ref: 3b14f94924987fc66302bb8b803ad9065816f93e
dir: dav1d/src/wedge.c
/*
* Copyright © 2018, VideoLAN and dav1d authors
* Copyright © 2018, Two Orioles, LLC
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice, this
* list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#include <stdint.h>
#include <string.h>
#include "common/intops.h"
#include "src/wedge.h"
enum WedgeDirectionType {
WEDGE_HORIZONTAL = 0,
WEDGE_VERTICAL = 1,
WEDGE_OBLIQUE27 = 2,
WEDGE_OBLIQUE63 = 3,
WEDGE_OBLIQUE117 = 4,
WEDGE_OBLIQUE153 = 5,
N_WEDGE_DIRECTIONS
};
typedef struct {
enum WedgeDirectionType direction;
int x_offset;
int y_offset;
} wedge_code_type;
static const wedge_code_type wedge_codebook_16_hgtw[16] = {
{ WEDGE_OBLIQUE27, 4, 4 }, { WEDGE_OBLIQUE63, 4, 4 },
{ WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
{ WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 4 },
{ WEDGE_HORIZONTAL, 4, 6 }, { WEDGE_VERTICAL, 4, 4 },
{ WEDGE_OBLIQUE27, 4, 2 }, { WEDGE_OBLIQUE27, 4, 6 },
{ WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
{ WEDGE_OBLIQUE63, 2, 4 }, { WEDGE_OBLIQUE63, 6, 4 },
{ WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};
static const wedge_code_type wedge_codebook_16_hltw[16] = {
{ WEDGE_OBLIQUE27, 4, 4 }, { WEDGE_OBLIQUE63, 4, 4 },
{ WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
{ WEDGE_VERTICAL, 2, 4 }, { WEDGE_VERTICAL, 4, 4 },
{ WEDGE_VERTICAL, 6, 4 }, { WEDGE_HORIZONTAL, 4, 4 },
{ WEDGE_OBLIQUE27, 4, 2 }, { WEDGE_OBLIQUE27, 4, 6 },
{ WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
{ WEDGE_OBLIQUE63, 2, 4 }, { WEDGE_OBLIQUE63, 6, 4 },
{ WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};
static const wedge_code_type wedge_codebook_16_heqw[16] = {
{ WEDGE_OBLIQUE27, 4, 4 }, { WEDGE_OBLIQUE63, 4, 4 },
{ WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
{ WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 6 },
{ WEDGE_VERTICAL, 2, 4 }, { WEDGE_VERTICAL, 6, 4 },
{ WEDGE_OBLIQUE27, 4, 2 }, { WEDGE_OBLIQUE27, 4, 6 },
{ WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
{ WEDGE_OBLIQUE63, 2, 4 }, { WEDGE_OBLIQUE63, 6, 4 },
{ WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};
static uint8_t ALIGN(wedge_masks_444_32x32[2 * 16 * 32 * 32], 64);
static uint8_t ALIGN(wedge_masks_444_32x16[2 * 16 * 32 * 16], 64);
static uint8_t ALIGN(wedge_masks_444_32x8 [2 * 16 * 32 * 8], 64);
static uint8_t ALIGN(wedge_masks_444_16x32[2 * 16 * 16 * 32], 64);
static uint8_t ALIGN(wedge_masks_444_16x16[2 * 16 * 16 * 16], 64);
static uint8_t ALIGN(wedge_masks_444_16x8 [2 * 16 * 16 * 8], 64);
static uint8_t ALIGN(wedge_masks_444_8x32 [2 * 16 * 8 * 32], 64);
static uint8_t ALIGN(wedge_masks_444_8x16 [2 * 16 * 8 * 16], 64);
static uint8_t ALIGN(wedge_masks_444_8x8 [2 * 16 * 8 * 8], 64);
static uint8_t ALIGN(wedge_masks_422_16x32[2 * 16 * 16 * 32], 64);
static uint8_t ALIGN(wedge_masks_422_16x16[2 * 16 * 16 * 16], 64);
static uint8_t ALIGN(wedge_masks_422_16x8 [2 * 16 * 16 * 8], 64);
static uint8_t ALIGN(wedge_masks_422_8x32 [2 * 16 * 8 * 32], 64);
static uint8_t ALIGN(wedge_masks_422_8x16 [2 * 16 * 8 * 16], 64);
static uint8_t ALIGN(wedge_masks_422_8x8 [2 * 16 * 8 * 8], 64);
static uint8_t ALIGN(wedge_masks_422_4x32 [2 * 16 * 4 * 32], 64);
static uint8_t ALIGN(wedge_masks_422_4x16 [2 * 16 * 4 * 16], 64);
