nihav-duck/src/codecs/vpcommon.rs

   1 use nihav_core::codecs::*;
   2 use nihav_core::codecs::blockdsp::*;
   3
   4 #[derive(Clone,Copy,Debug,PartialEq)]
   5 #[allow(dead_code)]
   6 pub enum VPMBType {
   7     Intra,
   8     InterNoMV,
   9     InterMV,
  10     InterNearest,
  11     InterNear,
  12     InterFourMV,
  13     GoldenNoMV,
  14     GoldenMV,
  15     GoldenNearest,
  16     GoldenNear,
  17 }
  18
  19 pub const VP_REF_INTER: u8 = 1;
  20 pub const VP_REF_GOLDEN: u8 = 2;
  21
  22 #[allow(dead_code)]
  23 impl VPMBType {
  24     pub fn is_intra(self) -> bool { self == VPMBType::Intra }
  25     pub fn get_ref_id(self) -> u8 {
  26         match self {
  27             VPMBType::Intra         => 0,
  28             VPMBType::InterNoMV     |
  29             VPMBType::InterMV       |
  30             VPMBType::InterNearest  |
  31             VPMBType::InterNear     |
  32             VPMBType::InterFourMV   => VP_REF_INTER,
  33             _                       => VP_REF_GOLDEN,
  34         }
  35     }
  36 }
  37
  38 impl Default for VPMBType {
  39     fn default() -> Self { VPMBType::Intra }
  40 }
  41
  42 #[derive(Default)]
  43 pub struct VPShuffler {
  44     lastframe: Option<NAVideoBufferRef<u8>>,
  45     goldframe: Option<NAVideoBufferRef<u8>>,
  46 }
  47
  48 impl VPShuffler {
  49     pub fn new() -> Self { VPShuffler { lastframe: None, goldframe: None } }
  50     pub fn clear(&mut self) { self.lastframe = None; self.goldframe = None; }
  51     pub fn add_frame(&mut self, buf: NAVideoBufferRef<u8>) {
  52         self.lastframe = Some(buf);
  53     }
  54     pub fn add_golden_frame(&mut self, buf: NAVideoBufferRef<u8>) {
  55         self.goldframe = Some(buf);
  56     }
  57     pub fn get_last(&mut self) -> Option<NAVideoBufferRef<u8>> {
  58         if let Some(ref frm) = self.lastframe {
  59             Some(frm.clone())
  60         } else {
  61             None
  62         }
  63     }
  64     pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
  65         if let Some(ref frm) = self.goldframe {
  66             Some(frm.clone())
  67         } else {
  68             None
  69         }
  70     }
  71     pub fn has_refs(&self) -> bool {
  72         self.lastframe.is_some()
  73     }
  74 }
  75
  76 pub const VP56_COEF_BASE: [i16; 6] = [ 5, 7, 11, 19, 35, 67 ];
  77 pub const VP56_COEF_ADD_PROBS: [[u8; 12]; 6] = [
  78     [ 159, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
  79     [ 165, 145, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
  80     [ 173, 148, 140, 128,   0,   0,   0,   0,   0,   0,   0,   0 ],
  81     [ 176, 155, 140, 135, 128,   0,   0,   0,   0,   0,   0,   0 ],
  82     [ 180, 157, 141, 134, 130, 128,   0,   0,   0,   0,   0,   0 ],
  83     [ 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 128 ],
  84 ];
  85
  86 #[allow(dead_code)]
  87 pub struct BoolCoder<'a> {
  88     pub src:    &'a [u8],
  89     pos:    usize,
  90     value:  u32,
  91     range:  u32,
  92     bits:   i32,
  93 }
  94
  95 #[allow(dead_code)]
  96 impl<'a> BoolCoder<'a> {
  97     pub fn new(src: &'a [u8]) -> DecoderResult<Self> {
  98         if src.len() < 3 { return Err(DecoderError::ShortData); }
  99         let value = ((src[0] as u32) << 24) | ((src[1] as u32) << 16) | ((src[2] as u32) << 8) | (src[3] as u32);
 100         Ok(Self { src, pos: 4, value, range: 255, bits: 8 })
 101     }
 102     pub fn read_bool(&mut self) -> bool {
 103         self.read_prob(128)
 104     }
 105     pub fn read_prob(&mut self, prob: u8) -> bool {
 106         self.renorm();
 107         let split = 1 + (((self.range - 1) * (prob as u32)) >> 8);
 108         let bit;
 109         if self.value < (split << 24) {
 110             self.range = split;
 111             bit = false;
 112         } else {
 113             self.range -= split;
 114             self.value -= split << 24;
 115             bit = true;
 116         }
 117         bit
 118     }
 119     pub fn read_bits(&mut self, bits: u8) -> u32 {
 120         let mut val = 0u32;
 121         for _ in 0..bits {
 122             val = (val << 1) | (self.read_prob(128) as u32);
 123         }
 124         val
 125     }
 126     pub fn read_byte(&mut self) -> u8 {
