[nihav.git] / nihav-duck / src / codecs / vpcommon.rs

use nihav_core::codecs::*;
use nihav_codec_support::codecs::blockdsp;
use nihav_codec_support::codecs::blockdsp::*;

pub const VP_YUVA420_FORMAT: NAPixelFormaton = NAPixelFormaton{
        model:      ColorModel::YUV(YUVSubmodel::YUVJ),
        components: 4,
        comp_info:  [
                Some(NAPixelChromaton{ h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 0, next_elem: 1}),
                Some(NAPixelChromaton{ h_ss: 1, v_ss: 1, packed: false, depth: 8, shift: 0, comp_offs: 1, next_elem: 1}),
                Some(NAPixelChromaton{ h_ss: 1, v_ss: 1, packed: false, depth: 8, shift: 0, comp_offs: 2, next_elem: 1}),
                Some(NAPixelChromaton{ h_ss: 0, v_ss: 0, packed: false, depth: 8, shift: 0, comp_offs: 3, next_elem: 1}),
                None ],
        elem_size:  0,
        be:         false,
        alpha:      true,
        palette:    false
    };

#[derive(Clone,Copy,Debug,PartialEq,Default)]
pub enum VPMBType {
    #[default]
    Intra,
    InterNoMV,
    InterMV,
    InterNearest,
    InterNear,
    InterFourMV,
    GoldenNoMV,
    GoldenMV,
    GoldenNearest,
    GoldenNear,
}

pub const VP_REF_INTER: u8 = 1;
pub const VP_REF_GOLDEN: u8 = 2;

impl VPMBType {
    pub fn is_intra(self) -> bool { self == VPMBType::Intra }
    pub fn get_ref_id(self) -> u8 {
        match self {
            VPMBType::Intra         => 0,
            VPMBType::InterNoMV     |
            VPMBType::InterMV       |
            VPMBType::InterNearest  |
            VPMBType::InterNear     |
            VPMBType::InterFourMV   => VP_REF_INTER,
            _                       => VP_REF_GOLDEN,
        }
    }
}

#[derive(Default)]
pub struct VPShuffler {
    lastframe: Option<NAVideoBufferRef<u8>>,
    goldframe: Option<NAVideoBufferRef<u8>>,
}

impl VPShuffler {
    pub fn new() -> Self { VPShuffler { lastframe: None, goldframe: None } }
    pub fn clear(&mut self) { self.lastframe = None; self.goldframe = None; }
    pub fn add_frame(&mut self, buf: NAVideoBufferRef<u8>) {
        self.lastframe = Some(buf);
    }
    pub fn add_golden_frame(&mut self, buf: NAVideoBufferRef<u8>) {
        self.goldframe = Some(buf);
    }
    pub fn get_last(&mut self) -> Option<NAVideoBufferRef<u8>> {
        self.lastframe.as_ref().cloned()
    }
    pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
        self.goldframe.as_ref().cloned()
    }
    pub fn has_refs(&self) -> bool {
        self.lastframe.is_some()
    }
}

pub const VP56_COEF_BASE: [i16; 6] = [ 5, 7, 11, 19, 35, 67 ];
pub const VP56_COEF_ADD_PROBS: [[u8; 12]; 6] = [
    [ 159, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
    [ 165, 145, 128,   0,   0,   0,   0,   0,   0,   0,   0,   0 ],
    [ 173, 148, 140, 128,   0,   0,   0,   0,   0,   0,   0,   0 ],
    [ 176, 155, 140, 135, 128,   0,   0,   0,   0,   0,   0,   0 ],
    [ 180, 157, 141, 134, 130, 128,   0,   0,   0,   0,   0,   0 ],
    [ 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 128 ],
];

pub struct BoolCoder<'a> {
    pub src:    &'a [u8],
    pos:    usize,
    value:  u32,
    range:  u32,
    bits:   i32,
}

