fn bink2_idct_old(coeffs: &mut [f32; 64]) {
let mut tmp: [f32; 64] = [0.0; 64];
+ coeffs[0] += 512.5;
for i in 0..8 {
idct!(float; coeffs, 8, i, tmp, 8, i, 0, 0);
}
let t0 = el!($src, $off - 0 * $step) + el!($src, $off + 1 * $step);
let t1 = el!($src, $off - 1 * $step) + el!($src, $off + 2 * $step);
let t2 = el!($src, $off - 2 * $step) + el!($src, $off + 3 * $step);
- clip8((((t0 * 19) >> 1) - t1 * 2 + (t2 >> 1) + 8) >> 4)
+ (((t0 * 19) >> 1) - t1 * 2 + (t2 >> 1) + 8) >> 4
});
}
};
}
+macro_rules! avg_tree {
+ ($a: expr, $b: expr) => (($a + $b + 1) >> 1);
+ ($a: expr, $b: expr, $c: expr, $d: expr) => (avg_tree!(avg_tree!($a, $b), avg_tree!($c, $d)));
+}
+
impl Bink2DSP {
fn calc_dc(src: &[u8], stride: usize) -> i32 {
+ let mut sums = [0u16; 8];
+ for i in 0..8 {
+ let s0 = src[i + stride * 0] as u16;
+ let s1 = src[i + stride * 1] as u16;
+ let s2 = src[i + stride * 2] as u16;
+ let s3 = src[i + stride * 3] as u16;
+ let s4 = src[i + stride * 4] as u16;
+ let s5 = src[i + stride * 5] as u16;
+ let s6 = src[i + stride * 6] as u16;
+ let s7 = src[i + stride * 7] as u16;
+ sums[i] = avg_tree!(avg_tree!(s0, s1, s2, s3), avg_tree!(s4, s5, s6, s7));
+ }
let mut sum = 0;
- for row in src.chunks(stride).take(8) {
- for i in 0..8 { sum += row[i] as i32; }
+ for e in sums.iter() {
+ sum += e;
}
- sum >> 3
+ sum as i32
}
fn put_mb4(dst: &mut [u8], mut off: usize, stride: usize, blk: &mut [[i32; 64]; 4]) {
bink2_idct(&mut blk[0]);
{
let dout = &mut dst[off..];
for (row, (b0, b1)) in dout.chunks_mut(stride).zip(blk[0].chunks(8).zip(blk[1].chunks(8))) {
- for i in 0..8 { row[i + 0] = clip8(b0[i] as i32); }
- for i in 0..8 { row[i + 8] = clip8(b1[i] as i32); }
+ for i in 0..8 { row[i + 0] = clip8((b0[i] as i32) - 512); }
+ for i in 0..8 { row[i + 8] = clip8((b1[i] as i32) - 512); }
}
}
off += stride * 8;
{
let dout = &mut dst[off..];
for (row, (b2, b3)) in dout.chunks_mut(stride).zip(blk[2].chunks(8).zip(blk[3].chunks(8))) {
- for i in 0..8 { row[i + 0] = clip8(b2[i] as i32); }
- for i in 0..8 { row[i + 8] = clip8(b3[i] as i32); }
+ for i in 0..8 { row[i + 0] = clip8((b2[i] as i32) - 512); }
+ for i in 0..8 { row[i + 8] = clip8((b3[i] as i32) - 512); }
}
}
}
{
let dout = &mut dst[off..];
for (row, (b0, b1)) in dout.chunks_mut(stride).zip(blk[0].chunks(8).zip(blk[1].chunks(8))) {
- for i in 0..8 { row[i + 0] = clip8((row[i + 0] as i32) + (b0[i] as i32)); }
- for i in 0..8 { row[i + 8] = clip8((row[i + 8] as i32) + (b1[i] as i32)); }
+ for i in 0..8 { row[i + 0] = clip8((row[i + 0] as i32) + (b0[i] as i32) - 512); }
+ for i in 0..8 { row[i + 8] = clip8((row[i + 8] as i32) + (b1[i] as i32) - 512); }
}
}
off += stride * 8;
{
let dout = &mut dst[off..];
for (row, (b2, b3)) in dout.chunks_mut(stride).zip(blk[2].chunks(8).zip(blk[3].chunks(8))) {
- for i in 0..8 { row[i + 0] = clip8((row[i + 0] as i32) + (b2[i] as i32)); }
- for i in 0..8 { row[i + 8] = clip8((row[i + 8] as i32) + (b3[i] as i32)); }
+ for i in 0..8 { row[i + 0] = clip8((row[i + 0] as i32) + (b2[i] as i32) - 512); }
+ for i in 0..8 { row[i + 8] = clip8((row[i + 8] as i32) + (b3[i] as i32) - 512); }
