { |
| 127 | *pix = clip_u8(((sline[x] - 5 * sline[x + 1] + 20 * sline[x + 2] + 20 * sline[x + 3] - 5 * sline[x + 4] + sline[x + 5] + 512) >> 10) as i16); |
| 128 | } |
| 129 | } |
| 130 | } |
| 131 | |
| 132 | fn h264_mc23(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 133 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 134 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 135 | h264_mc22(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 136 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h); |
| 137 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 138 | } |
| 139 | |
| 140 | fn h264_mc30(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 141 | interp_block1(dst, dstride, &src[2..], sstride, w, h, false, false); |
| 142 | } |
| 143 | |
| 144 | fn h264_mc31(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 145 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 146 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 147 | h264_mc20(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 148 | h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h); |
| 149 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 150 | } |
| 151 | |
| 152 | fn h264_mc32(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 153 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 154 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 155 | h264_mc22(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 156 | h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h); |
| 157 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 158 | } |
| 159 | |
| 160 | fn h264_mc33(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 161 | let mut tmp = [0u8; TMP_BUF_STRIDE * 16]; |
| 162 | let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16]; |
| 163 | h264_mc20(&mut tmp, TMP_BUF_STRIDE, &src[1..], sstride, w, h); |
| 164 | h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h); |
| 165 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 166 | } |
| 167 | |
| 168 | |
| 169 | pub fn chroma_interp(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, w: usize, h: usize) { |
| 170 | let a0 = 8 - dx; |
| 171 | let a1 = dx; |
| 172 | let b0 = 8 - dy; |
| 173 | let b1 = dy; |
| 174 | |
| 175 | let src1 = &src[sstride..]; |
| 176 | if a0 == 8 && b0 == 8 { |
| 177 | for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) { |
| 178 | drow[..w].copy_from_slice(&line[..w]); |
| 179 | } |
| 180 | } else if a0 == 8 { |
| 181 | for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) { |
| 182 | for (pix, (&a, &b)) in drow.iter_mut().take(w).zip(line0.iter().zip(line1.iter())) { |
| 183 | *pix = ((u16::from(a) * b0 + u16::from(b) * b1 + 4) >> 3) as u8; |
| 184 | } |
| 185 | } |
| 186 | } else if b0 == 8 { |
| 187 | for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) { |
| 188 | let mut a = line[0]; |
| 189 | for (pix, &b) in drow.iter_mut().take(w).zip(line.iter().skip(1)) { |
| 190 | *pix = ((u16::from(a) * a0 + u16::from(b) * a1 + 4) >> 3) as u8; |
| 191 | a = b; |
| 192 | } |
| 193 | } |
| 194 | } else { |
| 195 | for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) { |
| 196 | let mut a = line0[0]; |
| 197 | let mut c = line1[0]; |
| 198 | for (pix, (&b, &d)) in drow.iter_mut().take(w).zip(line0[1..].iter().zip(line1[1..].iter())) { |
| 199 | *pix = ((u16::from(a) * a0 * b0 + u16::from(b) * a1 * b0 + u16::from(c) * a0 * b1 + u16::from(d) * a1 * b1 + 0x20) >> 6) as u8; |
| 200 | a = b; |
| 201 | c = d; |
| 202 | } |
| 203 | } |
| 204 | } |
| 205 | } |
| 206 | |
| 207 | pub const H264_LUMA_INTERP: &[BlkInterpFunc] = &[ |
| 208 | h264_mc00, h264_mc01, h264_mc02, h264_mc03, |
| 209 | h264_mc10, h264_mc11, h264_mc12, h264_mc13, |
| 210 | h264_mc20, h264_mc21, h264_mc22, h264_mc23, |
| 211 | h264_mc30, h264_mc31, h264_mc32, h264_mc33 |
| 212 | ]; |