| 1 | use nihav_codec_support::codecs::blockdsp::*; |
| 2 | |
| 3 | use super::clip_u8; |
| 4 | |
| 5 | const TMP_BUF_STRIDE: usize = 32; |
| 6 | |
| 7 | fn interp_block1(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool, avg0: bool) { |
| 8 | unsafe { |
| 9 | let step = if hor { 1 } else { sstride }; |
| 10 | let avgidx = if avg0 { step * 2 } else { step * 3 }; |
| 11 | let mut src = src.as_ptr(); |
| 12 | let mut dst = dst.as_mut_ptr(); |
| 13 | for _ in 0..h { |
| 14 | for _ in 0..w { |
| 15 | let t = clip_u8(( i16::from(*src) |
| 16 | - 5 * i16::from(*src.add(step)) |
| 17 | + 20 * i16::from(*src.add(step * 2)) |
| 18 | + 20 * i16::from(*src.add(step * 3)) |
| 19 | - 5 * i16::from(*src.add(step * 4)) |
| 20 | + i16::from(*src.add(step * 5)) |
| 21 | + 16) >> 5); |
| 22 | *dst = ((u16::from(t) + u16::from(*src.add(avgidx)) + 1) >> 1) as u8; |
| 23 | src = src.add(1); |
| 24 | dst = dst.add(1); |
| 25 | } |
| 26 | dst = dst.sub(w).add(dstride); |
| 27 | src = src.sub(w).add(sstride); |
| 28 | } |
| 29 | } |
| 30 | } |
| 31 | |
| 32 | fn interp_block2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool) { |
| 33 | unsafe { |
| 34 | let step = if hor { 1 } else { sstride }; |
| 35 | let mut pix = dst.as_mut_ptr(); |
| 36 | let mut src = src.as_ptr(); |
| 37 | for _ in 0..h { |
| 38 | for x in 0..w { |
| 39 | *pix.add(x) = clip_u8(( i16::from(*src) |
| 40 | - 5 * i16::from(*src.add(step)) |
| 41 | + 20 * i16::from(*src.add(step * 2)) |
| 42 | + 20 * i16::from(*src.add(step * 3)) |
| 43 | - 5 * i16::from(*src.add(step * 4)) |
| 44 | + i16::from(*src.add(step * 5)) |
| 45 | + 16) >> 5); |
| 46 | src = src.add(1); |
| 47 | } |
| 48 | pix = pix.add(dstride); |
| 49 | src = src.sub(w); |
| 50 | src = src.add(sstride); |
| 51 | } |
| 52 | } |
| 53 | } |
| 54 | |
| 55 | fn mc_avg_tmp(dst: &mut [u8], dstride: usize, w: usize, h: usize, tmp: &[u8], tmp2: &[u8]) { |
| 56 | unsafe { |
| 57 | let mut src1 = tmp.as_ptr(); |
| 58 | let mut src2 = tmp2.as_ptr(); |
| 59 | let mut dst = dst.as_mut_ptr(); |
| 60 | for _ in 0..h { |
| 61 | for x in 0..w { |
| 62 | let a = *src1.add(x); |
| 63 | let b = *src2.add(x); |
| 64 | *dst.add(x) = ((u16::from(a) + u16::from(b) + 1) >> 1) as u8; |
| 65 | } |
| 66 | dst = dst.add(dstride); |
| 67 | src1 = src1.add(TMP_BUF_STRIDE); |
| 68 | src2 = src2.add(TMP_BUF_STRIDE); |
| 69 | } |
| 70 | } |
| 71 | } |
| 72 | |
| 73 | fn h264_mc00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 74 | unsafe { |
| 75 | let mut src = src.as_ptr(); |
| 76 | let mut dst = dst.as_mut_ptr(); |
| 77 | for _ in 0..h { |
| 78 | std::ptr::copy_nonoverlapping(src, dst, w); |
| 79 | src = src.add(sstride); |
| 80 | dst = dst.add(dstride); |
| 81 | } |
| 82 | } |
| 83 | } |
| 84 | |
| 85 | fn h264_mc01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 86 | interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, true); |
| 87 | } |
| 88 | |
| 89 | fn h264_mc02(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 90 | interp_block2(dst, dstride, &src[sstride * 2..], sstride, w, h, true); |
| 91 | } |
| 92 | |
| 93 | fn h264_mc03(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 94 | interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, false); |
| 95 | } |
| 96 | |
| 97 | fn h264_mc10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 98 | interp_block1(dst, dstride, &src[2..], sstride, w, h, false, true); |
| 99 | } |
| 100 | |
| 101 | fn h264_mc11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 102 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 103 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 104 | h264_mc02(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 105 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h); |