static uint8_t ALIGN(wedge_masks_422_4x8 [2 * 16 * 4 * 8], 32);
static uint8_t ALIGN(wedge_masks_420_16x16[2 * 16 * 16 * 16], 64);
static uint8_t ALIGN(wedge_masks_420_16x8 [2 * 16 * 16 * 8], 64);
static uint8_t ALIGN(wedge_masks_420_16x4 [2 * 16 * 16 * 4], 64);
static uint8_t ALIGN(wedge_masks_420_8x16 [2 * 16 * 8 * 16], 64);
static uint8_t ALIGN(wedge_masks_420_8x8 [2 * 16 * 8 * 8], 64);
static uint8_t ALIGN(wedge_masks_420_8x4 [2 * 16 * 8 * 4], 64);
static uint8_t ALIGN(wedge_masks_420_4x16 [2 * 16 * 4 * 16], 64);
static uint8_t ALIGN(wedge_masks_420_4x8 [2 * 16 * 4 * 8], 32);
static uint8_t ALIGN(wedge_masks_420_4x4 [2 * 16 * 4 * 4], 16);
const uint8_t *dav1d_wedge_masks[N_BS_SIZES][3][2][16];
static void insert_border(uint8_t *const dst, const uint8_t *const src,
const int ctr)
{
if (ctr > 4) memset(dst, 0, ctr - 4);
memcpy(dst + imax(ctr, 4) - 4, src + imax(4 - ctr, 0), imin(64 - ctr, 8));
if (ctr < 64 - 4)
memset(dst + ctr + 4, 64, 64 - 4 - ctr);
}
static void transpose(uint8_t *const dst, const uint8_t *const src) {
for (int y = 0, y_off = 0; y < 64; y++, y_off += 64)
for (int x = 0, x_off = 0; x < 64; x++, x_off += 64)
dst[x_off + y] = src[y_off + x];
}
static void hflip(uint8_t *const dst, const uint8_t *const src) {
for (int y = 0, y_off = 0; y < 64; y++, y_off += 64)
for (int x = 0; x < 64; x++)
dst[y_off + 64 - 1 - x] = src[y_off + x];
}
static void invert(uint8_t *const dst, const uint8_t *const src,
const int w, const int h)
{
for (int y = 0, y_off = 0; y < h; y++, y_off += w)
for (int x = 0; x < w; x++)
dst[y_off + x] = 64 - src[y_off + x];
}
static void copy2d(uint8_t *dst, const uint8_t *src,
const int w, const int h, const int x_off, const int y_off)
{
src += y_off * 64 + x_off;
for (int y = 0; y < h; y++) {
memcpy(dst, src, w);
src += 64;
dst += w;
}
}
static COLD void init_chroma(uint8_t *chroma, const uint8_t *luma,
const int sign, const int w, const int h,
const int ss_ver)
{
for (int y = 0; y < h; y += 1 + ss_ver) {
for (int x = 0; x < w; x += 2) {
int sum = luma[x] + luma[x + 1] + 1;
if (ss_ver) sum += luma[w + x] + luma[w + x + 1] + 1;
chroma[x >> 1] = (sum - sign) >> (1 + ss_ver);
}
luma += w << ss_ver;
chroma += w >> 1;
}
}
static COLD void fill2d_16x2(uint8_t *dst, const int w, const int h,
const enum BlockSize bs,
const uint8_t (*const master)[64 * 64],
const wedge_code_type *const cb,
uint8_t *masks_444, uint8_t *masks_422,
uint8_t *masks_420, const unsigned signs)
{
uint8_t *ptr = dst;
for (int n = 0; n < 16; n++) {
copy2d(ptr, master[cb[n].direction], w, h,
32 - (w * cb[n].x_offset >> 3), 32 - (h * cb[n].y_offset >> 3));
ptr += w * h;
}
for (int n = 0, off = 0; n < 16; n++, off += w * h)
invert(ptr + off, dst + off, w, h);
const int n_stride_444 = (w * h);
const int n_stride_422 = n_stride_444 >> 1;
const int n_stride_420 = n_stride_444 >> 2;
const int sign_stride_444 = 16 * n_stride_444;
const int sign_stride_422 = 16 * n_stride_422;
const int sign_stride_420 = 16 * n_stride_420;
// assign pointers in externally visible array
for (int n = 0; n < 16; n++) {
const int sign = (signs >> n) & 1;
dav1d_wedge_masks[bs][0][0][n] = &masks_444[ sign * sign_stride_444];
// not using !sign is intentional here, since 444 does not require
// any rounding since no chroma subsampling is applied.