impl<'a> BoolCoder<'a> {
    pub fn new(src: &'a [u8]) -> DecoderResult<Self> {
        if src.len() < 3 { return Err(DecoderError::ShortData); }
        let value = (u32::from(src[0]) << 24) | (u32::from(src[1]) << 16) | (u32::from(src[2]) << 8) | u32::from(src[3]);
        Ok(Self { src, pos: 4, value, range: 255, bits: 8 })
    }
    pub fn read_bool(&mut self) -> bool {
        self.read_prob(128)
    }
    pub fn read_prob(&mut self, prob: u8) -> bool {
        self.renorm();
        let split = 1 + (((self.range - 1) * u32::from(prob)) >> 8);
        let bit;
        if self.value < (split << 24) {
            self.range = split;
            bit = false;
        } else {
            self.range -= split;
            self.value -= split << 24;
            bit = true;
        }
        bit
    }
    pub fn read_bits(&mut self, bits: u8) -> u32 {
        let mut val = 0u32;
        for _ in 0..bits {
            val = (val << 1) | (self.read_prob(128) as u32);
        }
        val
    }
    pub fn read_byte(&mut self) -> u8 {
        let mut val = 0u8;
        for _ in 0..8 {
            val = (val << 1) | (self.read_prob(128) as u8);
        }
        val
    }
    pub fn read_sbits(&mut self, bits: u8) -> i32 {
        let mut val = if self.read_prob(128) { -1i32 } else { 0i32 };
        for _ in 1..bits {
            val = (val << 1) | (self.read_prob(128) as i32);
        }
        val
    }
    pub fn read_probability(&mut self) -> u8 {
        let val = self.read_bits(7) as u8;
        if val == 0 {
            1
        } else {
            val << 1
        }
    }
    fn renorm(&mut self) {
        let shift = self.range.leading_zeros() & 7;
        self.range <<= shift;
        self.value <<= shift;
        self.bits   -= shift as i32;
        if (self.bits <= 0) && (self.pos < self.src.len()) {
            self.value |= u32::from(self.src[self.pos]) << (-self.bits as u8);
            self.pos += 1;
            self.bits += 8;
        }
/*        while self.range < 0x80 {
            self.range <<= 1;
            self.value <<= 1;
            self.bits   -= 1;
            if (self.bits <= 0) && (self.pos < self.src.len()) {
                self.value |= u32::from(self.src[self.pos]);
                self.pos += 1;
                self.bits = 8;
            }
        }*/
    }
    pub fn skip_bytes(&mut self, nbytes: usize) {
        for _ in 0..nbytes {
            self.value <<= 8;
            if self.pos < self.src.len() {
                self.value |= u32::from(self.src[self.pos]);
                self.pos += 1;
            }
        }
    }
}

#[allow(dead_code)]
pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
    (((i32::from(prob) * i32::from(weights[0]) + 128) >> 8) + i32::from(weights[1])).min(maxval).max(1) as u8
}

macro_rules! vp_tree {
    ($bc: expr, $prob: expr, $node1: expr, $node2: expr) => {
        if !$bc.read_prob($prob) {
            $node1
        } else {
            $node2
        }
    };
    ($leaf: expr) => { $leaf }
}

const C1S7: i32 = 64277;
const C2S6: i32 = 60547;
const C3S5: i32 = 54491;
const C4S4: i32 = 46341;
const C5S3: i32 = 36410;
const C6S2: i32 = 25080;
const C7S1: i32 = 12785;

