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

use nihav_core::codecs::*;

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

#[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   => 1,
            _                       => 2,
        }
    }
}

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
        }
    }
}

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_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_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;
    }
}
Commit	Line	Data
5b24175d KS	1	use nihav_core::codecs::*;
	2
	3	#[derive(Clone,Copy,Debug,PartialEq)]
	4	#[allow(dead_code)]
	5	pub enum VPMBType {
	6	Intra,
	7	InterNoMV,
	8	InterMV,
	9	InterNearest,
	10	InterNear,
	11	InterFourMV,
	12	GoldenNoMV,
	13	GoldenMV,
	14	GoldenNearest,
	15	GoldenNear,
	16	}
	17
	18	#[allow(dead_code)]
	19	impl VPMBType {
	20	pub fn is_intra(self) -> bool { self == VPMBType::Intra }
	21	pub fn get_ref_id(self) -> u8 {
	22	match self {
	23	VPMBType::Intra => 0,
	24	VPMBType::InterNoMV \|
	25	VPMBType::InterMV \|
	26	VPMBType::InterNearest \|
	27	VPMBType::InterNear \|
	28	VPMBType::InterFourMV => 1,
	29	_ => 2,
	30	}
	31	}
	32	}
	33
	34	impl Default for VPMBType {
	35	fn default() -> Self { VPMBType::Intra }
	36	}
	37
	38	#[derive(Default)]
	39	pub struct VPShuffler {
	40	lastframe: Option<NAVideoBufferRef<u8>>,
	41	goldframe: Option<NAVideoBufferRef<u8>>,
	42	}
	43
	44	impl VPShuffler {
	45	pub fn new() -> Self { VPShuffler { lastframe: None, goldframe: None } }
	46	pub fn clear(&mut self) { self.lastframe = None; self.goldframe = None; }
	47	pub fn add_frame(&mut self, buf: NAVideoBufferRef<u8>) {
	48	self.lastframe = Some(buf);
	49	}
	50	pub fn add_golden_frame(&mut self, buf: NAVideoBufferRef<u8>) {
	51	self.goldframe = Some(buf);
	52	}
	53	pub fn get_last(&mut self) -> Option<NAVideoBufferRef<u8>> {
	54	if let Some(ref frm) = self.lastframe {
	55	Some(frm.clone())
	56	} else {
	57	None
	58	}
	59	}
	60	pub fn get_golden(&mut self) -> Option<NAVideoBufferRef<u8>> {
	61	if let Some(ref frm) = self.goldframe {
	62	Some(frm.clone())
	63	} else {
	64	None
65	}
66	}
67	}
68
69	const C1S7: i32 = 64277;
70	const C2S6: i32 = 60547;
71	const C3S5: i32 = 54491;
72	const C4S4: i32 = 46341;
73	const C5S3: i32 = 36410;
74	const C6S2: i32 = 25080;
75	const C7S1: i32 = 12785;
76
77	fn mul16(a: i32, b: i32) -> i32 {
78	(a * b) >> 16
79	}
80
81	macro_rules! idct_step {
82	($s0:expr, $s1:expr, $s2:expr, $s3:expr, $s4:expr, $s5:expr, $s6:expr, $s7:expr,
83	$d0:expr, $d1:expr, $d2:expr, $d3:expr, $d4:expr, $d5:expr, $d6:expr, $d7:expr,
84	$bias:expr, $shift:expr, $otype:ty) => {
85	let t_a = mul16(C1S7, i32::from($s1)) + mul16(C7S1, i32::from($s7));
86	let t_b = mul16(C7S1, i32::from($s1)) - mul16(C1S7, i32::from($s7));
87	let t_c = mul16(C3S5, i32::from($s3)) + mul16(C5S3, i32::from($s5));
88	let t_d = mul16(C3S5, i32::from($s5)) - mul16(C5S3, i32::from($s3));
89	let t_a1 = mul16(C4S4, t_a - t_c);
90	let t_b1 = mul16(C4S4, t_b - t_d);
91	let t_c = t_a + t_c;
92	let t_d = t_b + t_d;
93	let t_e = mul16(C4S4, i32::from($s0 + $s4)) + $bias;
94	let t_f = mul16(C4S4, i32::from($s0 - $s4)) + $bias;
95	let t_g = mul16(C2S6, i32::from($s2)) + mul16(C6S2, i32::from($s6));
96	let t_h = mul16(C6S2, i32::from($s2)) - mul16(C2S6, i32::from($s6));
97	let t_e1 = t_e - t_g;
98	let t_g = t_e + t_g;
99	let t_a = t_f + t_a1;
100	let t_f = t_f - t_a1;
101	let t_b = t_b1 - t_h;
102	let t_h = t_b1 + t_h;
103
104	$d0 = ((t_g + t_c) >> $shift) as $otype;
