nihav-duck/src/codecs/vpcommon.rs

   1 use nihav_core::codecs::*;
   2 use nihav_codec_support::codecs::blockdsp;
   3 use nihav_codec_support::codecs::blockdsp::*;
   4
   5 pub const VP_YUVA420_FORMAT: NAPixelFormaton = NAPixelFormaton{
   6         model:      ColorModel::YUV(YUVSubmodel::YUVJ),
   7         components: 4,
   8         comp_info:  [
   9                 Some(NAPixelChromaton{ h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 0, next_elem: 1}),
  10                 Some(NAPixelChromaton{ h_ss: 1, v_ss: 1, packed: false, depth: 8, shift: 0, comp_offs: 1, next_elem: 1}),
  11                 Some(NAPixelChromaton{ h_ss: 1, v_ss: 1, packed: false, depth: 8, shift: 0, comp_offs: 2, next_elem: 1}),
  12                 Some(NAPixelChromaton{ h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 3, next_elem: 1}),
  13                 None ],
  14         elem_size:  0,
  15         be:         false,
  16         alpha:      true,
  17         palette:    false
  18     };
  19
  20 #[derive(Clone,Copy,Debug,PartialEq)]
  21 #[allow(dead_code)]
  22 pub enum VPMBType {
  23     Intra,
  24     InterNoMV,
  25     InterMV,
  26     InterNearest,
  27     InterNear,
  28     InterFourMV,
  29     GoldenNoMV,
  30     GoldenMV,
  31     GoldenNearest,
  32     GoldenNear,
  33 }
  34
  35 pub const VP_REF_INTER: u8 = 1;
  36 pub const VP_REF_GOLDEN: u8 = 2;
  37
  38 #[allow(dead_code)]
  39 impl VPMBType {
  40     pub fn is_intra(self) -> bool { self == VPMBType::Intra }
  41     pub fn get_ref_id(self) -> u8 {
  42         match self {
  43             VPMBType::Intra         => 0,
  44             VPMBType::InterNoMV     |
  45             VPMBType::InterMV       |
  46             VPMBType::InterNearest  |
  47             VPMBType::InterNear     |
  48             VPMBType::InterFourMV   => VP_REF_INTER,
  49             _                       => VP_REF_GOLDEN,
  50         }
  51     }
  52 }
  53
  54 impl Default for VPMBType {
  55     fn default() -> Self { VPMBType::Intra }
  56 }
  57
  58 #[derive(Default)]
  59 pub struct VPShuffler {
  60     lastframe: Option<NAVideoBufferRef<u8>>,
  61     goldframe: Option<NAVideoBufferRef<u8>>,
  62 }
  63
  64 impl VPShuffler {
  65     pub fn new() -> Self { VPShuffler { lastframe: None, goldframe: None } }
  66     pub fn clear(&mut self) { self.lastframe = None; self.goldframe = None; }
  67     pub fn add_frame(&mut self, buf: NAVideoBufferRef<u8>) {
  68         self.lastframe = Some(buf);
  69     }
  70     pub fn add_golden_frame(&mut self, buf: NAVideoBufferRef<u8>) {
  71         self.goldframe = Some(buf);
  72     }
  73     pub fn get_last(&mut self) -> Option<NAVideoBufferRef<u8>> {
  74         if let Some(ref frm) = self.lastframe {
  75             Some(frm.clone())
  76         } else {
  77             None
  78         }
  79     }
  80     pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
  81         if let Some(ref frm) = self.goldframe {
  82             Some(frm.clone())
  83         } else {
  84             None
  85         }
  86     }
  87     pub fn has_refs(&self) -> bool {
  88         self.lastframe.is_some()
  89     }
  90 }
  91
  92 pub const VP56_COEF_BASE: [i16; 6] = [ 5, 7, 11, 19, 35, 67 ];
  93 pub const VP56_COEF_ADD_PROBS: [[u8; 12]; 6] = [
  94     [ 159, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
  95     [ 165, 145, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
  96     [ 173, 148, 140, 128,   0,   0,   0,   0,   0,   0,   0,   0 ],
  97     [ 176, 155, 140, 135, 128,   0,   0,   0,   0,   0,   0,   0 ],