fn mul16(a: i32, b: i32) -> i32 {
    (a * b) >> 16
}

macro_rules! idct_step {
    ($s0:expr, $s1:expr, $s2:expr, $s3:expr, $s4:expr, $s5:expr, $s6:expr, $s7:expr,
     $d0:expr, $d1:expr, $d2:expr, $d3:expr, $d4:expr, $d5:expr, $d6:expr, $d7:expr,
     $bias:expr, $shift:expr, $otype:ty) => {
        let t_a  = mul16(C1S7, i32::from($s1)) + mul16(C7S1, i32::from($s7));
        let t_b  = mul16(C7S1, i32::from($s1)) - mul16(C1S7, i32::from($s7));
        let t_c  = mul16(C3S5, i32::from($s3)) + mul16(C5S3, i32::from($s5));
        let t_d  = mul16(C3S5, i32::from($s5)) - mul16(C5S3, i32::from($s3));
        let t_a1 = mul16(C4S4, t_a - t_c);
        let t_b1 = mul16(C4S4, t_b - t_d);
        let t_c  = t_a + t_c;
        let t_d  = t_b + t_d;
        let t_e  = mul16(C4S4, i32::from($s0 + $s4)) + $bias;
        let t_f  = mul16(C4S4, i32::from($s0 - $s4)) + $bias;
        let t_g  = mul16(C2S6, i32::from($s2)) + mul16(C6S2, i32::from($s6));
        let t_h  = mul16(C6S2, i32::from($s2)) - mul16(C2S6, i32::from($s6));
        let t_e1 = t_e  - t_g;
        let t_g  = t_e  + t_g;
        let t_a  = t_f  + t_a1;
        let t_f  = t_f  - t_a1;
        let t_b  = t_b1 - t_h;
        let t_h  = t_b1 + t_h;

        $d0 = ((t_g  + t_c) >> $shift) as $otype;
        $d7 = ((t_g  - t_c) >> $shift) as $otype;
        $d1 = ((t_a  + t_h) >> $shift) as $otype;
        $d2 = ((t_a  - t_h) >> $shift) as $otype;
        $d3 = ((t_e1 + t_d) >> $shift) as $otype;
        $d4 = ((t_e1 - t_d) >> $shift) as $otype;
        $d5 = ((t_f  + t_b) >> $shift) as $otype;
        $d6 = ((t_f  - t_b) >> $shift) as $otype;
    }
}

pub fn vp_idct(coeffs: &mut [i16; 64]) {
    let mut tmp = [0i32; 64];
    for (src, dst) in coeffs.chunks(8).zip(tmp.chunks_mut(8)) {
        idct_step!(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
                   dst[0], dst[1], dst[2], dst[3], dst[4], dst[5], dst[6], dst[7], 0, 0, i32);
    }
    let src = &tmp;
    let dst = coeffs;
    for i in 0..8 {
        idct_step!(src[0 * 8 + i], src[1 * 8 + i], src[2 * 8 + i], src[3 * 8 + i],
                   src[4 * 8 + i], src[5 * 8 + i], src[6 * 8 + i], src[7 * 8 + i],
                   dst[0 * 8 + i], dst[1 * 8 + i], dst[2 * 8 + i], dst[3 * 8 + i],
                   dst[4 * 8 + i], dst[5 * 8 + i], dst[6 * 8 + i], dst[7 * 8 + i], 8, 4, i16);
    }
}

pub fn vp_idct_dc(coeffs: &mut [i16; 64]) {
    let dc = ((mul16(C4S4, mul16(C4S4, i32::from(coeffs[0]))) + 8) >> 4) as i16;
    for coef in coeffs.iter_mut() {
        *coef = dc;
    }
}

pub fn unquant(coeffs: &mut [i16; 64], qmat: &[i16; 64]) {
    for i in 1..64 {
        coeffs[i] = coeffs[i].wrapping_mul(qmat[i]);
    }
}

pub fn vp_put_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
    vp_idct(coeffs);
    let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
    for y in 0..8 {
        for x in 0..8 {
            frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
        }
        off += frm.stride[plane];
    }
}

pub fn vp_put_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
    vp_idct(coeffs);
    let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
    for y in 0..8 {
        for x in 0..8 {
            frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
        }
        off += frm.stride[plane] * 2;
    }
}

pub fn vp_put_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
    vp_idct_dc(coeffs);
    let dc = (coeffs[0] + 128).min(255).max(0) as u8;
    let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
    for _ in 0..8 {
        for x in 0..8 {
            frm.data[off + x] = dc;
        }
        off += frm.stride[plane];
    }
}

pub fn vp_add_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
    vp_idct(coeffs);
    let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
    for y in 0..8 {
        for x in 0..8 {
            frm.data[off + x] = (coeffs[x + y * 8] + i16::from(frm.data[off + x])).min(255).max(0) as u8;
        }
        off += frm.stride[plane];
    }
}