}
}
}
let (d_y, add_y) = if (mv.y & 1) != 0 { (2, 5) } else { (0, 0) };
let (w, h) = ref_pic.get_dimensions(plane);
- validate!((sx - d_x >= 0) && (sx - d_x + add_x + 16 <= (w as isize)));
- validate!((sy - d_y >= 0) && (sy - d_y + add_y + 16 <= (h as isize)));
+ let align_w = ((w + 31) & !31) as isize;
+ let align_h = ((h + 31) & !31) as isize;
+ validate!((sx - d_x >= 0) && (sx - d_x + add_x + 16 <= align_w));
+ validate!((sy - d_y >= 0) && (sy - d_y + add_y + 16 <= align_h));
let pstride = ref_pic.get_stride(plane);
let mut poff = ref_pic.get_offset(plane) + (sx as usize) + (sy as usize) * pstride;
let pdata = ref_pic.get_data();
1 => {
for out in dst.chunks_mut(stride).take(16) {
for i in 0..16 {
- out[i] = luma_filter!(ppix, poff + i, 1);
+ out[i] = clip8(luma_filter!(ppix, poff + i, 1));
}
poff += pstride;
}
2 => {
for out in dst.chunks_mut(stride).take(16) {
for i in 0..16 {
- out[i] = luma_filter!(ppix, poff + i, pstride);
+ out[i] = clip8(luma_filter!(ppix, poff + i, pstride));
}
poff += pstride;
}
},
3 => {
- let mut tmp = [0u8; 21 * 16];
+ let mut tmp = [0i16; 21 * 16];
for out in tmp.chunks_mut(16) {
for i in 0..16 {
- out[i] = luma_filter!(ppix, poff - 2 * pstride + i, 1);
+ out[i] = luma_filter!(ppix, poff - 2 * pstride + i, 1) as i16;
}
poff += pstride;
}
for (row, out) in dst.chunks_mut(stride).take(16).enumerate() {
for i in 0..16 {
- out[i] = luma_filter!(tmp, (row + 2) * 16 + i, 16);
+ out[i] = clip8(luma_filter!(tmp, (row + 2) * 16 + i, 16));
}
}
},
let add_y = if my != 0 { 1 } else { 0 };
let (w, h) = ref_pic.get_dimensions(plane);
- validate!((sx >= 0) && (sx + add_x + 8 <= (w as isize)));
- validate!((sy >= 0) && (sy + add_y + 8 <= (h as isize)));
+ let align_w = ((w + 15) & !15) as isize;
+ let align_h = ((h + 15) & !15) as isize;
+ validate!((sx >= 0) && (sx + add_x + 8 <= align_w));
+ validate!((sy >= 0) && (sy + add_y + 8 <= align_h));
let pstride = ref_pic.get_stride(plane);
let poff = ref_pic.get_offset(plane) + (sx as usize) + (sy as usize) * pstride;
let pdata = ref_pic.get_data();
let mut data = buf.get_data_mut();
let dst = data.as_mut_slice();
let bw = (width + 31) >> 5;
- let bh = (height + 31) >> 6;
+ let bheight = (height + 31) >> 5;
self.cur_w = (width + 7) & !7;
self.cur_h = ((height + 7) & !7) >> 1;
println!("fill {:X}", frame_flags);
unimplemented!();
}
+ let mut row_flags: Vec<bool> = Vec::with_capacity(bheight * 4);
+ let mut col_flags: Vec<bool> = Vec::with_capacity(bw * 4);
+ if (frame_flags & 0x10000) != 0 {
+ if (frame_flags & 0x8000) == 0 {
+ decode_flags(br, &mut row_flags, 1, bheight * 4 - 1)?;
+ } else {
+ row_flags.resize(bheight * 4, false);
+ }
+ if (frame_flags & 0x4000) == 0 {
+ decode_flags(br, &mut col_flags, 1, bw * 4 - 1)?;
+ } else {
+ col_flags.resize(bw * 4, false);
+ }
+ } else {
+ row_flags.resize(bheight * 4, false);
+ col_flags.resize(bw * 4, false);
+ }
+ //store frame_flags * 8 & 0x7F8
+
for slice_no in 0..2 {
- if slice_no == 1 {
+ let bh;
+ let yoff;
+ if slice_no == 0 {
+ bh = bheight >> 1;
+ yoff = 0;
+ } else {
+ bh = bheight - (bheight >> 1);
br.seek(offset2 * 8)?;