  98     [ 180, 157, 141, 134, 130, 128,   0,   0,   0,   0,   0,   0 ],
  99     [ 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 128 ],
 100 ];
 101
 102 #[allow(dead_code)]
 103 pub struct BoolCoder<'a> {
 104     pub src:    &'a [u8],
 105     pos:    usize,
 106     value:  u32,
 107     range:  u32,
 108     bits:   i32,
 109 }
 110
 111 #[allow(dead_code)]
 112 impl<'a> BoolCoder<'a> {
 113     pub fn new(src: &'a [u8]) -> DecoderResult<Self> {
 114         if src.len() < 3 { return Err(DecoderError::ShortData); }
 115         let value = (u32::from(src[0]) << 24) | (u32::from(src[1]) << 16) | (u32::from(src[2]) << 8) | u32::from(src[3]);
 116         Ok(Self { src, pos: 4, value, range: 255, bits: 8 })
 117     }
 118     pub fn read_bool(&mut self) -> bool {
 119         self.read_prob(128)
 120     }
 121     pub fn read_prob(&mut self, prob: u8) -> bool {
 122         self.renorm();
 123         let split = 1 + (((self.range - 1) * u32::from(prob)) >> 8);
 124         let bit;
 125         if self.value < (split << 24) {
 126             self.range = split;
 127             bit = false;
 128         } else {
 129             self.range -= split;
 130             self.value -= split << 24;
 131             bit = true;
 132         }
 133         bit
 134     }
 135     pub fn read_bits(&mut self, bits: u8) -> u32 {
 136         let mut val = 0u32;
 137         for _ in 0..bits {
 138             val = (val << 1) | (self.read_prob(128) as u32);
 139         }
 140         val
 141     }
 142     pub fn read_byte(&mut self) -> u8 {
 143         let mut val = 0u8;
 144         for _ in 0..8 {
 145             val = (val << 1) | (self.read_prob(128) as u8);
 146         }
 147         val
 148     }
 149     pub fn read_sbits(&mut self, bits: u8) -> i32 {
 150         let mut val = if self.read_prob(128) { -1i32 } else { 0i32 };
 151         for _ in 1..bits {
 152             val = (val << 1) | (self.read_prob(128) as i32);
 153         }
 154         val
 155     }
 156     pub fn read_probability(&mut self) -> u8 {
 157         let val = self.read_bits(7) as u8;
 158         if val == 0 {
 159             1
 160         } else {
 161             val << 1
 162         }
 163     }
 164     fn renorm(&mut self) {
 165         let shift = self.range.leading_zeros() & 7;
 166         self.range <<= shift;
 167         self.value <<= shift;
 168         self.bits   -= shift as i32;
 169         if (self.bits <= 0) && (self.pos < self.src.len()) {
 170             self.value |= u32::from(self.src[self.pos]) << (-self.bits as u8);
 171             self.pos += 1;
 172             self.bits += 8;
 173         }
 174 /*        while self.range < 0x80 {
 175             self.range <<= 1;
 176             self.value <<= 1;
 177             self.bits   -= 1;
 178             if (self.bits <= 0) && (self.pos < self.src.len()) {
 179                 self.value |= u32::from(self.src[self.pos]);
 180                 self.pos += 1;
 181                 self.bits = 8;
 182             }
 183         }*/
 184     }
 185     pub fn skip_bytes(&mut self, nbytes: usize) {
 186         for _ in 0..nbytes {
 187             self.value <<= 8;
 188             if self.pos < self.src.len() {
 189                 self.value |= u32::from(self.src[self.pos]);
 190                 self.pos += 1;
 191             }
 192         }
 193     }
 194 }
 195
 196 #[allow(dead_code)]
 197 #[allow(clippy::trivially_copy_pass_by_ref)]
 198 pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
 199     (((i32::from(prob) * i32::from(weights[0]) + 128) >> 8) + i32::from(weights[1])).min(maxval).max(1) as u8
