codecs/vpcommon.rs

   1 use nihav_core::codecs::*;
   2 use nihav_core::codecs::blockdsp::*;
   3
   4 pub const VP_YUVA420_FORMAT: NAPixelFormaton = NAPixelFormaton{
   5         model:      ColorModel::YUV(YUVSubmodel::YUVJ),
   6         components: 4,
   7         comp_info:  [
   8                 Some(NAPixelChromaton{ h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 0, next_elem: 1}),
   9                 Some(NAPixelChromaton{ h_ss: 1, v_ss: 1, packed: false, depth: 8, shift: 0, comp_offs: 1, next_elem: 1}),
  10                 Some(NAPixelChromaton{ h_ss: 1, v_ss: 1, packed: false, depth: 8, shift: 0, comp_offs: 2, next_elem: 1}),
  11                 Some(NAPixelChromaton{ h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 3, next_elem: 1}),
  12                 None ],
  13         elem_size:  0,
  14         be:         false,
  15         alpha:      true,
  16         palette:    false
  17     };
  18
  19 #[derive(Clone,Copy,Debug,PartialEq)]
  20 #[allow(dead_code)]
  21 pub enum VPMBType {
  22     Intra,
  23     InterNoMV,
  24     InterMV,
  25     InterNearest,
  26     InterNear,
  27     InterFourMV,
  28     GoldenNoMV,
  29     GoldenMV,
  30     GoldenNearest,
  31     GoldenNear,
  32 }
  33
  34 pub const VP_REF_INTER: u8 = 1;
  35 pub const VP_REF_GOLDEN: u8 = 2;
  36
  37 #[allow(dead_code)]
  38 impl VPMBType {
  39     pub fn is_intra(self) -> bool { self == VPMBType::Intra }
  40     pub fn get_ref_id(self) -> u8 {
  41         match self {
  42             VPMBType::Intra         => 0,
  43             VPMBType::InterNoMV     |
  44             VPMBType::InterMV       |
  45             VPMBType::InterNearest  |
  46             VPMBType::InterNear     |
  47             VPMBType::InterFourMV   => VP_REF_INTER,
  48             _                       => VP_REF_GOLDEN,
  49         }
  50     }
  51 }
  52
  53 impl Default for VPMBType {
  54     fn default() -> Self { VPMBType::Intra }
  55 }
  56
  57 #[derive(Default)]
  58 pub struct VPShuffler {
  59     lastframe: Option<NAVideoBufferRef<u8>>,
  60     goldframe: Option<NAVideoBufferRef<u8>>,
  61 }
  62
  63 impl VPShuffler {
  64     pub fn new() -> Self { VPShuffler { lastframe: None, goldframe: None } }
  65     pub fn clear(&mut self) { self.lastframe = None; self.goldframe = None; }
  66     pub fn add_frame(&mut self, buf: NAVideoBufferRef<u8>) {
  67         self.lastframe = Some(buf);
  68     }
  69     pub fn add_golden_frame(&mut self, buf: NAVideoBufferRef<u8>) {
  70         self.goldframe = Some(buf);
  71     }
  72     pub fn get_last(&mut self) -> Option<NAVideoBufferRef<u8>> {
  73         if let Some(ref frm) = self.lastframe {
  74             Some(frm.clone())
  75         } else {
  76             None
  77         }
  78     }
  79     pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
  80         if let Some(ref frm) = self.goldframe {
  81             Some(frm.clone())
  82         } else {
  83             None
  84         }
  85     }
  86     pub fn has_refs(&self) -> bool {
  87         self.lastframe.is_some()
  88     }
  89 }
  90
  91 pub const VP56_COEF_BASE: [i16; 6] = [ 5, 7, 11, 19, 35, 67 ];
  92 pub const VP56_COEF_ADD_PROBS: [[u8; 12]; 6] = [
  93     [ 159, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
  94     [ 165, 145, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
  95     [ 173, 148, 140, 128,   0,   0,   0,   0,   0,   0,   0,   0 ],
  96     [ 176, 155, 140, 135, 128,   0,   0,   0,   0,   0,   0,   0 ],