dav1d_wedge_masks[bs][0][1][n] = &masks_444[ sign * sign_stride_444];
dav1d_wedge_masks[bs][1][0][n] = &masks_422[ sign * sign_stride_422];
dav1d_wedge_masks[bs][1][1][n] = &masks_422[!sign * sign_stride_422];
dav1d_wedge_masks[bs][2][0][n] = &masks_420[ sign * sign_stride_420];
dav1d_wedge_masks[bs][2][1][n] = &masks_420[!sign * sign_stride_420];
masks_444 += n_stride_444;
masks_422 += n_stride_422;
masks_420 += n_stride_420;
// since the pointers come from inside, we know that
// violation of the const is OK here. Any other approach
// means we would have to duplicate the sign correction
// logic in two places, which isn't very nice, or mark
// the table faced externally as non-const, which also sucks
init_chroma((uint8_t *)dav1d_wedge_masks[bs][1][0][n],
dav1d_wedge_masks[bs][0][0][n], 0, w, h, 0);
init_chroma((uint8_t *)dav1d_wedge_masks[bs][1][1][n],
dav1d_wedge_masks[bs][0][0][n], 1, w, h, 0);
init_chroma((uint8_t *)dav1d_wedge_masks[bs][2][0][n],
dav1d_wedge_masks[bs][0][0][n], 0, w, h, 1);
init_chroma((uint8_t *)dav1d_wedge_masks[bs][2][1][n],
dav1d_wedge_masks[bs][0][0][n], 1, w, h, 1);
}
}
COLD void dav1d_init_wedge_masks(void) {
// This function is guaranteed to be called only once
enum WedgeMasterLineType {
WEDGE_MASTER_LINE_ODD,
WEDGE_MASTER_LINE_EVEN,
WEDGE_MASTER_LINE_VERT,
N_WEDGE_MASTER_LINES,
};
static const uint8_t wedge_master_border[N_WEDGE_MASTER_LINES][8] = {
[WEDGE_MASTER_LINE_ODD] = { 1, 2, 6, 18, 37, 53, 60, 63 },
[WEDGE_MASTER_LINE_EVEN] = { 1, 4, 11, 27, 46, 58, 62, 63 },
[WEDGE_MASTER_LINE_VERT] = { 0, 2, 7, 21, 43, 57, 62, 64 },
};
uint8_t master[6][64 * 64];
// create master templates
for (int y = 0, off = 0; y < 64; y++, off += 64)
insert_border(&master[WEDGE_VERTICAL][off],
wedge_master_border[WEDGE_MASTER_LINE_VERT], 32);
for (int y = 0, off = 0, ctr = 48; y < 64; y += 2, off += 128, ctr--)
{
insert_border(&master[WEDGE_OBLIQUE63][off],
wedge_master_border[WEDGE_MASTER_LINE_EVEN], ctr);
insert_border(&master[WEDGE_OBLIQUE63][off + 64],
wedge_master_border[WEDGE_MASTER_LINE_ODD], ctr - 1);
}
transpose(master[WEDGE_OBLIQUE27], master[WEDGE_OBLIQUE63]);
transpose(master[WEDGE_HORIZONTAL], master[WEDGE_VERTICAL]);
hflip(master[WEDGE_OBLIQUE117], master[WEDGE_OBLIQUE63]);
hflip(master[WEDGE_OBLIQUE153], master[WEDGE_OBLIQUE27]);
#define fill(w, h, sz_422, sz_420, hvsw, signs) \
fill2d_16x2((uint8_t *) wedge_masks_444_##w##x##h, w, h, BS_##w##x##h, \
master, wedge_codebook_16_##hvsw, wedge_masks_444_##w##x##h, \
wedge_masks_422_##sz_422, wedge_masks_420_##sz_420, signs)
fill(32, 32, 16x32, 16x16, heqw, 0x7bfb);
fill(32, 16, 16x16, 16x8, hltw, 0x7beb);
fill(32, 8, 16x8, 16x4, hltw, 0x6beb);
fill(16, 32, 8x32, 8x16, hgtw, 0x7beb);
fill(16, 16, 8x16, 8x8, heqw, 0x7bfb);
fill(16, 8, 8x8, 8x4, hltw, 0x7beb);
fill( 8, 32, 4x32, 4x16, hgtw, 0x7aeb);