 200 }
 201
 202 macro_rules! vp_tree {
 203     ($bc: expr, $prob: expr, $node1: expr, $node2: expr) => {
 204         if !$bc.read_prob($prob) {
 205             $node1
 206         } else {
 207             $node2
 208         }
 209     };
 210     ($leaf: expr) => { $leaf }
 211 }
 212
 213 const C1S7: i32 = 64277;
 214 const C2S6: i32 = 60547;
 215 const C3S5: i32 = 54491;
 216 const C4S4: i32 = 46341;
 217 const C5S3: i32 = 36410;
 218 const C6S2: i32 = 25080;
 219 const C7S1: i32 = 12785;
 220
 221 fn mul16(a: i32, b: i32) -> i32 {
 222     (a * b) >> 16
 223 }
 224
 225 macro_rules! idct_step {
 226     ($s0:expr, $s1:expr, $s2:expr, $s3:expr, $s4:expr, $s5:expr, $s6:expr, $s7:expr,
 227      $d0:expr, $d1:expr, $d2:expr, $d3:expr, $d4:expr, $d5:expr, $d6:expr, $d7:expr,
 228      $bias:expr, $shift:expr, $otype:ty) => {
 229         let t_a  = mul16(C1S7, i32::from($s1)) + mul16(C7S1, i32::from($s7));
 230         let t_b  = mul16(C7S1, i32::from($s1)) - mul16(C1S7, i32::from($s7));
 231         let t_c  = mul16(C3S5, i32::from($s3)) + mul16(C5S3, i32::from($s5));
 232         let t_d  = mul16(C3S5, i32::from($s5)) - mul16(C5S3, i32::from($s3));
 233         let t_a1 = mul16(C4S4, t_a - t_c);
 234         let t_b1 = mul16(C4S4, t_b - t_d);
 235         let t_c  = t_a + t_c;
 236         let t_d  = t_b + t_d;
 237         let t_e  = mul16(C4S4, i32::from($s0 + $s4)) + $bias;
 238         let t_f  = mul16(C4S4, i32::from($s0 - $s4)) + $bias;
 239         let t_g  = mul16(C2S6, i32::from($s2)) + mul16(C6S2, i32::from($s6));
 240         let t_h  = mul16(C6S2, i32::from($s2)) - mul16(C2S6, i32::from($s6));
 241         let t_e1 = t_e  - t_g;
 242         let t_g  = t_e  + t_g;
 243         let t_a  = t_f  + t_a1;
 244         let t_f  = t_f  - t_a1;
 245         let t_b  = t_b1 - t_h;
 246         let t_h  = t_b1 + t_h;
 247
 248         $d0 = ((t_g  + t_c) >> $shift) as $otype;
 249         $d7 = ((t_g  - t_c) >> $shift) as $otype;
 250         $d1 = ((t_a  + t_h) >> $shift) as $otype;
 251         $d2 = ((t_a  - t_h) >> $shift) as $otype;
 252         $d3 = ((t_e1 + t_d) >> $shift) as $otype;
 253         $d4 = ((t_e1 - t_d) >> $shift) as $otype;
 254         $d5 = ((t_f  + t_b) >> $shift) as $otype;
 255         $d6 = ((t_f  - t_b) >> $shift) as $otype;
 256     }
 257 }
 258
 259 pub fn vp_idct(coeffs: &mut [i16; 64]) {
 260     let mut tmp = [0i32; 64];
 261     for (src, dst) in coeffs.chunks(8).zip(tmp.chunks_mut(8)) {
 262         idct_step!(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
 263                    dst[0], dst[1], dst[2], dst[3], dst[4], dst[5], dst[6], dst[7], 0, 0, i32);
 264     }
 265     let src = &tmp;
 266     let dst = coeffs;
 267     for i in 0..8 {
 268         idct_step!(src[0 * 8 + i], src[1 * 8 + i], src[2 * 8 + i], src[3 * 8 + i],
 269                    src[4 * 8 + i], src[5 * 8 + i], src[6 * 8 + i], src[7 * 8 + i],
 270                    dst[0 * 8 + i], dst[1 * 8 + i], dst[2 * 8 + i], dst[3 * 8 + i],
 271                    dst[4 * 8 + i], dst[5 * 8 + i], dst[6 * 8 + i], dst[7 * 8 + i], 8, 4, i16);
 272     }
 273 }
 274
 275 pub fn vp_idct_dc(coeffs: &mut [i16; 64]) {
 276     let dc = ((mul16(C4S4, mul16(C4S4, i32::from(coeffs[0]))) + 8) >> 4) as i16;
 277     for i in 0..64 {
 278         coeffs[i] = dc;
 279     }
 280 }
 281
 282 pub fn unquant(coeffs: &mut [i16; 64], qmat: &[i16; 64]) {
 283     for i in 1..64 {
 284         coeffs[i] = coeffs[i].wrapping_mul(qmat[i]);
 285     }
 286 }
 287
 288 pub fn vp_put_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 289     vp_idct(coeffs);