 127         let mut val = 0u8;
 128         for _ in 0..8 {
 129             val = (val << 1) | (self.read_prob(128) as u8);
 130         }
 131         val
 132     }
 133     pub fn read_sbits(&mut self, bits: u8) -> i32 {
 134         let mut val = if self.read_prob(128) { -1i32 } else { 0i32 };
 135         for _ in 1..bits {
 136             val = (val << 1) | (self.read_prob(128) as i32);
 137         }
 138         val
 139     }
 140     pub fn read_probability(&mut self) -> u8 {
 141         let val = self.read_bits(7) as u8;
 142         if val == 0 {
 143             1
 144         } else {
 145             val << 1
 146         }
 147     }
 148     fn renorm(&mut self) {
 149         let shift = self.range.leading_zeros() & 7;
 150         self.range <<= shift;
 151         self.value <<= shift;
 152         self.bits   -= shift as i32;
 153         if (self.bits <= 0) && (self.pos < self.src.len()) {
 154             self.value |= (self.src[self.pos] as u32) << (-self.bits as u8);
 155             self.pos += 1;
 156             self.bits += 8;
 157         }
 158 /*        while self.range < 0x80 {
 159             self.range <<= 1;
 160             self.value <<= 1;
 161             self.bits   -= 1;
 162             if (self.bits <= 0) && (self.pos < self.src.len()) {
 163                 self.value |= self.src[self.pos] as u32;
 164                 self.pos += 1;
 165                 self.bits = 8;
 166             }
 167         }*/
 168     }
 169     pub fn skip_bytes(&mut self, nbytes: usize) {
 170         for _ in 0..nbytes {
 171             self.value <<= 8;
 172             if self.pos < self.src.len() {
 173                 self.value |= self.src[self.pos] as u32;
 174                 self.pos += 1;
 175             }
 176         }
 177     }
 178 }
 179
 180 #[allow(dead_code)]
 181 pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
 182     ((((prob as i32) * (weights[0] as i32) + 128) >> 8) + (weights[1] as i32)).min(maxval).max(1) as u8
 183 }
 184
 185 #[macro_export]
 186 macro_rules! vp_tree {
 187     ($bc: expr, $prob: expr, $node1: expr, $node2: expr) => {
 188         if !$bc.read_prob($prob) {
 189             $node1
 190         } else {
 191             $node2
 192         }
 193     };
 194     ($leaf: expr) => { $leaf }
 195 }
 196
 197 const C1S7: i32 = 64277;
 198 const C2S6: i32 = 60547;
 199 const C3S5: i32 = 54491;
 200 const C4S4: i32 = 46341;
 201 const C5S3: i32 = 36410;
 202 const C6S2: i32 = 25080;
 203 const C7S1: i32 = 12785;
 204
 205 fn mul16(a: i32, b: i32) -> i32 {
 206     (a * b) >> 16
 207 }
 208
 209 macro_rules! idct_step {
 210     ($s0:expr, $s1:expr, $s2:expr, $s3:expr, $s4:expr, $s5:expr, $s6:expr, $s7:expr,
 211      $d0:expr, $d1:expr, $d2:expr, $d3:expr, $d4:expr, $d5:expr, $d6:expr, $d7:expr,
 212      $bias:expr, $shift:expr, $otype:ty) => {
 213         let t_a  = mul16(C1S7, i32::from($s1)) + mul16(C7S1, i32::from($s7));
 214         let t_b  = mul16(C7S1, i32::from($s1)) - mul16(C1S7, i32::from($s7));
 215         let t_c  = mul16(C3S5, i32::from($s3)) + mul16(C5S3, i32::from($s5));
 216         let t_d  = mul16(C3S5, i32::from($s5)) - mul16(C5S3, i32::from($s3));
 217         let t_a1 = mul16(C4S4, t_a - t_c);
 218         let t_b1 = mul16(C4S4, t_b - t_d);
 219         let t_c  = t_a + t_c;
 220         let t_d  = t_b + t_d;
 221         let t_e  = mul16(C4S4, i32::from($s0 + $s4)) + $bias;
 222         let t_f  = mul16(C4S4, i32::from($s0 - $s4)) + $bias;
 223         let t_g  = mul16(C2S6, i32::from($s2)) + mul16(C6S2, i32::from($s6));
 224         let t_h  = mul16(C6S2, i32::from($s2)) - mul16(C2S6, i32::from($s6));
 225         let t_e1 = t_e  - t_g;
 226         let t_g  = t_e  + t_g;
 227         let t_a  = t_f  + t_a1;
 228         let t_f  = t_f  - t_a1;
 229         let t_b  = t_b1 - t_h;
 230         let t_h  = t_b1 + t_h;
 231
 232         $d0 = ((t_g  + t_c) >> $shift) as $otype;
 233         $d7 = ((t_g  - t_c) >> $shift) as $otype;
 234         $d1 = ((t_a  + t_h) >> $shift) as $otype;
 235         $d2 = ((t_a  - t_h) >> $shift) as $otype;
 236         $d3 = ((t_e1 + t_d) >> $shift) as $otype;
 237         $d4 = ((t_e1 - t_d) >> $shift) as $otype;
