X-Git-Url: https://git.nihav.org/?a=blobdiff_plain;f=nihav-duck%2Fsrc%2Fcodecs%2Fvp3.rs;h=e52c9400f1fddaef63542f6ec36f21d8930fba0e;hb=3cc76ad5da2772a179d6933695e84ec9212f84c2;hp=698d623746d1dcee1990bda9cbb655b54bdb9c5f;hpb=8e4b2f445d2c3e530e5fa517e19f9639142479cf;p=nihav.git diff --git a/nihav-duck/src/codecs/vp3.rs b/nihav-duck/src/codecs/vp3.rs index 698d623..e52c940 100644 --- a/nihav-duck/src/codecs/vp3.rs +++ b/nihav-duck/src/codecs/vp3.rs @@ -425,7 +425,7 @@ fn rescale_qmat(dst_qmat: &mut [i16; 64], base_qmat: &[i16; 64], dc_quant: i16, dst_qmat[0] = (base_qmat[0] * dc_quant / 100).max(minval * 2) << 2; } -fn expand_token(blk: &mut Block, br: &mut BitReader, eob_run: &mut usize, coef_no: usize, token: u8) -> DecoderResult<()> { +fn expand_token(blk: &mut Block, br: &mut BitReader, eob_run: &mut usize, token: u8) -> DecoderResult<()> { match token { // EOBs 0 | 1 | 2 => { *eob_run = (token as usize) + 1; }, @@ -527,7 +527,7 @@ fn expand_token(blk: &mut Block, br: &mut BitReader, eob_run: &mut usize, coef_n if *eob_run > 0 { blk.idx = 64; *eob_run -= 1; - } else if coef_no > 0 { + } else if (token > 8) && (blk.idx > 1) { blk.has_ac = true; } Ok(()) @@ -542,55 +542,6 @@ macro_rules! fill_dc_pred { }; } -fn vp3_interp00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize) -{ - let mut didx = 0; - let mut sidx = 0; - for _ in 0..bh { - for x in 0..bw { dst[didx + x] = src[sidx + x]; } - didx += dstride; - sidx += sstride; - } -} - -fn vp3_interp01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize) -{ - let mut didx = 0; - let mut sidx = 0; - for _ in 0..bh { - for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + 1] as u16)) >> 1) as u8; } - didx += dstride; - sidx += sstride; - } -} - -fn vp3_interp10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize) -{ - let mut didx = 0; - let mut sidx = 0; - for _ in 0..bh { - for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + sstride] as u16)) >> 1) as u8; } - didx += dstride; - sidx += sstride; - } -} - -fn vp3_interp11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize) -{ - let mut didx = 0; - let mut sidx = 0; - for _ in 0..bh { - for x in 0..bw { - dst[didx + x] = (((src[sidx + x] as u16) + - (src[sidx + x + 1] as u16) + - (src[sidx + x + sstride] as u16) + - (src[sidx + x + sstride + 1] as u16)) >> 2) as u8; - } - didx += dstride; - sidx += sstride; - } -} - fn vp31_loop_filter_v(frm: &mut NASimpleVideoFrame, x: usize, y: usize, plane: usize, loop_str: i16) { let off = frm.offset[plane] + x + y * frm.stride[plane]; vp31_loop_filter(frm.data, off, 1, frm.stride[plane], 8, loop_str); @@ -601,7 +552,14 @@ fn vp31_loop_filter_h(frm: &mut NASimpleVideoFrame, x: usize, y: usize, plan vp31_loop_filter(frm.data, off, frm.stride[plane], 1, 8, loop_str); } -pub const VP3_INTERP_FUNCS: &[blockdsp::BlkInterpFunc] = &[ vp3_interp00, vp3_interp01, vp3_interp10, vp3_interp11 ]; +fn vp3_mv_mode(mvx: i16, mvy: i16) -> usize { + let mode = ((mvx & 1) + (mvy & 1) * 2) as usize; + if (mode == 3) && (mvx ^ mvy < 0) { + 4 + } else { + mode + } +} impl VP34Decoder { fn new(version: u8) -> Self { @@ -871,7 +829,7 @@ println!("intra, ver {} (self {})", version, self.version); &codes.ac_p_cb[table] }; let token = br.read_cb(cb)?; - expand_token(blk, br, &mut self.eob_run, coef_no, token)?; + expand_token(blk, br, &mut self.eob_run, token)?; } Ok(()) } else { @@ -1064,7 +1022,7 @@ println!