  97     [ 180, 157, 141, 134, 130, 128,   0,   0,   0,   0,   0,   0 ],
  98     [ 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 128 ],
  99 ];
 100
 101 #[allow(dead_code)]
 102 pub struct BoolCoder<'a> {
 103     pub src:    &'a [u8],
 104     pos:    usize,
 105     value:  u32,
 106     range:  u32,
 107     bits:   i32,
 108 }
 109
 110 #[allow(dead_code)]
 111 impl<'a> BoolCoder<'a> {
 112     pub fn new(src: &'a [u8]) -> DecoderResult<Self> {
 113         if src.len() < 3 { return Err(DecoderError::ShortData); }
 114         let value = ((src[0] as u32) << 24) | ((src[1] as u32) << 16) | ((src[2] as u32) << 8) | (src[3] as u32);
 115         Ok(Self { src, pos: 4, value, range: 255, bits: 8 })
 116     }
 117     pub fn read_bool(&mut self) -> bool {
 118         self.read_prob(128)
 119     }
 120     pub fn read_prob(&mut self, prob: u8) -> bool {
 121         self.renorm();
 122         let split = 1 + (((self.range - 1) * (prob as u32)) >> 8);
 123         let bit;
 124         if self.value < (split << 24) {
 125             self.range = split;
 126             bit = false;
 127         } else {
 128             self.range -= split;
 129             self.value -= split << 24;
 130             bit = true;
 131         }
 132         bit
 133     }
 134     pub fn read_bits(&mut self, bits: u8) -> u32 {
 135         let mut val = 0u32;
 136         for _ in 0..bits {
 137             val = (val << 1) | (self.read_prob(128) as u32);
 138         }
 139         val
 140     }
 141     pub fn read_byte(&mut self) -> u8 {
 142         let mut val = 0u8;
 143         for _ in 0..8 {
 144             val = (val << 1) | (self.read_prob(128) as u8);
 145         }
 146         val
 147     }
 148     pub fn read_sbits(&mut self, bits: u8) -> i32 {
 149         let mut val = if self.read_prob(128) { -1i32 } else { 0i32 };
 150         for _ in 1..bits {
 151             val = (val << 1) | (self.read_prob(128) as i32);
 152         }
 153         val
 154     }
 155     pub fn read_probability(&mut self) -> u8 {
 156         let val = self.read_bits(7) as u8;
 157         if val == 0 {
 158             1
 159         } else {
 160             val << 1
 161         }
 162     }
 163     fn renorm(&mut self) {
 164         let shift = self.range.leading_zeros() & 7;
 165         self.range <<= shift;
 166         self.value <<= shift;
 167         self.bits   -= shift as i32;
 168         if (self.bits <= 0) && (self.pos < self.src.len()) {
 169             self.value |= (self.src[self.pos] as u32) << (-self.bits as u8);
 170             self.pos += 1;
 171             self.bits += 8;
 172         }
 173 /*        while self.range < 0x80 {
 174             self.range <<= 1;
 175             self.value <<= 1;
 176             self.bits   -= 1;
 177             if (self.bits <= 0) && (self.pos < self.src.len()) {
 178                 self.value |= self.src[self.pos] as u32;
 179                 self.pos += 1;
 180                 self.bits = 8;
 181             }
 182         }*/
 183     }
 184     pub fn skip_bytes(&mut self, nbytes: usize) {
 185         for _ in 0..nbytes {
 186             self.value <<= 8;
 187             if self.pos < self.src.len() {
 188                 self.value |= self.src[self.pos] as u32;
 189                 self.pos += 1;
 190             }
 191         }
 192     }
 193 }
 194
 195 #[allow(dead_code)]
 196 pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
 197     ((((prob as i32) * (weights[0] as i32) + 128) >> 8) + (weights[1] as i32)).min(maxval).max(1) as u8