105	$d7 = ((t_g - t_c) >> $shift) as $otype;
106	$d1 = ((t_a + t_h) >> $shift) as $otype;
107	$d2 = ((t_a - t_h) >> $shift) as $otype;
108	$d3 = ((t_e1 + t_d) >> $shift) as $otype;
109	$d4 = ((t_e1 - t_d) >> $shift) as $otype;
110	$d5 = ((t_f + t_b) >> $shift) as $otype;
111	$d6 = ((t_f - t_b) >> $shift) as $otype;
112	}
113	}
114
115	pub fn vp_idct(coeffs: &mut [i16; 64]) {
116	let mut tmp = [0i32; 64];
117	for (src, dst) in coeffs.chunks(8).zip(tmp.chunks_mut(8)) {
118	idct_step!(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
119	dst[0], dst[1], dst[2], dst[3], dst[4], dst[5], dst[6], dst[7], 0, 0, i32);
120	}
121	let src = &tmp;
122	let dst = coeffs;
123	for i in 0..8 {
124	idct_step!(src[0 * 8 + i], src[1 * 8 + i], src[2 * 8 + i], src[3 * 8 + i],
125	src[4 * 8 + i], src[5 * 8 + i], src[6 * 8 + i], src[7 * 8 + i],
126	dst[0 * 8 + i], dst[1 * 8 + i], dst[2 * 8 + i], dst[3 * 8 + i],
127	dst[4 * 8 + i], dst[5 * 8 + i], dst[6 * 8 + i], dst[7 * 8 + i], 8, 4, i16);
128	}
129	}
130
131	pub fn vp_idct_dc(coeffs: &mut [i16; 64]) {
132	let dc = ((mul16(C4S4, mul16(C4S4, i32::from(coeffs[0]))) + 8) >> 4) as i16;
133	for i in 0..64 {
134	coeffs[i] = dc;
135	}
136	}
137
138	pub fn unquant(coeffs: &mut [i16; 64], qmat: &[i16; 64]) {
139	for i in 1..64 {
140	coeffs[i] = coeffs[i].wrapping_mul(qmat[i]);
141	}
142	}
143
144	pub fn vp_put_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
145	vp_idct(coeffs);
146	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
147	for y in 0..8 {
148	for x in 0..8 {
149	frm.data[off + x] = (coeffs[x + y * 8] + 128).min(255).max(0) as u8;
150	}
151	off += frm.stride[plane];
152	}
153	}
154
155	pub fn vp_put_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
156	vp_idct_dc(coeffs);
157	let dc = (coeffs[0] + 128).min(255).max(0) as u8;
158	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
159	for _ in 0..8 {
160	for x in 0..8 {
161	frm.data[off + x] = dc;
162	}
163	off += frm.stride[plane];
164	}
165	}
166
167	pub fn vp_add_block(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
168	vp_idct(coeffs);
169	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
170	for y in 0..8 {
171	for x in 0..8 {
172	frm.data[off + x] = (coeffs[x + y * 8] + (frm.data[off + x] as i16)).min(255).max(0) as u8;
173	}
174	off += frm.stride[plane];
175	}
176	}
177
178	pub fn vp_add_block_dc(coeffs: &mut [i16; 64], bx: usize, by: usize, plane: usize, frm: &mut NASimpleVideoFrame<u8>) {
179	vp_idct_dc(coeffs);
180	let dc = coeffs[0];
181	let mut off = frm.offset[plane] + bx * 8 + by * 8 * frm.stride[plane];
182	for _ in 0..8 {
183	for x in 0..8 {
184	frm.data[off + x] = (dc + (frm.data[off + x] as i16)).min(255).max(0) as u8;
185	}
186	off += frm.stride[plane];
187	}
188	}
8d8ddfe1 KS	189
	190	pub fn vp31_loop_filter(data: &mut [u8], mut off: usize, step: usize, stride: usize,
	191	len: usize, loop_str: i16) {
	192	for _ in 0..len {
	193	let a = data[off - step * 2] as i16;
	194	let b = data[off - step] as i16;
	195	let c = data[off] as i16;
	196	let d = data[off + step] as i16;
	197	let mut diff = ((a - d) + 3 * (c - b) + 4) >> 3;
	198	if diff.abs() >= 2 * loop_str {
	199	diff = 0;
	200	} else if diff.abs() >= loop_str {
	201	if diff < 0 {
	202	diff = -diff - 2 * loop_str;
	203	} else {
	204	diff = -diff + 2 * loop_str;
	205	}
	206	}
	207	if diff != 0 {
	208	data[off - step] = (b + diff).max(0).min(255) as u8;
	209	data[off] = (c - diff).max(0).min(255) as u8;
	210	}
	211
	212	off += stride;
	213	}
	214	}
	215