/* * Copyright (c) 2016, Alliance for Open Media. All rights reserved. * * This source code is subject to the terms of the BSD 2 Clause License and * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License * was not distributed with this source code in the LICENSE file, you can * obtain it at www.aomedia.org/license/software. If the Alliance for Open * Media Patent License 1.0 was not distributed with this source code in the * PATENTS file, you can obtain it at www.aomedia.org/license/patent. */ #include <assert.h> #include <math.h> #include "config/aom_config.h" #include "config/aom_dsp_rtcd.h" #include "config/av1_rtcd.h" #include "aom_dsp/aom_dsp_common.h" #include "aom_mem/aom_mem.h" #include "aom_ports/aom_once.h" #include "aom_ports/mem.h" #include "av1/common/av1_common_int.h" #include "av1/common/cfl.h" #include "av1/common/reconintra.h" enum { … }; #define INTRA_EDGE_FILT … #define INTRA_EDGE_TAPS … #define MAX_UPSAMPLE_SZ … #define NUM_INTRA_NEIGHBOUR_PIXELS … static const uint8_t extend_modes[INTRA_MODES] = …; // Tables to store if the top-right reference pixels are available. The flags // are represented with bits, packed into 8-bit integers. E.g., for the 32x32 // blocks in a 128x128 superblock, the index of the "o" block is 10 (in raster // order), so its flag is stored at the 3rd bit of the 2nd entry in the table, // i.e. (table[10 / 8] >> (10 % 8)) & 1. // . . . . // . . . . // . . o . // . . . . static uint8_t has_tr_4x4[128] = …; static uint8_t has_tr_4x8[64] = …; static uint8_t has_tr_8x4[64] = …; static uint8_t has_tr_8x8[32] = …; static uint8_t has_tr_8x16[16] = …; static uint8_t has_tr_16x8[16] = …; static uint8_t has_tr_16x16[8] = …; static uint8_t has_tr_16x32[4] = …; static uint8_t has_tr_32x16[4] = …; static uint8_t has_tr_32x32[2] = …; static uint8_t has_tr_32x64[1] = …; static uint8_t has_tr_64x32[1] = …; static uint8_t has_tr_64x64[1] = …; static uint8_t has_tr_64x128[1] = …; static uint8_t has_tr_128x64[1] = …; static uint8_t has_tr_128x128[1] = …; static uint8_t has_tr_4x16[32] = …; static uint8_t has_tr_16x4[32] = …; static uint8_t has_tr_8x32[8] = …; static uint8_t has_tr_32x8[8] = …; static uint8_t has_tr_16x64[2] = …; static uint8_t has_tr_64x16[2] = …; static const uint8_t *const has_tr_tables[BLOCK_SIZES_ALL] = …; static uint8_t has_tr_vert_8x8[32] = …; static uint8_t has_tr_vert_16x16[8] = …; static uint8_t has_tr_vert_32x32[2] = …; static uint8_t has_tr_vert_64x64[1] = …; // The _vert_* tables are like the ordinary tables above, but describe the // order we visit square blocks when doing a PARTITION_VERT_A or // PARTITION_VERT_B. This is the same order as normal except for on the last // split where we go vertically (TL, BL, TR, BR). We treat the rectangular block // as a pair of squares, which means that these tables work correctly for both // mixed vertical partition types. // // There are tables for each of the square sizes. Vertical rectangles (like // BLOCK_16X32) use their respective "non-vert" table static const uint8_t *const has_tr_vert_tables[BLOCK_SIZES] = …; static const uint8_t *get_has_tr_table(PARTITION_TYPE partition, BLOCK_SIZE bsize) { … } static int has_top_right(BLOCK_SIZE sb_size, BLOCK_SIZE bsize, int mi_row, int mi_col, int top_available, int right_available, PARTITION_TYPE partition, TX_SIZE txsz, int row_off, int col_off, int ss_x, int ss_y) { … } // Similar to the has_tr_* tables, but store if the bottom-left reference // pixels are available. static