("intra, ver {} (self {})", version, self.version); } let cb = if (blkaddr & 3) == 0 { cbs[0] } else { cbs[1] }; let token = br.read_cb(cb)?; - expand_token(blk, br, &mut self.eob_run, coef_no, token)?; + expand_token(blk, br, &mut self.eob_run, token)?; } Ok(()) } else { @@ -1263,7 +1221,7 @@ println!("intra, ver {} (self {})", version, self.version); }; let cb = if (blkaddr & 3) == 0 { cbs[0] } else { cbs[1] }; let token = br.read_cb(cb)?; - expand_token(blk, br, &mut coef_eob[blk.idx], blk.idx, token)?; + expand_token(blk, br, &mut coef_eob[blk.idx], token)?; if blk.idx == 64 { break; } } let idx = blkaddr >> 2; @@ -1588,55 +1546,33 @@ println!("intra, ver {} (self {})", version, self.version); if self.blocks[blk_idx + bx].btype != VPMBType::InterFourMV { continue; } - let mv_a = self.blocks[blk_idx + bx].mv; - let mv_b = self.blocks[blk_idx + bx + 1].mv; - let mv_c = self.blocks[blk_idx + bx + bstride].mv; - let mv_d = self.blocks[blk_idx + bx + 1 + bstride].mv; - let mut mv_sum = mv_a + mv_b + mv_c + mv_d; + let mvs = [ self.blocks[blk_idx + bx].mv, + self.blocks[blk_idx + bx + 1].mv, + self.blocks[blk_idx + bx + bstride].mv, + self.blocks[blk_idx + bx + 1 + bstride].mv ]; + let mut mv_sum = mvs[0] + mvs[1] + mvs[2] + mvs[3]; mv_sum.x = (mv_sum.x + 2) >> 2; mv_sum.y = (mv_sum.y + 2) >> 2; let src = self.shuf.get_last().unwrap(); - let mode = ((mv_a.x & 1) + (mv_a.y & 1) * 2) as usize; - if self.version != 4 { - copy_block(frm, src.clone(), 0, bx * 8, by * 8, - mv_a.x >> 1, mv_a.y >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); - } else { - vp_copy_block(frm, src.clone(), 0, bx * 8, by * 8, - mv_a.x >> 1, mv_a.y >> 1, 0, 1, self.loop_str, - mode, VP3_INTERP_FUNCS, self.mc_buf.clone()); - } - let mode = ((mv_b.x & 1) + (mv_b.y & 1) * 2) as usize; - if self.version != 4 { - copy_block(frm, src.clone(), 0, bx * 8 + 8, by * 8, - mv_b.x >> 1, mv_b.y >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); - } else { - vp_copy_block(frm, src.clone(), 0, bx * 8 + 8, by * 8, - mv_b.x >> 1, mv_b.y >> 1, 0, 1, self.loop_str, - mode, VP3_INTERP_FUNCS, self.mc_buf.clone()); - } - let mode = ((mv_c.x & 1) + (mv_c.y & 1) * 2) as usize; - if self.version != 4 { - copy_block(frm, src.clone(), 0, bx * 8, by * 8 + 8, - mv_c.x >> 1, mv_c.y >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); - } else { - vp_copy_block(frm, src.clone(), 0, bx * 8, by * 8 + 8, - mv_c.x >> 1, mv_c.y >> 1, 0, 1, self.loop_str, - mode, VP3_INTERP_FUNCS, self.mc_buf.clone()); - } - let mode = ((mv_d.x & 1) + (mv_d.y & 1) * 2) as usize; - if self.version != 4 { - copy_block(frm, src.clone(), 0, bx * 8 + 8, by * 8 + 8, - mv_d.x >> 1, mv_d.y >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); - } else { - vp_copy_block(frm, src.clone(), 0, bx * 8 + 8, by * 8 + 8, - mv_d.x >> 1, mv_d.y >> 1, 0, 1, self.loop_str, - mode, VP3_INTERP_FUNCS, self.mc_buf.clone()); + for i in 0..4 { + let xoff = (i & 1) * 8; + let yoff = (i >> 1) * 8; + + let mode = vp3_mv_mode(mvs[i].x, mvs[i].y); + if self.version != 4 { + copy_block(frm, src.clone(), 0, bx * 8 + xoff, by * 8 + yoff, + mvs[i].x >> 1, mvs[i].y >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); + } else { + vp_copy_block(frm, src.clone(), 0, bx * 8 + xoff, by * 8 + yoff, + mvs[i].x >> 1, mvs[i].y >> 1, 0, 1, self.loop_str, + mode, VP3_INTERP_FUNCS, self.mc_buf.clone()); + } } let mx = (mv_sum.x >> 1) | (mv_sum.x & 1); let my = (mv_sum.y >> 1) | (mv_sum.y & 1); - let mode = ((mx & 1) + (my & 1) * 2) as usize; + let mode = vp3_mv_mode(mx, my); copy_block(frm, src.clone(), 1, bx * 4, by * 4, mx >> 1, my >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); copy_block(frm, src.clone(), 2, bx * 4, by * 4, mx >> 1, my >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); } @@ -1654,7 +1590,7 @@ println!