pub fn vp_add_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
    vp_idct(coeffs);
    let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
    for y in 0..8 {
        for x in 0..8 {
            frm.data[off + x] = (coeffs[x + y * 8] + i16::from(frm.data[off + x])).min(255).max(0) as u8;
        }
        off += frm.stride[plane] * 2;
    }
}

pub fn vp_add_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
    vp_idct_dc(coeffs);
    let dc = coeffs[0];
    let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
    for _ in 0..8 {
        for x in 0..8 {
            frm.data[off + x] = (dc + i16::from(frm.data[off + x])).min(255).max(0) as u8;
        }
        off += frm.stride[plane];
    }
}

pub fn vp31_loop_filter(data: &mut [u8], mut off: usize, step: usize, stride: usize,
                        len: usize, loop_str: i16) {
    for _ in 0..len {
        let a = i16::from(data[off - step * 2]);
        let b = i16::from(data[off - step]);
        let c = i16::from(data[off]);
        let d = i16::from(data[off + step]);
        let mut diff = ((a - d) + 3 * (c - b) + 4) >> 3;
        if diff.abs() >= 2 * loop_str {
            diff = 0;
        } else if diff.abs() >= loop_str {
            if diff < 0 {
                diff = -diff - 2 * loop_str;
            } else {
                diff = -diff + 2 * loop_str;
            }
        }
        if diff != 0 {
            data[off - step] = (b + diff).max(0).min(255) as u8;
            data[off]        = (c - diff).max(0).min(255) as u8;
        }

        off += stride;
    }
}

pub fn vp_copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, comp: usize,
                     dx: usize, dy: usize, mv_x: i16, mv_y: i16,
                     preborder: usize, postborder: usize, loop_str: i16,
                     mode: usize, interp: &[BlkInterpFunc], mut mc_buf: NAVideoBufferRef<u8>)
{
    let sx = (dx as isize) + (mv_x as isize);
    let sy = (dy as isize) + (mv_y as isize);
    if ((sx | sy) & 7) == 0 {
        copy_block(dst, src, comp, dx, dy, mv_x, mv_y, 8, 8, preborder, postborder, mode, interp);
        return;
    }
    let pre = preborder.max(2);
    let post = postborder.max(1);
    let bsize = 8 + pre + post;
    let src_x = sx - (pre as isize);
    let src_y = sy - (pre as isize);
    {
        let tmp_buf = NASimpleVideoFrame::from_video_buf(&mut mc_buf).unwrap();
        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);
//        copy_block(&mut tmp_buf, src, comp, 0, 0, src_x as i16, src_y as i16,
//                   bsize, bsize, 0, 0, 0, interp);
        if (sx & 7) != 0 {
            let foff = (8 - (sx & 7)) as usize;
            let off = pre + foff + tmp_buf.offset[comp];
            vp31_loop_filter(tmp_buf.data, off, 1, tmp_buf.stride[comp], bsize, loop_str);
        }
        if (sy & 7) != 0 {
            let foff = (8 - (sy & 7)) as usize;
            let off = (pre + foff) * tmp_buf.stride[comp] + tmp_buf.offset[comp];
            vp31_loop_filter(tmp_buf.data, off, tmp_buf.stride[comp], 1, bsize, loop_str);
        }
    }
    let dxoff = (pre as i16) - (dx as i16);
    let dyoff = (pre as i16) - (dy as i16);
    copy_block(dst, mc_buf, comp, dx, dy, dxoff, dyoff, 8, 8, preborder, postborder, mode, interp);
}

