[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)]
#[allow(dead_code)]
pub enum VPMBType {
    Intra,
    InterNoMV,
    InterMV,
    InterNearest,
    InterNear,
    InterFourMV,
    GoldenNoMV,
    GoldenMV,
    GoldenNearest,
    GoldenNear,
}

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

#[allow(dead_code)]
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,
        }
    }
}

impl Default for VPMBType {
    fn default() -> Self { VPMBType::Intra }
}

#[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>> {
        if let Some(ref frm) = self.lastframe {
            Some(frm.clone())
        } else {
            None
        }
    }
    pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
        if let Some(ref frm) = self.goldframe {
            Some(frm.clone())
        } else {
            None
        }
    }
    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 ],
];

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

#[allow(dead_code)]
impl<'a> BoolCoder<'a> {
    pub fn new(src: &'a [u8]) -> DecoderResult<Self> {
        if src.len() < 3 { return Err(DecoderError::ShortData); }
        let value = ((src[0] as u32) << 24) | ((src[1] as u32) << 16) | ((src[2] as u32) << 8) | (src[3] as u32);
        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) * (prob as u32)) >> 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 |= (self.src[self.pos] as u32) << (-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 |= self.src[self.pos] as u32;
                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 |= self.src[self.pos] as u32;
                self.pos += 1;
            }
        }
    }
}