| 106 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 107 | } |
| 108 | |
| 109 | fn h264_mc12(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 110 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 111 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 112 | h264_mc02(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 113 | h264_mc22(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h); |
| 114 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 115 | } |
| 116 | |
| 117 | fn h264_mc13(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 118 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 119 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 120 | h264_mc02(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 121 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h); |
| 122 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 123 | } |
| 124 | |
| 125 | fn h264_mc20(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 126 | interp_block2(dst, dstride, &src[2..], sstride, w, h, false); |
| 127 | } |
| 128 | |
| 129 | fn h264_mc21(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 130 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 131 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 132 | h264_mc22(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 133 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h); |
| 134 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 135 | } |
| 136 | |
| 137 | fn h264_mc22(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 138 | let mut tmp: [i32; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 139 | unsafe { |
| 140 | let mut src = src.as_ptr(); |
| 141 | let mut dst = tmp.as_mut_ptr(); |
| 142 | for _ in 0..h { |
| 143 | for _ in 0..w+5 { |
| 144 | *dst = i32::from(*src) |
| 145 | - 5 * i32::from(*src.add(sstride)) |
| 146 | + 20 * i32::from(*src.add(sstride * 2)) |
| 147 | + 20 * i32::from(*src.add(sstride * 3)) |
| 148 | - 5 * i32::from(*src.add(sstride * 4)) |
| 149 | + i32::from(*src.add(sstride * 5)); |
| 150 | dst = dst.add(1); |
| 151 | src = src.add(1); |
| 152 | } |
| 153 | src = src.sub(w+5).add(sstride); |
| 154 | dst = dst.sub(w+5).add(TMP_BUF_STRIDE); |
| 155 | } |
| 156 | } |
| 157 | unsafe { |
| 158 | let mut dst = dst.as_mut_ptr(); |
| 159 | let mut src = tmp.as_ptr(); |
| 160 | for _ in 0..h { |
| 161 | for _ in 0..w { |
| 162 | *dst = clip_u8(((*src - 5 * *src.add(1) + 20 * *src.add(2) + 20 * *src.add(3) - 5 * *src.add(4) + *src.add(5) + 512) >> 10) as i16); |
| 163 | dst = dst.add(1); |
| 164 | src = src.add(1); |
| 165 | } |
| 166 | dst = dst.sub(w).add(dstride); |
| 167 | src = src.sub(w).add(TMP_BUF_STRIDE); |
| 168 | } |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | fn h264_mc23(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 173 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 174 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 175 | h264_mc22(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 176 | h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h); |
| 177 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 178 | } |
| 179 | |
| 180 | fn h264_mc30(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 181 | interp_block1(dst, dstride, &src[2..], sstride, w, h, false, false); |
| 182 | } |
| 183 | |