 238         $d5 = ((t_f  + t_b) >> $shift) as $otype;
 239         $d6 = ((t_f  - t_b) >> $shift) as $otype;
 240     }
 241 }
 242
 243 pub fn vp_idct(coeffs: &mut [i16; 64]) {
 244     let mut tmp = [0i32; 64];
 245     for (src, dst) in coeffs.chunks(8).zip(tmp.chunks_mut(8)) {
 246         idct_step!(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
 247                    dst[0], dst[1], dst[2], dst[3], dst[4], dst[5], dst[6], dst[7], 0, 0, i32);
 248     }
 249     let src = &tmp;
 250     let dst = coeffs;
 251     for i in 0..8 {
 252         idct_step!(src[0 * 8 + i], src[1 * 8 + i], src[2 * 8 + i], src[3 * 8 + i],
 253                    src[4 * 8 + i], src[5 * 8 + i], src[6 * 8 + i], src[7 * 8 + i],
 254                    dst[0 * 8 + i], dst[1 * 8 + i], dst[2 * 8 + i], dst[3 * 8 + i],
 255                    dst[4 * 8 + i], dst[5 * 8 + i], dst[6 * 8 + i], dst[7 * 8 + i], 8, 4, i16);
 256     }
 257 }
 258
 259 pub fn vp_idct_dc(coeffs: &mut [i16; 64]) {
 260     let dc = ((mul16(C4S4, mul16(C4S4, i32::from(coeffs[0]))) + 8) >> 4) as i16;
 261     for i in 0..64 {
 262         coeffs[i] = dc;
 263     }
 264 }
 265
 266 pub fn unquant(coeffs: &mut [i16; 64], qmat: &[i16; 64]) {
 267     for i in 1..64 {
 268         coeffs[i] = coeffs[i].wrapping_mul(qmat[i]);
 269     }
 270 }
 271
 272 pub fn vp_put_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 273     vp_idct(coeffs);
 274     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 275     for y in 0..8 {
 276         for x in 0..8 {
 277             frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
 278         }
 279         off += frm.stride[plane];
 280     }
 281 }
 282
 283 pub fn vp_put_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 284     vp_idct(coeffs);
 285     let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
 286     for y in 0..8 {
 287         for x in 0..8 {
 288             frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
 289         }
 290         off += frm.stride[plane] * 2;
 291     }
 292 }
 293
 294 pub fn vp_put_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 295     vp_idct_dc(coeffs);
 296     let dc = (coeffs[0] + 128).min(255).max(0) as u8;
 297     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 298     for _ in 0..8 {
 299         for x in 0..8 {
 300             frm.data[off + x] = dc;
 301         }
 302         off += frm.stride[plane];
 303     }
 304 }
 305
 306 pub fn vp_add_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 307     vp_idct(coeffs);
 308     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 309     for y in 0..8 {
 310         for x in 0..8 {
 311             frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
 312         }
 313         off += frm.stride[plane];
 314     }
 315 }
 316
 317 pub fn vp_add_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 318     vp_idct(coeffs);
 319     let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
 320     for y in 0..8 {
 321         for x in 0..8 {
 322             frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
 323         }
 324         off += frm.stride[plane] * 2;
 325     }
 326 }
 327
 328 pub fn vp_add_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 329     vp_idct_dc(coeffs);
 330     let dc = coeffs[0];
 331     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 332     for _ in 0..8 {
 333         for x in 0..8 {
 334             frm.data[off + x] = (dc + (frm.data[off + x] as i16)).min(255).max(0) as u8;
 335         }
 336         off += frm.stride[plane];
 337     }
 338 }
 339
 340 pub fn vp31_loop_filter(data: &mut [u8], mut off: usize, step: usize, stride: usize,
 341                         len: usize, loop_str: i16) {
 342     for _ in 0..len {
 343         let a = data[off - step * 2] as i16;
 344         let b = data[off - step] as i16;
 345         let c = data[off] as i16;
 346         let d = data[off + step] as i16;
 347         let mut diff = ((a - d) + 3 * (c - b) + 4) >> 3;
 348         if diff.abs() >= 2 * loop_str {
 349             diff = 0;
 350         } else if diff.abs() >= loop_str {
 351             if diff < 0 {
 352                 diff = -diff - 2 * loop_str;
 353             } else {