 290     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 291     for y in 0..8 {
 292         for x in 0..8 {
 293             frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
 294         }
 295         off += frm.stride[plane];
 296     }
 297 }
 298
 299 pub fn vp_put_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 300     vp_idct(coeffs);
 301     let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
 302     for y in 0..8 {
 303         for x in 0..8 {
 304             frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
 305         }
 306         off += frm.stride[plane] * 2;
 307     }
 308 }
 309
 310 pub fn vp_put_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 311     vp_idct_dc(coeffs);
 312     let dc = (coeffs[0] + 128).min(255).max(0) as u8;
 313     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 314     for _ in 0..8 {
 315         for x in 0..8 {
 316             frm.data[off + x] = dc;
 317         }
 318         off += frm.stride[plane];
 319     }
 320 }
 321
 322 pub fn vp_add_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 323     vp_idct(coeffs);
 324     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 325     for y in 0..8 {
 326         for x in 0..8 {
 327             frm.data[off + x] = (coeffs[x + y * 8] + i16::from(frm.data[off + x])).min(255).max(0) as u8;
 328         }
 329         off += frm.stride[plane];
 330     }
 331 }
 332
 333 pub fn vp_add_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 334     vp_idct(coeffs);
 335     let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
 336     for y in 0..8 {
 337         for x in 0..8 {
 338             frm.data[off + x] = (coeffs[x + y * 8] + i16::from(frm.data[off + x])).min(255).max(0) as u8;
 339         }
 340         off += frm.stride[plane] * 2;
 341     }
 342 }
 343
 344 pub fn vp_add_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
 345     vp_idct_dc(coeffs);
 346     let dc = coeffs[0];
 347     let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
 348     for _ in 0..8 {
 349         for x in 0..8 {
 350             frm.data[off + x] = (dc + i16::from(frm.data[off + x])).min(255).max(0) as u8;
 351         }
 352         off += frm.stride[plane];
 353     }
 354 }
 355
 356 pub fn vp31_loop_filter(data: &mut [u8], mut off: usize, step: usize, stride: usize,
 357                         len: usize, loop_str: i16) {
 358     for _ in 0..len {
 359         let a = i16::from(data[off - step * 2]);
 360         let b = i16::from(data[off - step]);
 361         let c = i16::from(data[off]);
 362         let d = i16::from(data[off + step]);
 363         let mut diff = ((a - d) + 3 * (c - b) + 4) >> 3;
 364         if diff.abs() >= 2 * loop_str {
 365             diff = 0;
 366         } else if diff.abs() >= loop_str {
 367             if diff < 0 {
 368                 diff = -diff - 2 * loop_str;
 369             } else {
 370                 diff = -diff + 2 * loop_str;
 371             }
 372         }
 373         if diff != 0 {
 374             data[off - step] = (b + diff).max(0).min(255) as u8;
 375             data[off]        = (c - diff).max(0).min(255) as u8;
 376         }
 377
 378         off += stride;
 379     }
 380 }
 381
 382 pub fn vp_copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, comp: usize,
 383                      dx: usize, dy: usize, mv_x: i16, mv_y: i16,
 384                      preborder: usize, postborder: usize, loop_str: i16,
 385                      mode: usize, interp: &[BlkInterpFunc], mut mc_buf: NAVideoBufferRef<u8>)
 386 {
 387     let sx = (dx as isize) + (mv_x as isize);