| 184 | fn h264_mc31(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 185 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 186 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 187 | h264_mc20(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 188 | h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h); |
| 189 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 190 | } |
| 191 | |
| 192 | fn h264_mc32(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 193 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 194 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 195 | h264_mc22(&mut tmp, TMP_BUF_STRIDE, src, sstride, w, h); |
| 196 | h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h); |
| 197 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 198 | } |
| 199 | |
| 200 | fn h264_mc33(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) { |
| 201 | let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 202 | let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() }; |
| 203 | h264_mc20(&mut tmp, TMP_BUF_STRIDE, &src[1..], sstride, w, h); |
| 204 | h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h); |
| 205 | mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2); |
| 206 | } |
| 207 | |
| 208 | |
| 209 | pub fn chroma_interp(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, w: usize, h: usize) { |
| 210 | let a0 = 8 - dx; |
| 211 | let a1 = dx; |
| 212 | let b0 = 8 - dy; |
| 213 | let b1 = dy; |
| 214 | |
| 215 | if a0 == 8 && b0 == 8 { |
| 216 | unsafe { |
| 217 | let mut src = src.as_ptr(); |
| 218 | let mut dst = dst.as_mut_ptr(); |
| 219 | for _ in 0..h { |
| 220 | std::ptr::copy_nonoverlapping(src, dst, w); |
| 221 | src = src.add(sstride); |
| 222 | dst = dst.add(dstride); |
| 223 | } |
| 224 | } |
| 225 | } else if a0 == 8 { |
| 226 | unsafe { |
| 227 | let mut src0 = src.as_ptr(); |
| 228 | let mut src1 = src0.add(sstride); |
| 229 | let mut dst = dst.as_mut_ptr(); |
| 230 | for _ in 0..h { |
| 231 | for x in 0..w { |
| 232 | let a = *src0.add(x); |
| 233 | let b = *src1.add(x); |
| 234 | *dst.add(x) = ((u16::from(a) * b0 + u16::from(b) * b1 + 4) >> 3) as u8; |
| 235 | } |
| 236 | src0 = src0.add(sstride); |
| 237 | src1 = src1.add(sstride); |
| 238 | dst = dst.add(dstride); |
| 239 | } |
| 240 | } |
| 241 | } else if b0 == 8 { |
| 242 | unsafe { |
| 243 | let mut src = src.as_ptr(); |
| 244 | let mut dst = dst.as_mut_ptr(); |
| 245 | for _ in 0..h { |
| 246 | let mut a = *src; |
| 247 | for x in 0..w { |
| 248 | let b = *src.add(x + 1); |
| 249 | *dst.add(x) = ((u16::from(a) * a0 + u16::from(b) * a1 + 4) >> 3) as u8; |
| 250 | a = b; |
| 251 | } |
| 252 | src = src.add(sstride); |
| 253 | dst = dst.add(dstride); |
| 254 | } |
| 255 | } |
| 256 | } else { |
| 257 | unsafe { |
| 258 | let mut src0 = src.as_ptr(); |
| 259 | let mut src1 = src0.add(sstride); |
| 260 | let mut dst = dst.as_mut_ptr(); |
| 261 | for _ in 0..h { |
| 262 | let mut a = *src0; |
| 263 | let mut c = *src1; |
| 264 | for x in 0..w { |
| 265 | let b = *src0.add(x + 1); |
| 266 | let d = *src1.add(x + 1); |
| 267 | *dst.add(x) = ((u16::from(a) * a0 * b0 + u16::from(b) * a1 * b0 + u16::from(c) * a0 * b1 + u16::from(d) * a1 * b1 + 0x20) >> 6) as u8; |
| 268 | a = b; |
| 269 | c = d; |
| 270 | } |
| 271 | src0 = src0.add(sstride); |
| 272 | src1 = src1.add(sstride); |
| 273 | dst = dst.add(dstride); |
| 274 | } |
| 275 | } |
| 276 | } |
| 277 | } |
| 278 | |
| 279 | pub const H264_LUMA_INTERP: &[BlkInterpFunc] = &[ |
| 280 | h264_mc00, h264_mc01, h264_mc02, h264_mc03, |
| 281 | h264_mc10, h264_mc11, h264_mc12, h264_mc13, |
| 282 | h264_mc20, h264_mc21, h264_mc22, h264_mc23, |
| 283 | h264_mc30, h264_mc31, h264_mc32, h264_mc33 |
| 284 | ]; |