uint8_t has_bl_4x4[128] = …; static uint8_t has_bl_4x8[64] = …; static uint8_t has_bl_8x4[64] = …; static uint8_t has_bl_8x8[32] = …; static uint8_t has_bl_8x16[16] = …; static uint8_t has_bl_16x8[16] = …; static uint8_t has_bl_16x16[8] = …; static uint8_t has_bl_16x32[4] = …; static uint8_t has_bl_32x16[4] = …; static uint8_t has_bl_32x32[2] = …; static uint8_t has_bl_32x64[1] = …; static uint8_t has_bl_64x32[1] = …; static uint8_t has_bl_64x64[1] = …; static uint8_t has_bl_64x128[1] = …; static uint8_t has_bl_128x64[1] = …; static uint8_t has_bl_128x128[1] = …; static uint8_t has_bl_4x16[32] = …; static uint8_t has_bl_16x4[32] = …; static uint8_t has_bl_8x32[8] = …; static uint8_t has_bl_32x8[8] = …; static uint8_t has_bl_16x64[2] = …; static uint8_t has_bl_64x16[2] = …; static const uint8_t *const has_bl_tables[BLOCK_SIZES_ALL] = …; static uint8_t has_bl_vert_8x8[32] = …; static uint8_t has_bl_vert_16x16[8] = …; static uint8_t has_bl_vert_32x32[2] = …; static uint8_t has_bl_vert_64x64[1] = …; // The _vert_* tables are like the ordinary tables above, but describe the // order we visit square blocks when doing a PARTITION_VERT_A or // PARTITION_VERT_B. This is the same order as normal except for on the last // split where we go vertically (TL, BL, TR, BR). We treat the rectangular block // as a pair of squares, which means that these tables work correctly for both // mixed vertical partition types. // // There are tables for each of the square sizes. Vertical rectangles (like // BLOCK_16X32) use their respective "non-vert" table static const uint8_t *const has_bl_vert_tables[BLOCK_SIZES] = …; static const uint8_t *get_has_bl_table(PARTITION_TYPE partition, BLOCK_SIZE bsize) { … } static int has_bottom_left(BLOCK_SIZE sb_size, BLOCK_SIZE bsize, int mi_row, int mi_col, int bottom_available, int left_available, PARTITION_TYPE partition, TX_SIZE txsz, int row_off, int col_off, int ss_x, int ss_y) { … } intra_pred_fn; static intra_pred_fn pred[INTRA_MODES][TX_SIZES_ALL]; static intra_pred_fn dc_pred[2][2][TX_SIZES_ALL]; #if CONFIG_AV1_HIGHBITDEPTH typedef void (*intra_high_pred_fn)(uint16_t *dst, ptrdiff_t stride, const uint16_t *above, const uint16_t *left, int bd); static intra_high_pred_fn pred_high[INTRA_MODES][TX_SIZES_ALL]; static intra_high_pred_fn dc_pred_high[2][2][TX_SIZES_ALL]; #endif static void init_intra_predictors_internal(void) { … } // Directional prediction, zone 1: 0 < angle < 90 void av1_dr_prediction_z1_c(uint8_t *dst, ptrdiff_t stride, int bw, int bh, const uint8_t *above, const uint8_t *left, int upsample_above, int dx, int dy) { … } // Directional prediction, zone 2: 90 < angle < 180 void av1_dr_prediction_z2_c(uint8_t *dst, ptrdiff_t stride, int bw, int bh, const uint8_t *above, const uint8_t *left, int upsample_above, int upsample_left, int dx, int dy) { … } // Directional prediction, zone 3: 180 < angle < 270 void av1_dr_prediction_z3_c(uint8_t *dst, ptrdiff_t stride, int bw, int bh, const uint8_t *above, const uint8_t *left, int upsample_left, int dx, int dy) { … } static void dr_predictor(uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left, int upsample_above, int upsample_left, int angle) { … } #if CONFIG_AV1_HIGHBITDEPTH // Directional prediction, zone 1: 0 < angle < 90 void av1_highbd_dr_prediction_z1_c(uint16_t *dst, ptrdiff_t stride, int bw, int bh, const uint16_t *above, const uint16_t *left, int upsample_above, int dx, int dy, int bd) { int r, c, x, base, shift, val; (void)left; (void)dy; (void)bd; assert(dy == 1); assert(dx > 0); const int max_base_x = ((bw + bh) - 1) << upsample_above; const int frac_bits = 6 - upsample_above; const int base_inc = 1 << upsample_above; x = dx; for (r = 0; r < bh; ++r, dst += stride, x += dx) { base = x >> frac_bits; shift = ((x << upsample_above) & 0x3F) >> 1; if (base >= max_base_x) { for (int i = r; i < bh; ++i) { aom_memset16(dst, above[max_base_x], bw); dst += stride; } return; } for (c = 0; c < bw; ++c, base += base_inc) { if (base < max_base_x) { val = above[base] * (32 - shift) + above[base + 1] * shift; dst[c] = ROUND_POWER_OF_TWO(val, 5); } else { dst[c] = above[max_base_x]; } } } } // Directional prediction, zone 2: 90 < angle < 180 void av1_highbd_dr_prediction_z2_c(uint16_t *dst, ptrdiff_t stride, int bw, int bh, const uint16_t *above, const uint16_t *left, int upsample_above, int upsample_left, int dx, int dy, int bd) { (void)bd; assert(dx > 0); assert(dy > 0); const int min_base_x = -(1 << upsample_above); const int min_base_y = -(1 << upsample_left); (void)min_base_y; const int frac_bits_x = 6 - upsample_above; const int frac_bits_y = 6 - upsample_left; for (int r = 0; r < bh; ++r) { for (int c = 0; c < bw; ++c) { int val; int y = r + 1; int x = (c << 6) - y * dx; const int base_x = x >> frac_bits_x; if (base_x >= min_base_x) { const int shift = ((x * (1 << upsample_above)) & 0x3F) >> 1; val = above[base_x] * (32 - shift) + above[base_x + 1] * shift; val = ROUND_POWER_OF_TWO(val, 5); } else { x = c + 1; y = (r << 6) - x * dy; const int base_y = y >> frac_bits_y; assert(base_y >= min_base_y); const int shift = ((y * (1 << upsample_left)) & 0x3F) >> 1; val = left[base_y] * (32 - shift) + left[base_y + 1] * shift; val = ROUND_POWER_OF_TWO(val, 5); } dst[c] = val; } dst += stride; } } // Directional prediction, zone 3: 180 < angle < 270 void av1_highbd_dr_prediction_z3_c(uint16_t *dst, ptrdiff_t stride, int bw, int bh, const uint16_t *above, const uint16_t *left, int upsample_left, int dx, int dy, int bd) { int r, c, y, base, shift, val; (void)above; (void)dx; (void)bd; assert(dx == 1); assert(dy > 0); const int max_base_y = (bw + bh - 1) << upsample_left; const int frac_bits = 6 - upsample_left; const int base_inc = 1 << upsample_left; y = dy; for (c = 0; c < bw; ++c, y += dy) { base = y >> frac_bits; shift = ((y << upsample_left) & 0x3F) >> 1; for (r = 0; r < bh; ++r, base += base_inc) { if (base < max_base_y) { val = left[base] * (32 - shift) + left[base + 1] * shift; dst[r * stride + c] = ROUND_POWER_OF_TWO(val, 5); } else { for (; r < bh; ++r) dst[r * stride + c] = left[max_base_y]; break; } } } } static void highbd_dr_predictor(uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int upsample_above, int upsample_left, int angle, int bd) { const int dx = av1_get_dx(angle); const int dy = av1_get_dy(angle); const int bw = tx_size_wide[tx_size]; const int bh = tx_size_high[tx_size]; assert(angle > 0 && angle < 270); if (angle > 0 && angle < 90) { av1_highbd_dr_prediction_z1(dst, stride, bw, bh, above, left, upsample_above, dx, dy, bd); } else if (angle > 90 && angle < 180) { av1_highbd_dr_prediction_z2(dst, stride, bw, bh, above, left, upsample_above, upsample_left, dx, dy, bd); } else if (angle > 180 && angle < 270) { av1_highbd_dr_prediction_z3(dst, stride, bw, bh, above, left, upsample_left, dx, dy, bd); } else if (angle == 90) { pred_high[V_PRED][tx_size](dst, stride, above, left, bd); } else if (angle == 180) { pred_high[H_PRED][tx_size](dst, stride, above, left, bd); } } #endif // CONFIG_AV1_HIGHBITDEPTH DECLARE_ALIGNED(16, const int8_t, av1_filter_intra_taps[FILTER_INTRA_MODES][8][8]) = …; void av1_filter_intra_predictor_c(uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left, int mode) { … } #if CONFIG_AV1_HIGHBITDEPTH static void highbd_filter_intra_predictor(uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int mode, int bd) { int r, c; uint16_t buffer[33][33]; const int bw = tx_size_wide[tx_size]; const int bh = tx_size_high[tx_size]; assert(bw <= 32 && bh <= 32); for (r = 0; r < bh; ++r) buffer[r + 1][0] = left[r]; memcpy(buffer[0], &above[-1], (bw + 1) * sizeof(buffer[0][0])); for (r = 1; r < bh + 1; r += 2) for (c = 1; c < bw + 1; c += 4) { const uint16_t p0 = buffer[r - 1][c - 1]; const uint16_t p1 = buffer[r - 1][c]; const uint16_t p2 = buffer[r - 1][c + 1]; const uint16_t p3 = buffer[r - 1][c + 2]; const uint16_t p4 = buffer[r - 1][c + 3]; const uint16_t p5 = buffer[r][c - 1]; const uint16_t p6 = buffer[r + 1][c - 1]; for (int k = 0; k < 8; ++k) { int r_offset = k >> 2; int c_offset = k & 0x03; int pr = av1_filter_intra_taps[mode][k][0] * p0 + av1_filter_intra_taps[mode][k][1] * p1 + av1_filter_intra_taps[mode][k][2] * p2 + av1_filter_intra_taps[mode][k][3] * p3 + av1_filter_intra_taps[mode][k][4] * p4 + av1_filter_intra_taps[mode][k][5] * p5 + av1_filter_intra_taps[mode][k][6] * p6; // Section 7.11.2.3 specifies the right-hand side of the assignment as // Clip1( Round2Signed( pr, INTRA_FILTER_SCALE_BITS ) ). // Since Clip1() clips a negative value to 0, it is safe to replace // Round2Signed() with Round2(). buffer[r + r_offset][c + c_offset] = clip_pixel_highbd( ROUND_POWER_OF_TWO(pr, FILTER_INTRA_SCALE_BITS), bd); } } for (r = 0; r < bh; ++r) { memcpy(dst, &buffer[r + 1][1], bw * sizeof(dst[0])); dst += stride; } } #endif // CONFIG_AV1_HIGHBITDEPTH static int is_smooth(const MB_MODE_INFO *mbmi, int plane) { … } static int get_intra_edge_filter_type(const MACROBLOCKD *xd, int plane) { … } static int intra_edge_filter_strength(int bs0, int bs1, int delta, int type) { … } void av1_filter_intra_edge_c(uint8_t *p, int sz, int strength) { … } static void filter_intra_edge_corner(uint8_t *p_above, uint8_t *p_left) { … } void av1_upsample_intra_edge_c(uint8_t *p, int sz) { … } static void build_directional_and_filter_intra_predictors( const uint8_t *ref, int ref_stride, uint8_t *dst, int dst_stride, PREDICTION_MODE mode, int p_angle, FILTER_INTRA_MODE filter_intra_mode, TX_SIZE tx_size, int disable_edge_filter, int n_top_px, int n_topright_px, int n_left_px, int n_bottomleft_px, int intra_edge_filter_type) { … } // This function generates the pred data of a given block for non-directional // intra prediction modes (i.e., DC, SMOOTH, SMOOTH_H, SMOOTH_V and PAETH). static void build_non_directional_intra_predictors( const uint8_t *ref, int ref_stride, uint8_t *dst, int dst_stride, PREDICTION_MODE mode, TX_SIZE tx_size, int n_top_px, int n_left_px) { … } #if CONFIG_AV1_HIGHBITDEPTH void av1_highbd_filter_intra_edge_c(uint16_t *p, int sz, int strength) { if (!strength) return; const int kernel[INTRA_EDGE_FILT][INTRA_EDGE_TAPS] = { { 0, 4, 8, 4, 0 }, { 0, 5, 6, 5, 0 }, { 2, 4, 4, 4, 2 } }; const int filt = strength - 1; uint16_t edge[129]; memcpy(edge, p, sz * sizeof(*p)); for (int i = 1; i < sz; i++) { int s = 0; for (int j = 0; j < INTRA_EDGE_TAPS; j++) { int k = i - 2 + j; k = (k < 0) ? 