#[allow(dead_code)]
pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
    ((((prob as i32) * (weights[0] as i32) + 128) >> 8) + (weights[1] as i32)).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 i in 0..64 {
        coeffs[i] = 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] + (frm.data[off + x] as i16)).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] + (frm.data[off + x] as i16)).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 + (frm.data[off + x] as i16)).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 = data[off - step * 2] as i16;
        let b = data[off - step] as i16;
        let c = data[off] as i16;
        let d = data[off + step] as i16;
        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);
//        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 {
        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_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] = (((src[sidx + x] as u16) +
                              (src[sidx + x + sstride + 1] as u16)) >> 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] = (((src[sidx + x + 1] as u16) +
                              (src[sidx + x + sstride] as u16)) >> 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
5b24175d KS	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
3584b223 KS	35	pub const VP_REF_INTER: u8 = 1;
	36	pub const VP_REF_GOLDEN: u8 = 2;
	37
5b24175d KS	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 \|
3584b223 KS	48	VPMBType::InterFourMV => VP_REF_INTER,
3584b223 KS	49	_ => VP_REF_GOLDEN,
5b24175d KS	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	}
6e24ec0b KS	87	pub fn has_refs(&self) -> bool {
	88	self.lastframe.is_some()
	89	}
5b24175d KS	90	}
5b24175d KS	91
3f67638d KS	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
3584b223 KS	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 = ((src[0] as u32) << 24) \| ((src[1] as u32) << 16) \| ((src[2] as u32) << 8) \| (src[3] as u32);
	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) * (prob as u32)) >> 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	}
587a6d78 KS	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	}
3584b223 KS	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 \|= (self.src[self.pos] as u32) << (-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 \|= self.src[self.pos] as u32;
	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 \|= self.src[self.pos] as u32;
	190	self.pos += 1;
	191	}
	192	}
	193	}
	194	}
	195
	196	#[allow(dead_code)]
	197	pub fn rescale_prob(prob: u8, weights: &[i16; 2], maxval: i32) -> u8 {
	198	((((prob as i32) * (weights[0] as i32) + 128) >> 8) + (weights[1] as i32)).min(maxval).max(1) as u8
	199	}
	200
3584b223 KS	201	macro_rules! vp_tree {
	202	($bc: expr, $prob: expr, $node1: expr, $node2: expr) => {
	203	if !$bc.read_prob($prob) {
	204	$node1
	205	} else {
	206	$node2
	207	}
	208	};
	209	($leaf: expr) => { $leaf }
	210	}
	211
5b24175d KS	212	const C1S7: i32 = 64277;
	213	const C2S6: i32 = 60547;
	214	const C3S5: i32 = 54491;
	215	const C4S4: i32 = 46341;
	216	const C5S3: i32 = 36410;
	217	const C6S2: i32 = 25080;
	218	const C7S1: i32 = 12785;
	219
	220	fn mul16(a: i32, b: i32) -> i32 {
	221	(a * b) >> 16
	222	}
	223
	224	macro_rules! idct_step {
	225	($s0:expr, $s1:expr, $s2:expr, $s3:expr, $s4:expr, $s5:expr, $s6:expr, $s7:expr,
	226	$d0:expr, $d1:expr, $d2:expr, $d3:expr, $d4:expr, $d5:expr, $d6:expr, $d7:expr,
	227	$bias:expr, $shift:expr, $otype:ty) => {
	228	let t_a = mul16(C1S7, i32::from($s1)) + mul16(C7S1, i32::from($s7));
	229	let t_b = mul16(C7S1, i32::from($s1)) - mul16(C1S7, i32::from($s7));
	230	let t_c = mul16(C3S5, i32::from($s3)) + mul16(C5S3, i32::from($s5));
	231	let t_d = mul16(C3S5, i32::from($s5)) - mul16(C5S3, i32::from($s3));
	232	let t_a1 = mul16(C4S4, t_a - t_c);
	233	let t_b1 = mul16(C4S4, t_b - t_d);
	234	let t_c = t_a + t_c;
	235	let t_d = t_b + t_d;
	236	let t_e = mul16(C4S4, i32::from($s0 + $s4)) + $bias;
	237	let t_f = mul16(C4S4, i32::from($s0 - $s4)) + $bias;