- off_y = ooff_y + stride_y * bh * 32;
- off_u = ooff_u + stride_u * bh * 16;
- off_v = ooff_v + stride_v * bh * 16;
- }
- let mut row_flags: Vec<u8> = Vec::with_capacity(height >> 3);
- let mut col_flags: Vec<u8> = Vec::with_capacity(width >> 3);
- if (frame_flags & 0x1000) != 0 {
- if (frame_flags & 0x8000) != 0 {
- decode_flags(br, &mut row_flags, 1, (height >> 3) - 1)?;
- }
- if (frame_flags & 0x4000) != 0 {
- decode_flags(br, &mut col_flags, 1, (width >> 3) - 1)?;
- }
+ off_y = ooff_y + stride_y * (bheight >> 1) * 32;
+ off_u = ooff_u + stride_u * (bheight >> 1) * 16;
+ off_v = ooff_v + stride_v * (bheight >> 1) * 16;
+ yoff = bheight >> 1;
}
- row_flags.resize(height >> 3, 0);
- col_flags.resize(width >> 3, 0);
- //store frame_flags * 8 & 0x7F8
let mut row_state = frame_flags & 0x2E000;
if is_intra {
let mut q_y_p = 8;
let mut q_u_p = 8;
let mut q_v_p = 8;
- let rflags = (row_flags[by >> 1] >> (if (by & 1) != 0 { 4 } else { 0 })) as u32;
+ let rflags = (row_flags[by + yoff] as u32) * 4;
row_state = (row_state & 0x3FFFFFF) | ((row_state >> 4) & 0xC000000) | (rflags << 28);
if by == 0 {
row_state |= 0x80;
if (bx & 1) != 0 {
blk_state |= 0x200;
}
- let clflags = (col_flags[bx >> 1] >> (if (bx & 1) != 0 { 4 } else { 0 })) as u32;
+ let clflags = (col_flags[bx] as u32) * 4;
let edge_state_c = ((blk_state >> 4) & 0x3C0000) | (blk_state & 0xFC03FFFF) | ((clflags & 0xF) << 22);
let edge_state_y = (frame_flags & 0x40000) | (blk_state & 0x3FFFF);
edge_state = edge_state_c;
if let Some(ref ref_pic) = self.ips.get_ref() {
for blk_no in 0..4 {
let xoff = bx * 32 + (blk_no & 1) * 16;
- let yoff = slice_no * bh * 32 + by * 32 + (blk_no & 2) * 8;
+ let yoff = slice_no * (bheight >> 1) * 32 + by * 32 + (blk_no & 2) * 8;
Bink2DSP::mc_luma(&mut dst[off_y..], stride_y, ref_pic, xoff, yoff, ZERO_MV, 0)?;
Bink2DSP::mc_chroma(&mut dst[off_u..], stride_u, ref_pic, xoff >> 1, yoff >> 1, ZERO_MV, 1)?;
Bink2DSP::mc_chroma(&mut dst[off_v..], stride_v, ref_pic, xoff >> 1, yoff >> 1, ZERO_MV, 2)?;
if let Some(ref ref_pic) = self.ips.get_ref() {
for blk_no in 0..4 {
let xoff = bx * 32 + (blk_no & 1) * 16;
- let yoff = slice_no * bh * 32 + by * 32 + (blk_no & 2) * 8;
+ let yoff = slice_no * (bheight >> 1) * 32 + by * 32 + (blk_no & 2) * 8;
let mv = self.mvs.get_mv(bx, blk_no);
Bink2DSP::mc_luma(&mut dst[off_y..], stride_y, ref_pic, xoff, yoff, mv, 0)?;
Bink2DSP::mc_chroma(&mut dst[off_u..], stride_u, ref_pic, xoff >> 1, yoff >> 1, mv, 1)?;
if let Some(ref ref_pic) = self.ips.get_ref() {
for blk_no in 0..4 {
let xoff = bx * 32 + (blk_no & 1) * 16;
- let yoff = slice_no * bh * 32 + by * 32 + (blk_no & 2) * 8;
+ let yoff = slice_no * (bheight >> 1) * 32 + by * 32 + (blk_no & 2) * 8;
let mv = self.mvs.get_mv(bx, blk_no);
Bink2DSP::mc_luma(&mut dst[off_y..], stride_y, ref_pic, xoff, yoff, mv, 0)?;
Bink2DSP::mc_chroma(&mut dst[off_u..], stride_u, ref_pic, xoff >> 1, yoff >> 1, mv, 1)?;
}
}
-fn decode_flags(_br: &mut BitReader, _dst: &mut Vec<u8>, _start: usize, _nbits: usize) -> DecoderResult<u32> {
-unimplemented!();
-/* if !br.read_bool()? { // read bits into byte array?