 354                 diff = -diff + 2 * loop_str;
 355             }
 356         }
 357         if diff != 0 {
 358             data[off - step] = (b + diff).max(0).min(255) as u8;
 359             data[off]        = (c - diff).max(0).min(255) as u8;
 360         }
 361
 362         off += stride;
 363     }
 364 }
 365
 366 pub fn vp_copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, comp: usize,
 367                      dx: usize, dy: usize, mv_x: i16, mv_y: i16,
 368                      preborder: usize, postborder: usize, loop_str: i16,
 369                      mode: usize, interp: &[BlkInterpFunc], mut mc_buf: NAVideoBufferRef<u8>)
 370 {
 371     let sx = (dx as isize) + (mv_x as isize);
 372     let sy = (dy as isize) + (mv_y as isize);
 373     if ((sx | sy) & 7) == 0 {
 374         copy_block(dst, src, comp, dx, dy, mv_x, mv_y, 8, 8, preborder, postborder, mode, interp);
 375         return;
 376     }
 377     let pre = preborder.max(2);
 378     let post = postborder.max(1);
 379     let bsize = 8 + pre + post;
 380     let src_x = sx - (pre as isize);
 381     let src_y = sy - (pre as isize);
 382     {
 383         let tmp_buf = NASimpleVideoFrame::from_video_buf(&mut mc_buf).unwrap();
 384         edge_emu(src.as_ref(), src_x, src_y, bsize, bsize, &mut tmp_buf.data[tmp_buf.offset[comp]..], tmp_buf.stride[comp], comp);
 385 //        copy_block(&mut tmp_buf, src, comp, 0, 0, src_x as i16, src_y as i16,
 386 //                   bsize, bsize, 0, 0, 0, interp);
 387         if (sx & 7) != 0 {
 388             let foff = (8 - (sx & 7)) as usize;
 389             let off = pre + foff + tmp_buf.offset[comp];
 390             vp31_loop_filter(tmp_buf.data, off, 1, tmp_buf.stride[comp], bsize, loop_str);
 391         }
 392         if (sy & 7) != 0 {
 393             let foff = (8 - (sy & 7)) as usize;
 394             let off = (pre + foff) * tmp_buf.stride[comp] + tmp_buf.offset[comp];
 395             vp31_loop_filter(tmp_buf.data, off, tmp_buf.stride[comp], 1, bsize, loop_str);
 396         }
 397     }
 398     let dxoff = (pre as i16) - (dx as i16);
 399     let dyoff = (pre as i16) - (dy as i16);
 400     copy_block(dst, mc_buf, comp, dx, dy, dxoff, dyoff, 8, 8, preborder, postborder, mode, interp);
 401 }
 402
 403 fn vp3_interp00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 404 {
 405     let mut didx = 0;
 406     let mut sidx = 0;
 407     for _ in 0..bh {
 408         for x in 0..bw { dst[didx + x] = src[sidx + x]; }
 409         didx += dstride;
 410         sidx += sstride;
 411     }
 412 }
 413
 414 fn vp3_interp01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 415 {
 416     let mut didx = 0;
 417     let mut sidx = 0;
 418     for _ in 0..bh {
 419         for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + 1] as u16)) >> 1) as u8; }
 420         didx += dstride;
 421         sidx += sstride;
 422     }
 423 }
 424
 425 fn vp3_interp10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 426 {
 427     let mut didx = 0;
 428     let mut sidx = 0;
 429     for _ in 0..bh {
 430         for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + sstride] as u16)) >> 1) as u8; }
 431         didx += dstride;
 432         sidx += sstride;
 433     }
 434 }
 435
 436 fn vp3_interp1x(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 437 {
 438     let mut didx = 0;
 439     let mut sidx = 0;
 440     for _ in 0..bh {
 441         for x in 0..bw {
 442             dst[didx + x] = (((src[sidx + x] as u16) +
 443                               (src[sidx + x + sstride + 1] as u16)) >> 1) as u8;
 444         }
 445         didx += dstride;
 446         sidx += sstride;
 447     }
 448 }
 449
 450 fn vp3_interp1y(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 451 {
 452     let mut didx = 0;
 453     let mut sidx = 0;
 454     for _ in 0..bh {
 455         for x in 0..bw {
 456             dst[didx + x] = (((src[sidx + x + 1] as u16) +
 457                               (src[sidx + x + sstride] as u16)) >> 1) as u8;
 458         }
 459         didx += dstride;
 460         sidx += sstride;
 461     }
 462 }
 463
 464 pub const VP3_INTERP_FUNCS: &[blockdsp::BlkInterpFunc] = &[ vp3_interp00, vp3_interp01, vp3_interp10, vp3_interp1x, vp3_interp1y ];
 465