fill( 8, 16, 4x16, 4x8, hgtw, 0x7beb);
fill( 8, 8, 4x8, 4x4, heqw, 0x7bfb);
#undef fill
}
#define N_II_PRED_MODES (N_INTER_INTRA_PRED_MODES - 1)
static uint8_t ALIGN(ii_dc_mask[32 * 32], 64);
static uint8_t ALIGN(ii_nondc_mask_32x32[N_II_PRED_MODES][32 * 32], 64);
static uint8_t ALIGN(ii_nondc_mask_16x32[N_II_PRED_MODES][16 * 32], 64);
static uint8_t ALIGN(ii_nondc_mask_16x16[N_II_PRED_MODES][16 * 16], 64);
static uint8_t ALIGN(ii_nondc_mask_8x32 [N_II_PRED_MODES][ 8 * 32], 64);
static uint8_t ALIGN(ii_nondc_mask_8x16 [N_II_PRED_MODES][ 8 * 16], 64);
static uint8_t ALIGN(ii_nondc_mask_8x8 [N_II_PRED_MODES][ 8 * 8], 64);
static uint8_t ALIGN(ii_nondc_mask_4x16 [N_II_PRED_MODES][ 4 * 16], 64);
static uint8_t ALIGN(ii_nondc_mask_4x8 [N_II_PRED_MODES][ 4 * 8], 32);
static uint8_t ALIGN(ii_nondc_mask_4x4 [N_II_PRED_MODES][ 4 * 4], 16);
#undef N_II_PRED_MODES
#define set1(sz) \
[II_DC_PRED] = ii_dc_mask, \
[II_VERT_PRED] = ii_nondc_mask_##sz[II_VERT_PRED - 1], \
[II_HOR_PRED] = ii_nondc_mask_##sz[II_HOR_PRED - 1], \
[II_SMOOTH_PRED] = ii_nondc_mask_##sz[II_SMOOTH_PRED - 1]
#define set(sz_444, sz_422, sz_420) \
{ { set1(sz_444) }, { set1(sz_422) }, { set1(sz_420) } }
const uint8_t *dav1d_ii_masks[N_BS_SIZES][3][N_INTER_INTRA_PRED_MODES] = {
[BS_8x8] = set( 8x8, 4x8, 4x4),
[BS_8x16] = set( 8x16, 4x16, 4x8),
[BS_16x8] = set(16x16, 8x8, 8x8),
[BS_16x16] = set(16x16, 8x16, 8x8),
[BS_16x32] = set(16x32, 8x32, 8x16),
[BS_32x16] = set(32x32, 16x16, 16x16),
[BS_32x32] = set(32x32, 16x32, 16x16),
};
#undef set
#undef set1
static COLD void build_nondc_ii_masks(uint8_t *const mask_v,
uint8_t *const mask_h,
uint8_t *const mask_sm,
const int w, const int h, const int step)
{
static const uint8_t ii_weights_1d[] = {
60, 52, 45, 39, 34, 30, 26, 22, 19, 17, 15, 13, 11, 10, 8, 7,
6, 6, 5, 4, 4, 3, 3, 2, 2, 2, 2, 1, 1, 1, 1, 1,
};
for (int y = 0, off = 0; y < h; y++, off += w) {
memset(&mask_v[off], ii_weights_1d[y * step], w);
for (int x = 0; x < w; x++) {
mask_sm[off + x] = ii_weights_1d[imin(x, y) * step];
mask_h[off + x] = ii_weights_1d[x * step];
}
}
}
COLD void dav1d_init_interintra_masks(void) {
// This function is guaranteed to be called only once
memset(ii_dc_mask, 32, 32 * 32);
#define set(a) a[II_VERT_PRED - 1], a[II_HOR_PRED - 1], a[II_SMOOTH_PRED - 1]
build_nondc_ii_masks(set(ii_nondc_mask_32x32), 32, 32, 1);
build_nondc_ii_masks(set(ii_nondc_mask_16x32), 16, 32, 1);
build_nondc_ii_masks(set(ii_nondc_mask_16x16), 16, 16, 2);
build_nondc_ii_masks(set(ii_nondc_mask_8x32), 8, 32, 1);
build_nondc_ii_masks(set(ii_nondc_mask_8x16), 8, 16, 2);
build_nondc_ii_masks(set(ii_nondc_mask_8x8), 8, 8, 4);
build_nondc_ii_masks(set(ii_nondc_mask_4x16), 4, 16, 2);
build_nondc_ii_masks(set(ii_nondc_mask_4x8), 4, 8, 4);
build_nondc_ii_masks(set(ii_nondc_mask_4x4), 4, 4, 8);
#undef set
}