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 {
        dst[didx..][..bw].copy_from_slice(&src[sidx..][..bw]);
        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] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + 1])) >> 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] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + sstride])) >> 1) as u8; }
        didx += dstride;
        sidx += sstride;
    }
}

fn vp3_interp1x(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] = ((u16::from(src[sidx + x]) +
                              u16::from(src[sidx + x + sstride + 1])) >> 1) as u8;
        }
        didx += dstride;
        sidx += sstride;
    }
}

fn vp3_interp1y(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] = ((u16::from(src[sidx + x + 1]) +
                              u16::from(src[sidx + x + sstride])) >> 1) as u8;
        }
        didx += dstride;
        sidx += sstride;
    }
}

pub const VP3_INTERP_FUNCS: &[blockdsp::BlkInterpFunc] = &[ vp3_interp00, vp3_interp01, vp3_interp10, vp3_interp1x, vp3_interp1y ];
Commit	Line	Data
5b24175d	1	use nihav_core::codecs::*;
b4d5b851 KS	2	use nihav_codec_support::codecs::blockdsp;
b4d5b851 KS	3	use nihav_codec_support::codecs::blockdsp::*;
5b24175d	4
93bbc2b0 KS	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
e6aaad5c	20	#[derive(Clone,Copy,Debug,PartialEq,Default)]
5b24175d	21	pub enum VPMBType {
e6aaad5c	22	#[default]
5b24175d KS	23	Intra,
	24	InterNoMV,
	25	InterMV,
	26	InterNearest,
	27	InterNear,
	28	InterFourMV,
	29	GoldenNoMV,
	30	GoldenMV,
	31	GoldenNearest,
	32	GoldenNear,
	33	}
	34
3584b223 KS	35	pub const VP_REF_INTER: u8 = 1;
	36	pub const VP_REF_GOLDEN: u8 = 2;
	37
5b24175d KS	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 \|
3584b223 KS	47	VPMBType::InterFourMV => VP_REF_INTER,
3584b223 KS	48	_ => VP_REF_GOLDEN,
5b24175d KS	49	}
	50	}
	51	}
	52
5b24175d KS	53	#[derive(Default)]
	54	pub struct VPShuffler {
	55	lastframe: Option<NAVideoBufferRef<u8>>,
	56	goldframe: Option<NAVideoBufferRef<u8>>,
	57	}
	58
	59	impl VPShuffler {
	60	pub fn new() -> Self { VPShuffler { lastframe: None, goldframe: None } }
	61	pub fn clear(&mut self) { self.lastframe = None; self.goldframe = None; }
	62	pub fn add_frame(&mut self, buf: NAVideoBufferRef<u8>) {
	63	self.lastframe = Some(buf);
	64	}
	65	pub fn add_golden_frame(&mut self, buf: NAVideoBufferRef<u8>) {
	66	self.goldframe = Some(buf);
	67	}
	68	pub fn get_last(&mut self) -> Option<NAVideoBufferRef<u8>> {
e6aaad5c	69	self.lastframe.as_ref().cloned()
5b24175d KS	70	}
5b24175d KS	71	pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
e6aaad5c	72	self.goldframe.as_ref().cloned()
5b24175d	73	}
6e24ec0b KS	74	pub fn has_refs(&self) -> bool {
	75	self.lastframe.is_some()
	76	}
5b24175d KS	77	}
5b24175d KS	78
3f67638d KS	79	pub const VP56_COEF_BASE: [i16; 6] = [ 5, 7, 11, 19, 35, 67 ];
	80	pub const VP56_COEF_ADD_PROBS: [[u8; 12]; 6] = [
	81	[ 159, 128, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ],
	82	[ 165, 145, 128, 0, 0, 0, 0, 0, 0, 0, 0, 0 ],
	83	[ 173, 148, 140, 128, 0, 0, 0, 0, 0, 0, 0, 0 ],
	84	[ 176, 155, 140, 135, 128, 0, 0, 0, 0, 0, 0, 0 ],
	85	[ 180, 157, 141, 134, 130, 128, 0, 0, 0, 0, 0, 0 ],
	86	[ 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 128 ],
	87	];
	88
3584b223 KS	89	pub struct BoolCoder<'a> {
	90	pub src: &'a [u8],
	91	pos: usize,
