nihav-itu/src/codecs/h264/dsp/mc/debug.rs

   1 use super::clip_u8;
   2
   3 const TMP_BUF_STRIDE: usize = 32;
   4
   5 fn interp_block1(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool, avg0: bool) {
   6
   7     let step = if hor { 1 } else { sstride };
   8     let mut idx = 0;
   9     let avgidx = if avg0 { step * 2 } else { step * 3 };
  10
  11     for dline in dst.chunks_mut(dstride).take(h) {
  12         for (x, pix) in dline.iter_mut().take(w).enumerate() {
  13             let t = clip_u8((       i16::from(src[idx + x])
  14                              - 5  * i16::from(src[idx + x + step])
  15                              + 20 * i16::from(src[idx + x + step * 2])
  16                              + 20 * i16::from(src[idx + x + step * 3])
  17                              - 5  * i16::from(src[idx + x + step * 4])
  18                              +      i16::from(src[idx + x + step * 5])
  19                              + 16) >> 5);
  20             *pix = ((u16::from(t) + u16::from(src[idx + x + avgidx]) + 1) >> 1) as u8;
  21         }
  22         idx += sstride;
  23     }
  24 }
  25
  26 fn interp_block2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool) {
  27     let step = if hor { 1 } else { sstride };
  28     let mut idx = 0;
  29     for dline in dst.chunks_mut(dstride).take(h) {
  30         for (x, pix) in dline.iter_mut().take(w).enumerate() {
  31             *pix = clip_u8((       i16::from(src[idx + x])
  32                             - 5  * i16::from(src[idx + x + step])
  33                             + 20 * i16::from(src[idx + x + step * 2])
  34                             + 20 * i16::from(src[idx + x + step * 3])
  35                             - 5  * i16::from(src[idx + x + step * 4])
  36                             +      i16::from(src[idx + x + step * 5])
  37                             + 16) >> 5);
  38         }
  39         idx += sstride;
  40     }
  41 }
  42
  43 fn mc_avg_tmp(dst: &mut [u8], dstride: usize, w: usize, h: usize, tmp: &[u8], tmp2: &[u8]) {
  44     for (dline, (sline0, sline1)) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE).zip(tmp2.chunks(TMP_BUF_STRIDE))).take(h) {
  45         for (pix, (&a, &b)) in dline.iter_mut().zip(sline0.iter().zip(sline1.iter())).take(w) {
  46             *pix = ((u16::from(a) + u16::from(b) + 1) >> 1) as u8;
  47         }
  48     }
  49 }
  50
  51 fn h264_mc00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  52     for (dline, sline) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
  53         dline[..w].copy_from_slice(&sline[..w]);
  54     }
  55 }
  56
  57 fn h264_mc01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  58     interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, true);
  59 }
  60
  61 fn h264_mc02(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  62     interp_block2(dst, dstride, &src[sstride * 2..], sstride, w, h, true);
  63 }
  64
  65 fn h264_mc03(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  66     interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, false);
  67 }
  68
  69 fn h264_mc10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  70     interp_block1(dst, dstride, &src[2..], sstride, w, h, false, true);
  71 }
  72
  73 fn h264_mc11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  74     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
  75     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
  76     h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
  77     h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
  78     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
  79 }
  80
  81 fn h264_mc12(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  82     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
  83     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
  84     h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
  85     h264_mc22(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
  86     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
  87 }
  88
  89 fn h264_mc13(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  90     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
  91     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
  92     h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
  93     h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
  94     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
  95 }
  96
  97 fn h264_mc20(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
  98     interp_block2(dst, dstride, &src[2..], sstride, w, h, false);
  99 }
 100
 101 fn h264_mc21(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 102     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
 103     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
 104     h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
 105     h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
 106     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
 107 }
 108
 109 fn h264_mc22(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 110     let mut tmp = [0i32; TMP_BUF_STRIDE * 16];
 111     let mut idx = 0;
 112     for dline in tmp.chunks_mut(TMP_BUF_STRIDE).take(h) {
 113         for (x, pix) in dline.iter_mut().take(w + 5).enumerate() {
 114             *pix =        i32::from(src[idx + x])
 115                    - 5  * i32::from(src[idx + x + sstride])
 116                    + 20 * i32::from(src[idx + x + sstride * 2])
 117                    + 20 * i32::from(src[idx + x + sstride * 3])
 118                    - 5  * i32::from(src[idx + x + sstride * 4])
 119                    +      i32::from(src[idx + x + sstride * 5]);
 120         }
 121         idx += sstride;
 122     }
 123     for (dline, sline) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE)).take(h) {
 124         for (x, pix) in dline.iter_mut().take(w).enumerate() {
 125             *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);
 126         }
 127     }
 128 }
 129
 130 fn h264_mc23(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 131     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
 132     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
 133     h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