("intra, ver {} (self {})", version, self.version); } else { self.shuf.get_golden().unwrap() }; - let mode = ((blk.mv.x & 1) + (blk.mv.y & 1) * 2) as usize; + let mode = vp3_mv_mode(blk.mv.x, blk.mv.y); if self.version != 4 { copy_block(frm, src.clone(), 0, bx * 8, by * 8, blk.mv.x >> 1, blk.mv.y >> 1, 16, 16, 0, 1, mode, VP3_INTERP_FUNCS); @@ -1674,7 +1610,7 @@ println!("intra, ver {} (self {})", version, self.version); } let mx = (blk.mv.x >> 1) | (blk.mv.x & 1); let my = (blk.mv.y >> 1) | (blk.mv.y & 1); - let mode = ((mx & 1) + (my & 1) * 2) as usize; + let mode = vp3_mv_mode(mx, my); if self.version != 4 { copy_block(frm, src.clone(), 1, bx * 4, by * 4, mx >> 1, my >> 1, 8, 8, 0, 1, mode, VP3_INTERP_FUNCS); @@ -1760,12 +1696,27 @@ println!("intra, ver {} (self {})", version, self.version); } blk_idx += blk_w; } -/* for plane in 1..3 { + let blk_w = self.mb_w; + for plane in 1..3 { for by in 0..self.mb_h { for bx in 0..self.mb_w { + let blk = &self.blocks[blk_idx + bx]; + if (bx > 0) && blk.coded { + vp31_loop_filter_v(frm, bx * 8, by * 8, plane, self.loop_str); + } + if (by > 0) && blk.coded { + vp31_loop_filter_h(frm, bx * 8, by * 8, plane, self.loop_str); + } + if (bx < blk_w - 1) && !self.blocks[blk_idx + bx + 1].coded { + vp31_loop_filter_v(frm, bx * 8 + 8, by * 8, plane, self.loop_str); + } + if (by < self.mb_h - 1) && !self.blocks[blk_idx + bx + blk_w].coded { + vp31_loop_filter_h(frm, bx * 8, by * 8 + 8, plane, self.loop_str); + } } + blk_idx += blk_w; } - }*/ + } } fn generate_block_addr(&mut self) { let sb_w_y = (self.width + 31) >> 5; @@ -1845,7 +1796,7 @@ impl NADecoder for VP34Decoder { let myinfo = NACodecTypeInfo::Video(NAVideoInfo::new(vinfo.get_width(), vinfo.get_height(), true, fmt)); self.info = NACodecInfo::new_ref(info.get_name(), myinfo, info.get_extradata()).into_ref(); supp.pool_u8.set_dec_bufs(3); - supp.pool_u8.prealloc_video(NAVideoInfo::new(vinfo.get_width(), vinfo.get_height(), false, fmt), 4)?; + supp.pool_u8.prealloc_video(NAVideoInfo::new(vinfo.get_width(), vinfo.get_height(), true, fmt), 4)?; self.generate_block_addr(); if self.version == 4 { @@ -1865,6 +1816,10 @@ impl NADecoder for VP34Decoder { self.parse_header(&mut br)?; if self.is_intra { self.shuf.clear(); + } else { + if !self.shuf.has_refs() { + return Err(DecoderError::MissingReference); + } } let ret = supp.pool_u8.get_free(); @@ -1890,13 +1845,16 @@ impl NADecoder for VP34Decoder { frm.set_frame_type(if self.is_intra { FrameType::I } else { FrameType::P }); Ok(frm.into_ref()) } + fn flush(&mut self) { + self.shuf.clear(); + } } -pub fn get_decoder_vp3() -> Box { +pub fn get_decoder_vp3() -> Box { Box::new(VP34Decoder::new(3)) } -pub fn get_decoder_vp4() -> Box { +pub fn get_decoder_vp4() -> Box { Box::new(VP34Decoder::new(4)) } @@ -2048,17 +2006,6 @@ const VP3_QMAT_INTER: &[i16; 64] = &[ 40, 48, 64, 64, 64, 96, 128, 128 ]; -const ZIGZAG: [usize; 64] = [ - 0, 1, 8, 16, 9, 2, 3, 10, - 17, 24, 32, 25, 18, 11, 4, 5, - 12, 19, 26, 33, 40, 48, 41, 34, - 27, 20, 13, 6, 7, 14, 21, 28, - 35, 42, 49, 56, 57, 50, 43, 36, - 29, 22, 15, 23, 30, 37, 44, 51, - 58, 59, 52, 45, 38, 31, 39, 46, - 53, 60, 61, 54, 47, 55, 62, 63 -]; - const VP31_DC_CODES: [[u16; 32]; 16] = [ [ 0x002D, 0x0026, 0x0166, 0x004E, 0x02CE, 0x059E, 0x027D, 0x0008,