	92	value: u32,
	93	range: u32,
	94	bits: i32,
	95	}
	96
3584b223 KS	97	impl<'a> BoolCoder<'a> {
	98	pub fn new(src: &'a [u8]) -> DecoderResult<Self> {
	99	if src.len() < 3 { return Err(DecoderError::ShortData); }
47933c6d	100	let value = (u32::from(src[0]) << 24) \| (u32::from(src[1]) << 16) \| (u32::from(src[2]) << 8) \| u32::from(src[3]);
3584b223 KS	101	Ok(Self { src, pos: 4, value, range: 255, bits: 8 })
	102	}
	103	pub fn read_bool(&mut self) -> bool {
	104	self.read_prob(128)
	105	}
	106	pub fn read_prob(&mut self, prob: u8) -> bool {
	107	self.renorm();
47933c6d	108	let split = 1 + (((self.range - 1) * u32::from(prob)) >> 8);
3584b223 KS	109	let bit;
	110	if self.value < (split << 24) {
	111	self.range = split;
	112	bit = false;
	113	} else {
	114	self.range -= split;
	115	self.value -= split << 24;
	116	bit = true;
	117	}
	118	bit
	119	}
	120	pub fn read_bits(&mut self, bits: u8) -> u32 {
	121	let mut val = 0u32;
	122	for _ in 0..bits {
	123	val = (val << 1) \| (self.read_prob(128) as u32);
	124	}
	125	val
	126	}
587a6d78 KS	127	pub fn read_byte(&mut self) -> u8 {
	128	let mut val = 0u8;
	129	for _ in 0..8 {
	130	val = (val << 1) \| (self.read_prob(128) as u8);
	131	}
	132	val
	133	}
	134	pub fn read_sbits(&mut self, bits: u8) -> i32 {
	135	let mut val = if self.read_prob(128) { -1i32 } else { 0i32 };
	136	for _ in 1..bits {
	137	val = (val << 1) \| (self.read_prob(128) as i32);
	138	}
	139	val
	140	}
3584b223 KS	141	pub fn read_probability(&mut self) -> u8 {
	142	let val = self.read_bits(7) as u8;
	143	if val == 0 {
	144	1
	145	} else {
	146	val << 1
	147	}
	148	}
	149	fn renorm(&mut self) {
	150	let shift = self.range.leading_zeros() & 7;
	151	self.range <<= shift;
	152	self.value <<= shift;
	153	self.bits -= shift as i32;
	154	if (self.bits <= 0) && (self.pos < self.src.len()) {
47933c6d	155	self.value \|= u32::from(self.src[self.pos]) << (-self.bits as u8);
3584b223 KS	156	self.pos += 1;
	157	self.bits += 8;
	158	}
	159	/* while self.range < 0x80 {
	160	self.range <<= 1;
	161	self.value <<= 1;
	162	self.bits -= 1;
	163	if (self.bits <= 0) && (self.pos < self.src.len()) {
47933c6d	164	self.value \|= u32::from(self.src[self.pos]);
3584b223 KS	165	self.pos += 1;
	166	self.bits = 8;
	167	}
	168	}*/
	169	}
	170	pub fn skip_bytes(&mut self, nbytes: usize) {
	171	for _ in 0..nbytes {
	172	self.value <<= 8;
	173	if self.pos < self.src.len() {
47933c6d	174	self.value \|= u32::from(self.src[self.pos]);
3584b223 KS	175	self.pos += 1;
	176	}
	177	}
	178	}
	179	}
	180
	181	#[allow(dead_code)]
	182	pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
47933c6d	183	(((i32::from(prob) * i32::from(weights[0]) + 128) >> 8) + i32::from(weights[1])).min(maxval).max(1) as u8
3584b223 KS	184	}
3584b223 KS	185
3584b223 KS	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) => { $leaf }
	195	}
	196
5b24175d KS	197	const C1S7: i32 = 64277;
	198	const C2S6: i32 = 60547;
	199	const C3S5: i32 = 54491;
	200	const C4S4: i32 = 46341;
	201	const C5S3: i32 = 36410;
	202	const C6S2: i32 = 25080;
	203	const C7S1: i32 = 12785;
	204
	205	fn mul16(a: i32, b: i32) -> i32 {
	206	(a * b) >> 16
	207	}
	208
	209	macro_rules! idct_step {
	210	($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;
20b5a55f KS	261	for coef in coeffs.iter_mut() {
20b5a55f KS	262	*coef = dc;
5b24175d KS	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