- if nbits == 0 { return Ok(()); }
- if nbits < 9 {
- shift = in_shift;
- pfx = 0;
- } else {
- shift = in_shift;
- loop {
- pfx |= br.read(8)? << shift;
- dst.push((pfx & 0xFF) as u8);
- pfx >>= 8;
- shift -= 8;
- }
+fn decode_flags(br: &mut BitReader, dst: &mut Vec<bool>, start: usize, nbits: usize) -> DecoderResult<()> {
+ if start > 0 {
+ dst.push(false);
+ }
+ if !br.read_bool()? {
+ for _ in 0..nbits {
+ let bit = br.read_bool()?;
+ dst.push(bit);
}
- let val = br.read(cur_nbits)?;
- dst.push(pfx | (val << shift))
} else {
let mut cur_bits = nbits;
let mut mode = 0;
- let mut lastbit = 0;
+ let mut lastbit = false;
while cur_bits > 0 {
if !br.read_bool()? {
- lastbit = if mode == 3 { lastbit ^ 1 } else { br.read(1)? };
- let val1 = lastval | (lastbit << shift);
- let val2 = br.read(if cur_bits > 4 { 4 } else { cur_bits });
- let val = lastval | (lastbit << shift) | (val2 << (shift + 1));
- mode = 2;
- if oshift >= 8 {
- dst.push((val & 0xFF) as u8);
- oshift -= 8;
- val >>= 8;
+ lastbit = if mode == 3 { !lastbit } else { br.read_bool()? };
+ dst.push(lastbit);
+ cur_bits -= 1;
+ let len = cur_bits.min(4);
+ for _ in 0..len {
+ let bit = br.read_bool()?;
+ dst.push(bit);
}
- lastval = val;
+ cur_bits -= len;
+ mode = 2;
} else {
- let bread;
+ let bread: u8;
if cur_bits < 4 {
bread = 2;
} else if cur_bits < 16 {
} else {
bread = 4 | 1;
}
- lastbit = if mode == 3 { lastbit ^ 1 } else { br.read(1)? };
- run = (if mode == 3 { bread + 1 } else { bread + 2 }).min(cur_bits);
- if run == cur_bits {
- output lastbit x run
- } else {
+ lastbit = if mode == 3 { !lastbit } else { br.read_bool()? };
+ let mut run = (if mode == 3 { bread + 1 } else { bread + 2 } as usize).min(cur_bits);
+ if run != cur_bits {
let add_run = br.read(bread)? as usize;
run += add_run;
- output lastbit x run
- mode = if add_run == (1 << bread) - 1 { 3 } else { 1 };
+ mode = if add_run == (1 << bread) - 1 { 1 } else { 3 };
+ }
+ for _ in 0..run {
+ dst.push(lastbit);
}
+ cur_bits -= run;
}
}
}
- Ok(())*/
+ Ok(())
}
fn get_new_quant(br: &mut BitReader, prev_q: u8) -> DecoderResult<u8> {
let mut buf;
self.key_frame = pkt.is_keyframe();
- let bufret = alloc_video_buffer(self.info.get_properties().get_video_info().unwrap(), 4);
+ let bufret = alloc_video_buffer(self.info.get_properties().get_video_info().unwrap(), 5);
if let Err(_) = bufret { return Err(DecoderError::InvalidData); }
let bufinfo = bufret.unwrap();
buf = bufinfo.get_vbuf().unwrap();
//let file = "assets/RAD/sc13_01_partial.bk2";
let file = "assets/RAD/ge_video_86l.bk2";
//let file = "assets/RAD/eg_club_0.bk2";
- test_file_decoding("bink", file, Some(42), true, false, Some("bink2"), &dmx_reg, &dec_reg);
+ test_file_decoding("bink", file, Some(8), true, false, None/*Some("bink2")*/, &dmx_reg, &dec_reg);
}
}
projects / nihav.git / blobdiff