	238	let t_g = mul16(C2S6, i32::from($s2)) + mul16(C6S2, i32::from($s6));
	239	let t_h = mul16(C6S2, i32::from($s2)) - mul16(C2S6, i32::from($s6));
	240	let t_e1 = t_e - t_g;
	241	let t_g = t_e + t_g;
	242	let t_a = t_f + t_a1;
	243	let t_f = t_f - t_a1;
	244	let t_b = t_b1 - t_h;
	245	let t_h = t_b1 + t_h;
	246
	247	$d0 = ((t_g + t_c) >> $shift) as $otype;
	248	$d7 = ((t_g - t_c) >> $shift) as $otype;
	249	$d1 = ((t_a + t_h) >> $shift) as $otype;
	250	$d2 = ((t_a - t_h) >> $shift) as $otype;
	251	$d3 = ((t_e1 + t_d) >> $shift) as $otype;
	252	$d4 = ((t_e1 - t_d) >> $shift) as $otype;
	253	$d5 = ((t_f + t_b) >> $shift) as $otype;
	254	$d6 = ((t_f - t_b) >> $shift) as $otype;
	255	}
	256	}
	257
	258	pub fn vp_idct(coeffs: &mut [i16; 64]) {
	259	let mut tmp = [0i32; 64];
	260	for (src, dst) in coeffs.chunks(8).zip(tmp.chunks_mut(8)) {
	261	idct_step!(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
	262	dst[0], dst[1], dst[2], dst[3], dst[4], dst[5], dst[6], dst[7], 0, 0, i32);
	263	}
	264	let src = &tmp;
	265	let dst = coeffs;
	266	for i in 0..8 {
	267	idct_step!(src[0 * 8 + i], src[1 * 8 + i], src[2 * 8 + i], src[3 * 8 + i],
	268	src[4 * 8 + i], src[5 * 8 + i], src[6 * 8 + i], src[7 * 8 + i],
	269	dst[0 * 8 + i], dst[1 * 8 + i], dst[2 * 8 + i], dst[3 * 8 + i],
	270	dst[4 * 8 + i], dst[5 * 8 + i], dst[6 * 8 + i], dst[7 * 8 + i], 8, 4, i16);
	271	}
	272	}
	273
	274	pub fn vp_idct_dc(coeffs: &mut [i16; 64]) {
	275	let dc = ((mul16(C4S4, mul16(C4S4, i32::from(coeffs[0]))) + 8) >> 4) as i16;
276	for i in 0..64 {
277	coeffs[i] = dc;
278	}
279	}
280
281	pub fn unquant(coeffs: &mut [i16; 64], qmat: &[i16; 64]) {
282	for i in 1..64 {
283	coeffs[i] = coeffs[i].wrapping_mul(qmat[i]);
284	}
285	}
286
287	pub fn vp_put_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
288	vp_idct(coeffs);
289	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
290	for y in 0..8 {
291	for x in 0..8 {
292	frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
293	}
294	off += frm.stride[plane];
295	}
296	}
297
3584b223 KS	298	pub fn vp_put_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
	299	vp_idct(coeffs);
	300	let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
	301	for y in 0..8 {
	302	for x in 0..8 {
	303	frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
	304	}
	305	off += frm.stride[plane] * 2;
	306	}
	307	}
	308
5b24175d KS	309	pub fn vp_put_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
	310	vp_idct_dc(coeffs);
	311	let dc = (coeffs[0] + 128).min(255).max(0) as u8;
	312	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
	313	for _ in 0..8 {
	314	for x in 0..8 {
	315	frm.data[off + x] = dc;
	316	}
	317	off += frm.stride[plane];
	318	}
	319	}
	320
	321	pub fn vp_add_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
	322	vp_idct(coeffs);
	323	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
	324	for y in 0..8 {
	325	for x in 0..8 {
	326	frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
	327	}
	328	off += frm.stride[plane];
	329	}
	330	}
	331
3584b223 KS	332	pub fn vp_add_block_ilace(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
	333	vp_idct(coeffs);
	334	let mut off = frm.offset[plane] + bx * 8 + ((by & !1) * 8 + (by & 1)) * frm.stride[plane];
	335	for y in 0..8 {
	336	for x in 0..8 {
	337	frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
	338	}
	339	off += frm.stride[plane] * 2;
	340	}
	341	}
	342
5b24175d KS	343	pub fn vp_add_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
	344	vp_idct_dc(coeffs);
	345	let dc = coeffs[0];
	346	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
	347	for _ in 0..8 {
	348	for x in 0..8 {
	349	frm.data[off + x] = (dc + (frm.data[off + x] as i16)).min(255).max(0) as u8;
	350	}
	351	off += frm.stride[plane];
	352	}
	353	}
8d8ddfe1 KS	354
	355	pub fn vp31_loop_filter(data: &mut [u8], mut off: usize, step: usize, stride: usize,
	356	len: usize, loop_str: i16) {
	357	for _ in 0..len {
	358	let a = data[off - step * 2] as i16;
	359	let b = data[off - step] as i16;
	360	let c = data[off] as i16;
	361	let d = data[off + step] as i16;