 198 }
 199
 200 macro_rules! vp_tree {
 201     ($bc: expr, $prob: expr, $node1: expr, $node2: expr) => {
 202         if !$bc.read_prob($prob) {
 203             $node1
 204         } else {
 205             $node2
 206         }
 207     };
 208     ($leaf: expr) => { $leaf }
 209 }
 210
 211 const C1S7: i32 = 64277;
 212 const C2S6: i32 = 60547;
 213 const C3S5: i32 = 54491;
 214 const C4S4: i32 = 46341;
 215 const C5S3: i32 = 36410;
 216 const C6S2: i32 = 25080;
 217 const C7S1: i32 = 12785;
 218
 219 fn mul16(a: i32, b: i32) -> i32 {
 220     (a * b) >> 16
 221 }
 222
 223 macro_rules! idct_step {
 224     ($s0:expr, $s1:expr, $s2:expr, $s3:expr, $s4:expr, $s5:expr, $s6:expr, $s7:expr,
 225      $d0:expr, $d1:expr, $d2:expr, $d3:expr, $d4:expr, $d5:expr, $d6:expr, $d7:expr,
 226      $bias:expr, $shift:expr, $otype:ty) => {
 227         let t_a  = mul16(C1S7, i32::from($s1)) + mul16(C7S1, i32::from($s7));
 228         let t_b  = mul16(C7S1, i32::from($s1)) - mul16(C1S7, i32::from($s7));
 229         let t_c  = mul16(C3S5, i32::from($s3)) + mul16(C5S3, i32::from($s5));
 230         let t_d  = mul16(C3S5, i32::from($s5)) - mul16(C5S3, i32::from($s3));
 231         let t_a1 = mul16(C4S4, t_a - t_c);
 232         let t_b1 = mul16(C4S4, t_b - t_d);
 233         let t_c  = t_a + t_c;
 234         let t_d  = t_b + t_d;
 235         let t_e  = mul16(C4S4, i32::from($s0 + $s4)) + $bias;
 236         let t_f  = mul16(C4S4, i32::from($s0 - $s4)) + $bias;
 237         let t_g  = mul16(C2S6, i32::from($s2)) + mul16(C6S2, i32::from($s6));
 238         let t_h  = mul16(C6S2, i32::from($s2)) - mul16(C2S6, i32::from($s6));
 239         let t_e1 = t_e  - t_g;
 240         let t_g  = t_e  + t_g;
 241         let t_a  = t_f  + t_a1;
 242         let t_f  = t_f  - t_a1;
 243         let t_b  = t_b1 - t_h;
 244         let t_h  = t_b1 + t_h;
 245
 246         $d0 = ((t_g  + t_c) >> $shift) as $otype;
 247         $d7 = ((t_g  - t_c) >> $shift) as $otype;
 248         $d1 = ((t_a  + t_h) >> $shift) as $otype;
 249         $d2 = ((t_a  - t_h) >> $shift) as $otype;
 250         $d3 = ((t_e1 + t_d) >> $shift) as $otype;
 251         $d4 = ((t_e1 - t_d) >> $shift) as $otype;
 252         $d5 = ((t_f  + t_b) >> $shift) as $otype;
 253         $d6 = ((t_f  - t_b) >> $shift) as $otype;
 254     }
 255 }
 256
 257 pub fn vp_idct(coeffs: &mut [i16; 64]) {
 258     let mut tmp = [0i32; 64];
 259     for (src, dst) in coeffs.chunks(8).zip(tmp.chunks_mut(8)) {
 260         idct_step!(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
 261                    dst[0], dst[1], dst[2], dst[3], dst[4], dst[5], dst[6], dst[7], 0, 0, i32);
 262     }
 263     let src = &tmp;
 264     let dst = coeffs;
 265     for i in 0..8 {
 266         idct_step!(src[0 * 8 + i], src[1 * 8 + i], src[2 * 8 + i], src[3 * 8 + i],
 267                    src[4 * 8 + i], src[5 * 8 + i], src[6 * 8 + i], src[7 * 8 + i],
 268                    dst[0 * 8 + i], dst[1 * 8 + i], dst[2 * 8 + i], dst[3 * 8 + i],
 269                    dst[4 * 8 + i], dst[5 * 8 + i], dst[6 * 8 + i], dst[7 * 8 + i], 8, 4, i16);
 270     }
 271 }
 272
 273 pub fn vp_idct_dc(coeffs: &mut [i16; 64]) {
 274     let dc = ((mul16(C4S4, mul16(C4S4, i32::from(coeffs[0]))) + 8) >> 4) as i16;
 275     for i in 0..64 {
 276         coeffs[i] = dc;
 277     }
 278 }
 279
 280 pub fn unquant(coeffs: &mut [i16; 64], qmat: &[i16; 64]) {
 281     for i in 1..64 {
 282         coeffs[i] = coeffs[i].wrapping_mul(qmat[i]);