3584b223 KS	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
5b24175d KS	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 {
47933c6d	311	frm.data[off + x] = (coeffs[x + y * 8] + i16::from(frm.data[off + x])).min(255).max(0) as u8;
5b24175d KS	312	}
	313	off += frm.stride[plane];
	314	}
	315	}
	316
3584b223 KS	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 {
47933c6d	322	frm.data[off + x] = (coeffs[x + y * 8] + i16::from(frm.data[off + x])).min(255).max(0) as u8;
3584b223 KS	323	}
	324	off += frm.stride[plane] * 2;
	325	}
	326	}
	327
5b24175d KS	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 {
47933c6d	334	frm.data[off + x] = (dc + i16::from(frm.data[off + x])).min(255).max(0) as u8;
5b24175d KS	335	}
	336	off += frm.stride[plane];
	337	}
	338	}
8d8ddfe1 KS	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 {
47933c6d KS	343	let a = i16::from(data[off - step * 2]);
	344	let b = i16::from(data[off - step]);
	345	let c = i16::from(data[off]);
	346	let d = i16::from(data[off + step]);
8d8ddfe1 KS	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
8e4b2f44 KS	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	{
900c9c57	383	let tmp_buf = NASimpleVideoFrame::from_video_buf(&mut mc_buf).unwrap();
86081fed	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, 0);
900c9c57 KS	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 {
8e4b2f44	388	let foff = (8 - (sx & 7)) as usize;
900c9c57	389	let off = pre + foff + tmp_buf.offset[comp];
8e4b2f44 KS	390	vp31_loop_filter(tmp_buf.data, off, 1, tmp_buf.stride[comp], bsize, loop_str);
8e4b2f44 KS	391	}
900c9c57	392	if (sy & 7) != 0 {
547a8074	393	let foff = (8 - (sy & 7)) as usize;
900c9c57	394	let off = (pre + foff) * tmp_buf.stride[comp] + tmp_buf.offset[comp];
547a8074 KS	395	vp31_loop_filter(tmp_buf.data, off, tmp_buf.stride[comp], 1, bsize, loop_str);
547a8074 KS	396	}
8e4b2f44 KS	397	}
	398	let dxoff = (pre as i16) - (dx as i16);
	399	let dyoff = (pre as i16) - (dy as i16);
547a8074	400	copy_block(dst, mc_buf, comp, dx, dy, dxoff, dyoff, 8, 8, preborder, postborder, mode, interp);
8e4b2f44	401	}
3584b223 KS	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 {
47933c6d	408	dst[didx..][..bw].copy_from_slice(&src[sidx..][..bw]);
3584b223 KS	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 {
47933c6d	419	for x in 0..bw { dst[didx + x] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + 1])) >> 1) as u8; }
3584b223 KS	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 {
47933c6d	430	for x in 0..bw { dst[didx + x] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + sstride])) >> 1) as u8; }
3584b223 KS	431	didx += dstride;
	432	sidx += sstride;
	433	}
	434	}
	435
3cc76ad5	436	fn vp3_interp1x(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
3584b223 KS	437	{
	438	let mut didx = 0;
	439	let mut sidx = 0;
	440	for _ in 0..bh {
	441	for x in 0..bw {
47933c6d KS	442	dst[didx + x] = ((u16::from(src[sidx + x]) +
47933c6d KS	443	u16::from(src[sidx + x + sstride + 1])) >> 1) as u8;
3584b223 KS	444	}
	445	didx += dstride;
	446	sidx += sstride;
	447	}
	448	}
	449
3cc76ad5 KS	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 {
47933c6d KS	456	dst[didx + x] = ((u16::from(src[sidx + x + 1]) +
47933c6d KS	457	u16::from(src[sidx + x + sstride])) >> 1) as u8;
3cc76ad5 KS	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 ];
3584b223	465