	362	let mut diff = ((a - d) + 3 * (c - b) + 4) >> 3;
	363	if diff.abs() >= 2 * loop_str {
	364	diff = 0;
	365	} else if diff.abs() >= loop_str {
	366	if diff < 0 {
	367	diff = -diff - 2 * loop_str;
	368	} else {
	369	diff = -diff + 2 * loop_str;
	370	}
	371	}
	372	if diff != 0 {
	373	data[off - step] = (b + diff).max(0).min(255) as u8;
	374	data[off] = (c - diff).max(0).min(255) as u8;
	375	}
	376
	377	off += stride;
	378	}
	379	}
	380
8e4b2f44 KS	381	pub fn vp_copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, comp: usize,
	382	dx: usize, dy: usize, mv_x: i16, mv_y: i16,
	383	preborder: usize, postborder: usize, loop_str: i16,
	384	mode: usize, interp: &[BlkInterpFunc], mut mc_buf: NAVideoBufferRef<u8>)
	385	{
	386	let sx = (dx as isize) + (mv_x as isize);
	387	let sy = (dy as isize) + (mv_y as isize);
	388	if ((sx \| sy) & 7) == 0 {
	389	copy_block(dst, src, comp, dx, dy, mv_x, mv_y, 8, 8, preborder, postborder, mode, interp);
	390	return;
	391	}
	392	let pre = preborder.max(2);
	393	let post = postborder.max(1);
	394	let bsize = 8 + pre + post;
	395	let src_x = sx - (pre as isize);
	396	let src_y = sy - (pre as isize);
	397	{
900c9c57 KS	398	let tmp_buf = NASimpleVideoFrame::from_video_buf(&mut mc_buf).unwrap();
	399	edge_emu(src.as_ref(), src_x, src_y, bsize, bsize, &mut tmp_buf.data[tmp_buf.offset[comp]..], tmp_buf.stride[comp], comp);
	400	// copy_block(&mut tmp_buf, src, comp, 0, 0, src_x as i16, src_y as i16,
	401	// bsize, bsize, 0, 0, 0, interp);
	402	if (sx & 7) != 0 {
8e4b2f44	403	let foff = (8 - (sx & 7)) as usize;
900c9c57	404	let off = pre + foff + tmp_buf.offset[comp];
8e4b2f44 KS	405	vp31_loop_filter(tmp_buf.data, off, 1, tmp_buf.stride[comp], bsize, loop_str);
8e4b2f44 KS	406	}
900c9c57	407	if (sy & 7) != 0 {
547a8074	408	let foff = (8 - (sy & 7)) as usize;
900c9c57	409	let off = (pre + foff) * tmp_buf.stride[comp] + tmp_buf.offset[comp];
547a8074 KS	410	vp31_loop_filter(tmp_buf.data, off, tmp_buf.stride[comp], 1, bsize, loop_str);
547a8074 KS	411	}
8e4b2f44 KS	412	}
	413	let dxoff = (pre as i16) - (dx as i16);
	414	let dyoff = (pre as i16) - (dy as i16);
547a8074	415	copy_block(dst, mc_buf, comp, dx, dy, dxoff, dyoff, 8, 8, preborder, postborder, mode, interp);
8e4b2f44	416	}
3584b223 KS	417
	418	fn vp3_interp00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
	419	{
	420	let mut didx = 0;
	421	let mut sidx = 0;
	422	for _ in 0..bh {
	423	for x in 0..bw { dst[didx + x] = src[sidx + x]; }
	424	didx += dstride;
	425	sidx += sstride;
	426	}
	427	}
	428
	429	fn vp3_interp01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
	430	{
	431	let mut didx = 0;
	432	let mut sidx = 0;
	433	for _ in 0..bh {
	434	for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + 1] as u16)) >> 1) as u8; }
	435	didx += dstride;
	436	sidx += sstride;
	437	}
	438	}
	439
	440	fn vp3_interp10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
	441	{
	442	let mut didx = 0;
	443	let mut sidx = 0;
	444	for _ in 0..bh {
	445	for x in 0..bw { dst[didx + x] = (((src[sidx + x] as u16) + (src[sidx + x + sstride] as u16)) >> 1) as u8; }
	446	didx += dstride;
	447	sidx += sstride;
	448	}
	449	}
	450
3cc76ad5	451	fn vp3_interp1x(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
3584b223 KS	452	{
	453	let mut didx = 0;
	454	let mut sidx = 0;
	455	for _ in 0..bh {
	456	for x in 0..bw {
	457	dst[didx + x] = (((src[sidx + x] as u16) +
3cc76ad5	458	(src[sidx + x + sstride + 1] as u16)) >> 1) as u8;
3584b223 KS	459	}
	460	didx += dstride;
	461	sidx += sstride;
	462	}
	463	}
	464
3cc76ad5 KS	465	fn vp3_interp1y(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
	466	{
	467	let mut didx = 0;
	468	let mut sidx = 0;
	469	for _ in 0..bh {
	470	for x in 0..bw {
	471	dst[didx + x] = (((src[sidx + x + 1] as u16) +
	472	(src[sidx + x + sstride] as u16)) >> 1) as u8;
	473	}
	474	didx += dstride;
	475	sidx += sstride;
	476	}
	477	}
	478
	479	pub const VP3_INTERP_FUNCS: &[blockdsp::BlkInterpFunc] = &[ vp3_interp00, vp3_interp01, vp3_interp10, vp3_interp1x, vp3_interp1y ];
3584b223	480