 134     h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
 135     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
 136 }
 137
 138 fn h264_mc30(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 139     interp_block1(dst, dstride, &src[2..], sstride, w, h, false, false);
 140 }
 141
 142 fn h264_mc31(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 143     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
 144     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
 145     h264_mc20(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
 146     h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
 147     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
 148 }
 149
 150 fn h264_mc32(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 151     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
 152     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
 153     h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
 154     h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
 155     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
 156 }
 157
 158 fn h264_mc33(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
 159     let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
 160     let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
 161     h264_mc20(&mut tmp,  TMP_BUF_STRIDE, &src[1..], sstride, w, h);
 162     h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
 163     mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
 164 }
 165
 166
 167 fn chroma_interp(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, w: usize, h: usize) {
 168     let a0 = 8 - dx;
 169     let a1 = dx;
 170     let b0 = 8 - dy;
 171     let b1 = dy;
 172
 173     let src1 = &src[sstride..];
 174     if a0 == 8 && b0 == 8 {
 175         for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
 176             drow[..w].copy_from_slice(&line[..w]);
 177         }
 178     } else if a0 == 8 {
 179         for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) {
 180             for (pix, (&a, &b)) in drow.iter_mut().take(w).zip(line0.iter().zip(line1.iter())) {
 181                 *pix = ((u16::from(a) * b0 + u16::from(b) * b1 + 4) >> 3) as u8;
 182             }
 183         }
 184     } else if b0 == 8 {
 185         for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
 186             let mut a = line[0];
 187             for (pix, &b) in drow.iter_mut().take(w).zip(line.iter().skip(1)) {
 188                 *pix = ((u16::from(a) * a0 + u16::from(b) * a1 + 4) >> 3) as u8;
 189                 a = b;
 190             }
 191         }
 192     } else {
 193         for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) {
 194             let mut a = line0[0];
 195             let mut c = line1[0];
 196             for (pix, (&b, &d)) in drow.iter_mut().take(w).zip(line0[1..].iter().zip(line1[1..].iter())) {
 197                 *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;
 198                 a = b;
 199                 c = d;
 200             }
 201         }
 202     }
 203 }
 204
 205 pub fn chroma_interp_8(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, h: usize) {
 206     chroma_interp(dst, dstride, src, sstride, dx, dy, 8, h);
 207 }
 208
 209 pub fn chroma_interp_4(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, h: usize) {
 210     chroma_interp(dst, dstride, src, sstride, dx, dy, 4, h);
 211 }
 212
 213 pub fn chroma_interp_2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, h: usize) {
 214     chroma_interp(dst, dstride, src, sstride, dx, dy, 2, h);
 215 }
 216
 217 macro_rules! luma_mc {
 218     ($orig:ident, $func4:ident, $func8:ident, $func16:ident) => {
 219         fn $func4(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, h: usize) {
 220             $orig(dst, dstride, src, sstride, 4, h);
 221         }
 222         fn $func8(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, h: usize) {
 223             $orig(dst, dstride, src, sstride, 8, h);
 224         }
 225         fn $func16(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, h: usize) {
 226             $orig(dst, dstride, src, sstride, 16, h);
 227         }
 228     }
 229 }
 230
 231 luma_mc!(h264_mc00, h264_mc00_4, h264_mc00_8, h264_mc00_16);
 232 luma_mc!(h264_mc01, h264_mc01_4, h264_mc01_8, h264_mc01_16);
 233 luma_mc!(h264_mc02, h264_mc02_4, h264_mc02_8, h264_mc02_16);
 234 luma_mc!(h264_mc03, h264_mc03_4, h264_mc03_8, h264_mc03_16);
 235 luma_mc!(h264_mc10, h264_mc10_4, h264_mc10_8, h264_mc10_16);
 236 luma_mc!(h264_mc11, h264_mc11_4, h264_mc11_8, h264_mc11_16);
 237 luma_mc!(h264_mc12, h264_mc12_4, h264_mc12_8, h264_mc12_16);
 238 luma_mc!(h264_mc13, h264_mc13_4, h264_mc13_8, h264_mc13_16);
 239 luma_mc!(h264_mc20, h264_mc20_4, h264_mc20_8, h264_mc20_16);
 240 luma_mc!(h264_mc21, h264_mc21_4, h264_mc21_8, h264_mc21_16);
 241 luma_mc!(h264_mc22, h264_mc22_4, h264_mc22_8, h264_mc22_16);
 242 luma_mc!(h264_mc23, h264_mc23_4, h264_mc23_8, h264_mc23_16);
 243 luma_mc!(h264_mc30, h264_mc30_4, h264_mc30_8, h264_mc30_16);
 244 luma_mc!(h264_mc31, h264_mc31_4, h264_mc31_8, h264_mc31_16);
 245 luma_mc!(h264_mc32, h264_mc32_4, h264_mc32_8, h264_mc32_16);
 246 luma_mc!(h264_mc33, h264_mc33_4, h264_mc33_8, h264_mc33_16);
 247
 248 pub const H264_LUMA_INTERP: &[[super::MCFunc; 16]; 3] = &[
 249   [
 250     h264_mc00_4, h264_mc01_4, h264_mc02_4, h264_mc03_4,
 251     h264_mc10_4, h264_mc11_4, h264_mc12_4, h264_mc13_4,
 252     h264_mc20_4, h264_mc21_4, h264_mc22_4, h264_mc23_4,
 253     h264_mc30_4, h264_mc31_4, h264_mc32_4, h264_mc33_4
 254   ], [
 255     h264_mc00_8, h264_mc01_8, h264_mc02_8, h264_mc03_8,
 256     h264_mc10_8, h264_mc11_8, h264_mc12_8, h264_mc13_8,
 257     h264_mc20_8, h264_mc21_8, h264_mc22_8, h264_mc23_8,
 258     h264_mc30_8, h264_mc31_8, h264_mc32_8, h264_mc33_8
 259   ], [
 260     h264_mc00_16, h264_mc01_16, h264_mc02_16, h264_mc03_16,
 261     h264_mc10_16, h264_mc11_16, h264_mc12_16, h264_mc13_16,
 262     h264_mc20_16, h264_mc21_16, h264_mc22_16, h264_mc23_16,
 263     h264_mc30_16, h264_mc31_16, h264_mc32_16, h264_mc33_16
 264   ]
 265 ];
 266
 267 impl super::RegisterSIMD for super::H264MC {
 268     fn register_simd(&mut self) {}
 269 }