avg_buf: NAVideoBufferRef<u8>,
pub put_block_weighted: [fn (dst: &mut [u8], stride: usize, src: &[u8], h: usize, wparams: [i8; 3]); 4],
pub put_block_weighted2: [fn (dst: &mut [u8], stride: usize, src0: &[u8], src1: &[u8], h: usize, wparams: [i8; 5]); 4],
+ pub chroma_interp: [fn (dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, h: usize); 3],
+ avg: [fn (dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize); 4],
+
+ width: usize,
+ height: usize,
}
impl H264MC {
avg_buf,
put_block_weighted: [put_blk_w_2, put_blk_w_4, put_blk_w_8, put_blk_w_16],
put_block_weighted2: [put_blk_w2_2, put_blk_w2_4, put_blk_w2_8, put_blk_w2_16],
+ chroma_interp: [chroma_interp_2, chroma_interp_4, chroma_interp_8],
+ avg: [avg_2, avg_4, avg_8, avg_16],
+ width: 0, height: 0,
};
obj.register_simd();
obj
}
+ pub fn set_dimensions(&mut self, width: usize, height: usize) {
+ self.width = width;
+ self.height = height;
+ }
pub fn do_mc(&mut self, frm: &mut NASimpleVideoFrame<u8>, refpic: NAVideoBufferRef<u8>, xpos: usize, ypos: usize, w: usize, h: usize, mv: MV) {
let mut ebuf = [0u8; 22 * 22];
let mvx = mv.x >> 2;
let mode = ((mv.x & 3) + (mv.y & 3) * 4) as usize;
let pre = if mode != 0 { 2isize } else { 0 };
let post = if mode != 0 { 3isize } else { 0 };
- let (yw, yh) = refpic.get_dimensions(0);
+ let (yw, yh) = (self.width, self.height);
let src = refpic.get_data();
let systride = refpic.get_stride(0);
let src_x = (xpos as isize) + (mvx as isize);
};
(H264_LUMA_INTERP[wmode][mode])(&mut frm.data[frm.offset[0] + xpos + ypos * frm.stride[0]..], frm.stride[0], ysrc, ystride, h);
- let (cw, ch) = refpic.get_dimensions(1);
+ let (cw, ch) = (self.width >> 1, self.height >> 1);
let mvx = mv.x >> 3;
let mvy = mv.y >> 3;
let dx = (mv.x & 7) as u16;
};
for chroma in 1..3 {
let off = frm.offset[chroma] + xpos / 2 + (ypos / 2) * frm.stride[chroma];
- chroma_interp(&mut frm.data[off..], frm.stride[chroma], csrc[chroma - 1], cstride[chroma - 1], dx, dy, cbw, cbh);
+ (self.chroma_interp[wmode])(&mut frm.data[off..], frm.stride[chroma], csrc[chroma - 1], cstride[chroma - 1], dx, dy, cbh);
}
}
let pre = if mode != 0 { 2 } else { 0 };
let post = if mode != 0 { 3 } else { 0 };
- let (width, height) = refpic.get_dimensions(0);
+ let (width, height) = (self.width, self.height);
let sx = (xpos as isize) + ((mv.x >> 2) as isize);
let sy = (ypos as isize) + ((mv.y >> 2) as isize);
(H264_LUMA_INTERP[wmode][mode])(ydst, 16, &sbuf[saddr..], sstride, h);
}
- let (cw, ch) = refpic.get_dimensions(1);
+ let (cw, ch) = (self.width >> 1, self.height >> 1);
let mvx = mv.x >> 3;
let mvy = mv.y >> 3;
let dx = (mv.x & 7) as u16;
&src[svoff + (src_x as usize) + (src_y as usize) * svstride..]],
[sustride, svstride])
};
- chroma_interp(udst, 16, csrc[0], cstride[0], dx, dy, cbw, cbh);
- chroma_interp(vdst, 16, csrc[1], cstride[1], dx, dy, cbw, cbh);
- }
-
- pub fn avg(&mut self, dst: &mut [u8], dstride: usize, bw: usize, bh: usize, comp: usize) {
- let afrm = NASimpleVideoFrame::from_video_buf(&mut self.avg_buf).unwrap();
- let src = &afrm.data[afrm.offset[comp]..];
- let sstride = afrm.stride[comp];
- for (dline, sline) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(bh) {
- for (dst, src) in dline.iter_mut().zip(sline.iter()).take(bw) {
- *dst = ((u16::from(*dst) + u16::from(*src) + 1) >> 1) as u8;
- }
- }
+ (self.chroma_interp[wmode])(udst, 16, csrc[0], cstride[0], dx, dy, cbh);
+ (self.chroma_interp[wmode])(vdst, 16, csrc[1], cstride[1], dx, dy, cbh);
}
pub fn do_mc_avg(&mut self, frm: &mut NASimpleVideoFrame<u8>, refpic: NAVideoBufferRef<u8>, xpos: usize, ypos: usize, w: usize, h: usize, mv: MV) {
let mut afrm = NASimpleVideoFrame::from_video_buf(&mut abuf).unwrap();
let amv = MV { x: mv.x + (xpos as i16) * 4, y: mv.y + (ypos as i16) * 4 };
self.do_mc(&mut afrm, refpic, 0, 0, w, h, amv);
+ let wsize = match w {
+ 2 => 0,
+ 4 => 1,
+ 8 => 2,
+ _ => 3,
+ };
+ let src = self.avg_buf.get_data();
for comp in 0..3 {
let shift = if comp == 0 { 0 } else { 1 };
- self.avg(&mut frm.data[frm.offset[comp] + (xpos >> shift) + (ypos >> shift) * frm.stride[comp]..], frm.stride[comp], w >> shift, h >> shift, comp);
+ let sstride = self.avg_buf.get_stride(comp);
+ let soff = self.avg_buf.get_offset(comp);
+ (self.avg[wsize - shift])(&mut frm.data[frm.offset[comp] + (xpos >> shift) + (ypos >> shift) * frm.stride[comp]..], frm.stride[comp], &src[soff..], sstride, h >> shift);
}
}
}
}
+fn avg(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize) {
+ for (dline, sline) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(bh) {
+ for (dst, src) in dline.iter_mut().zip(sline.iter()).take(bw) {
+ *dst = ((u16::from(*dst) + u16::from(*src) + 1) >> 1) as u8;
+ }
+ }
+}
+
+fn avg_2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize) {
+ avg(dst, dstride, src, sstride, 2, bh);
+}
+fn avg_4(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize) {
+ avg(dst, dstride, src, sstride, 4, bh);
+}
+fn avg_8(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize) {
+ avg(dst, dstride, src, sstride, 8, bh);
+}
+fn avg_16(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize) {
+ avg(dst, dstride, src, sstride, 16, bh);
+}
+
fn put_block_weighted(dst: &mut [u8], stride: usize, src: &[u8], w: usize, h: usize, wparams: [i8; 3]) {
let weight = i16::from(wparams[0]);
let offset = i16::from(wparams[1]);
projects / nihav.git / blobdiff