 388     let sy = (dy as isize) + (mv_y as isize);
 389     if ((sx | sy) & 7) == 0 {
 390         copy_block(dst, src, comp, dx, dy, mv_x, mv_y, 8, 8, preborder, postborder, mode, interp);
 391         return;
 392     }
 393     let pre = preborder.max(2);
 394     let post = postborder.max(1);
 395     let bsize = 8 + pre + post;
 396     let src_x = sx - (pre as isize);
 397     let src_y = sy - (pre as isize);
 398     {
 399         let tmp_buf = NASimpleVideoFrame::from_video_buf(&mut mc_buf).unwrap();
 400         edge_emu(src.as_ref(), src_x, src_y, bsize, bsize, &mut tmp_buf.data[tmp_buf.offset[comp]..], tmp_buf.stride[comp], comp, 0);
 401 //        copy_block(&mut tmp_buf, src, comp, 0, 0, src_x as i16, src_y as i16,
 402 //                   bsize, bsize, 0, 0, 0, interp);
 403         if (sx & 7) != 0 {
 404             let foff = (8 - (sx & 7)) as usize;
 405             let off = pre + foff + tmp_buf.offset[comp];
 406             vp31_loop_filter(tmp_buf.data, off, 1, tmp_buf.stride[comp], bsize, loop_str);
 407         }
 408         if (sy & 7) != 0 {
 409             let foff = (8 - (sy & 7)) as usize;
 410             let off = (pre + foff) * tmp_buf.stride[comp] + tmp_buf.offset[comp];
 411             vp31_loop_filter(tmp_buf.data, off, tmp_buf.stride[comp], 1, bsize, loop_str);
 412         }
 413     }
 414     let dxoff = (pre as i16) - (dx as i16);
 415     let dyoff = (pre as i16) - (dy as i16);
 416     copy_block(dst, mc_buf, comp, dx, dy, dxoff, dyoff, 8, 8, preborder, postborder, mode, interp);
 417 }
 418
 419 fn vp3_interp00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 420 {
 421     let mut didx = 0;
 422     let mut sidx = 0;
 423     for _ in 0..bh {
 424         dst[didx..][..bw].copy_from_slice(&src[sidx..][..bw]);
 425         didx += dstride;
 426         sidx += sstride;
 427     }
 428 }
 429
 430 fn vp3_interp01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 431 {
 432     let mut didx = 0;
 433     let mut sidx = 0;
 434     for _ in 0..bh {
 435         for x in 0..bw { dst[didx + x] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + 1])) >> 1) as u8; }
 436         didx += dstride;
 437         sidx += sstride;
 438     }
 439 }
 440
 441 fn vp3_interp10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 442 {
 443     let mut didx = 0;
 444     let mut sidx = 0;
 445     for _ in 0..bh {
 446         for x in 0..bw { dst[didx + x] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + sstride])) >> 1) as u8; }
 447         didx += dstride;
 448         sidx += sstride;
 449     }
 450 }
 451
 452 fn vp3_interp1x(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 453 {
 454     let mut didx = 0;
 455     let mut sidx = 0;
 456     for _ in 0..bh {
 457         for x in 0..bw {
 458             dst[didx + x] = ((u16::from(src[sidx + x]) +
 459                               u16::from(src[sidx + x + sstride + 1])) >> 1) as u8;
 460         }
 461         didx += dstride;
 462         sidx += sstride;
 463     }
 464 }
 465
 466 fn vp3_interp1y(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
 467 {
 468     let mut didx = 0;
 469     let mut sidx = 0;
 470     for _ in 0..bh {
 471         for x in 0..bw {
 472             dst[didx + x] = ((u16::from(src[sidx + x + 1]) +
 473                               u16::from(src[sidx + x + sstride])) >> 1) as u8;
 474         }
 475         didx += dstride;
 476         sidx += sstride;
 477     }
 478 }
 479
 480 pub const VP3_INTERP_FUNCS: &[blockdsp::BlkInterpFunc] = &[ vp3_interp00, vp3_interp01, vp3_interp10, vp3_interp1x, vp3_interp1y ];
 481