 283     }
 284 }
 285
 286 pub fn vp_put_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 287     vp_idct(coeffs);
 288     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 289     for y in 0..8 {
 290         for x in 0..8 {
 291             frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
 292         }
 293         off += frm.stride[plane];
 294     }
 295 }
 296
 297 pub fn vp_put_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 298     vp_idct(coeffs);
 299     let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
 300     for y in 0..8 {
 301         for x in 0..8 {
 302             frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
 303         }
 304         off += frm.stride[plane] * 2;
 305     }
 306 }
 307
 308 pub fn vp_put_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 309     vp_idct_dc(coeffs);
 310     let dc = (coeffs[0] + 128).min(255).max(0) as u8;
 311     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 312     for _ in 0..8 {
 313         for x in 0..8 {
 314             frm.data[off + x] = dc;
 315         }
 316         off += frm.stride[plane];
 317     }
 318 }
 319
 320 pub fn vp_add_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 321     vp_idct(coeffs);
 322     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 323     for y in 0..8 {
 324         for x in 0..8 {
 325             frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
 326         }
 327         off += frm.stride[plane];
 328     }
 329 }
 330
 331 pub fn vp_add_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 332     vp_idct(coeffs);
 333     let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
 334     for y in 0..8 {
 335         for x in 0..8 {
 336             frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
 337         }
 338         off += frm.stride[plane] * 2;
 339     }
 340 }
 341
 342 pub fn vp_add_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 343     vp_idct_dc(coeffs);
 344     let dc = coeffs[0];
 345     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 346     for _ in 0..8 {
 347         for x in 0..8 {
 348             frm.data[off + x] = (dc + (frm.data[off + x] as i16)).min(255).max(0) as u8;
 349         }
 350         off += frm.stride[plane];
 351     }
 352 }
 353
 354 pub fn vp31_loop_filter(data: &mut [u8], mut off: usize, step: usize, stride: usize,
 355                         len: usize, loop_str: i16) {
 356     for _ in 0..len {
 357         let a = data[off - step * 2] as i16;
 358         let b = data[off - step] as i16;
 359         let c = data[off] as i16;
 360         let d = data[off + step] as i16;
 361         let mut diff = ((a - d) + 3 * (c - b) + 4) >> 3;
 362         if diff.abs() >= 2 * loop_str {
 363             diff = 0;
 364         } else if diff.abs() >= loop_str {
 365             if diff < 0 {
 366                 diff = -diff - 2 * loop_str;
 367             } else {
 368                 diff = -diff + 2 * loop_str;
 369             }
 370         }
 371         if diff != 0 {
 372             data[off - step] = (b + diff).max(0).min(255) as u8;
 373             data[off]        = (c - diff).max(0).min(255) as u8;
 374         }
 375
 376         off += stride;
 377     }
 378 }
 379
 380 pub fn vp_copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, comp: usize,
 381                      dx: usize, dy: usize, mv_x: i16, mv_y: i16,
 382                      preborder: usize, postborder: usize, loop_str: i16,
 383                      mode: usize, interp: &[BlkInterpFunc], mut mc_buf: NAVideoBufferRef<u8>)
 384 {
 385     let sx = (dx as isize) + (mv_x as isize);
 386     let sy = (dy as isize) + (mv_y as isize);
 387     if ((sx | sy) & 7) == 0 {
 388         copy_block(dst, src, comp, dx, dy, mv_x, mv_y, 8, 8, preborder, postborder, mode, interp);
 389         return;
 390     }
 391     let pre = preborder.max(2);
 392     let post = postborder.max(1);
 393     let bsize = 8 + pre + post;
 394     let src_x = sx - (pre as isize);
 395     let src_y = sy - (pre as isize);
 396     {
 397         let tmp_buf = NASimpleVideoFrame::from_video_buf(&mut mc_buf).unwrap();
 398         edge_emu(src.as_ref(), src_x, src_y, bsize, bsize, &mut tmp_buf.data[tmp_buf.offset[comp]..], tmp_buf.stride[comp], comp);
 399 //        copy_block(&mut tmp_buf, src, comp, 0, 0, src_x as i16, src_y as i16,
 400 //                   bsize, bsize, 0, 0, 0, interp);
 401         if (sx & 7) != 0 {
 402             let foff = (8 - (sx & 7)) as usize;
 403             let off = pre + foff + tmp_buf.offset[comp];
 404             vp31_loop_filter(tmp_buf.data, off, 1, tmp_buf.stride[comp], bsize, loop_str);
 405         }
 406         if (sy & 7) != 0 {
 407             let foff = (8 - (sy & 7)) as usize;
 408             let off = (pre + foff) * tmp_buf.stride[comp] + tmp_buf.offset[comp];
 409             vp31_loop_filter(tmp_buf.data, off, tmp_buf.stride[comp], 1, bsize, loop_str);
 410         }
 411     }
 412     let dxoff = (pre as i16) - (dx as i16);
 413     let dyoff = (pre as i16) - (dy as i16);
 414     copy_block(dst, mc_buf, comp, dx, dy, dxoff, dyoff, 8, 8, preborder, postborder, mode, interp);
 415 }
 416
 417 fn vp3_interp00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 418 {
 419     let mut didx = 0;
 420     let mut sidx = 0;
 421     for _ in 0..bh {
 422         for x in 0..bw { dst[didx + x] = src[sidx + x]; }
 423         didx += dstride;
 424         sidx += sstride;
 425     }
 426 }
 427
 428 fn vp3_interp01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 429 {
 430     let mut didx = 0;
 431     let mut sidx = 0;
 432     for _ in 0..bh {
 433         for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + 1] as u16)) >> 1) as u8; }
 434         didx += dstride;
 435         sidx += sstride;
 436     }
 437 }
 438
 439 fn vp3_interp10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 440 {
 441     let mut didx = 0;
 442     let mut sidx = 0;
 443     for _ in 0..bh {
 444         for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + sstride] as u16)) >> 1) as u8; }
 445         didx += dstride;
 446         sidx += sstride;
 447     }
 448 }
 449
 450 fn vp3_interp1x(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] as u16) +
 457                               (src[sidx + x + sstride + 1] as u16)) >> 1) as u8;
 458         }
 459         didx += dstride;
 460         sidx += sstride;
 461     }
 462 }
 463
 464 fn vp3_interp1y(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 465 {
 466     let mut didx = 0;
 467     let mut sidx = 0;
 468     for _ in 0..bh {
 469         for x in 0..bw {
 470             dst[didx + x] = (((src[sidx + x + 1] as u16) +
 471                               (src[sidx + x + sstride] as u16)) >> 1) as u8;
 472         }
 473         didx += dstride;
 474         sidx += sstride;
 475     }
 476 }
 477
 478 pub const VP3_INTERP_FUNCS: &[blockdsp::BlkInterpFunc] = &[ vp3_interp00, vp3_interp01, vp3_interp10, vp3_interp1x, vp3_interp1y ];
 479