0 : k; k = (k > sz - 1) ? sz - 1 : k; s += edge[k] * kernel[filt][j]; } s = (s + 8) >> 4; p[i] = s; } } static void highbd_filter_intra_edge_corner(uint16_t *p_above, uint16_t *p_left) { const int kernel[3] = { 5, 6, 5 }; int s = (p_left[0] * kernel[0]) + (p_above[-1] * kernel[1]) + (p_above[0] * kernel[2]); s = (s + 8) >> 4; p_above[-1] = s; p_left[-1] = s; } void av1_highbd_upsample_intra_edge_c(uint16_t *p, int sz, int bd) { // interpolate half-sample positions assert(sz <= MAX_UPSAMPLE_SZ); uint16_t in[MAX_UPSAMPLE_SZ + 3]; // copy p[-1..(sz-1)] and extend first and last samples in[0] = p[-1]; in[1] = p[-1]; for (int i = 0; i < sz; i++) { in[i + 2] = p[i]; } in[sz + 2] = p[sz - 1]; // interpolate half-sample edge positions p[-2] = in[0]; for (int i = 0; i < sz; i++) { int s = -in[i] + (9 * in[i + 1]) + (9 * in[i + 2]) - in[i + 3]; s = (s + 8) >> 4; s = clip_pixel_highbd(s, bd); p[2 * i - 1] = s; p[2 * i] = in[i + 2]; } } static void highbd_build_directional_and_filter_intra_predictors( const uint8_t *ref8, int ref_stride, uint8_t *dst8, int dst_stride, PREDICTION_MODE mode, int p_angle, FILTER_INTRA_MODE filter_intra_mode, TX_SIZE tx_size, int disable_edge_filter, int n_top_px, int n_topright_px, int n_left_px, int n_bottomleft_px, int intra_edge_filter_type, int bit_depth) { int i; uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); const uint16_t *const ref = CONVERT_TO_SHORTPTR(ref8); DECLARE_ALIGNED(16, uint16_t, left_data[NUM_INTRA_NEIGHBOUR_PIXELS]); DECLARE_ALIGNED(16, uint16_t, above_data[NUM_INTRA_NEIGHBOUR_PIXELS]); uint16_t *const above_row = above_data + 16; uint16_t *const left_col = left_data + 16; const int txwpx = tx_size_wide[tx_size]; const int txhpx = tx_size_high[tx_size]; int need_left = extend_modes[mode] & NEED_LEFT; int need_above = extend_modes[mode] & NEED_ABOVE; int need_above_left = extend_modes[mode] & NEED_ABOVELEFT; const uint16_t *above_ref = ref - ref_stride; const uint16_t *left_ref = ref - 1; const int is_dr_mode = av1_is_directional_mode(mode); const int use_filter_intra = filter_intra_mode != FILTER_INTRA_MODES; assert(use_filter_intra || is_dr_mode); const int base = 128 << (bit_depth - 8); // The left_data, above_data buffers must be zeroed to fix some intermittent // valgrind errors. Uninitialized reads in intra pred modules (e.g. width = 4 // path in av1_highbd_dr_prediction_z2_avx2()) from left_data, above_data are // seen to be the potential reason for this issue. aom_memset16(left_data, base + 1, NUM_INTRA_NEIGHBOUR_PIXELS); aom_memset16(above_data, base - 1, NUM_INTRA_NEIGHBOUR_PIXELS); // The default values if ref pixels are not available: // base base-1 base-1 .. base-1 base-1 base-1 base-1 base-1 base-1 // base+1 A B .. Y Z // base+1 C D .. W X // base+1 E F .. U V // base+1 G H .. S T T T T T if (is_dr_mode) { if (p_angle <= 90) need_above = 1, need_left = 0, need_above_left = 1; else if (p_angle < 180) need_above = 1, need_left = 1, need_above_left = 1; else need_above = 0, need_left = 1, need_above_left = 1; } if (use_filter_intra) need_left = need_above = need_above_left = 1; assert(n_top_px >= 0); assert(n_topright_px >= -1); assert(n_left_px >= 0); assert(n_bottomleft_px >= -1); if ((!need_above && n_left_px == 0) || (!need_left && n_top_px == 0)) { int val; if (need_left) { val = (n_top_px > 0) ? above_ref[0] : base + 1; } else { val = (n_left_px > 0) ? left_ref[0] : base - 1; } for (i = 0; i < txhpx; ++i) { aom_memset16(dst, val, txwpx); dst += dst_stride; } return; } // NEED_LEFT if (need_left) { const int num_left_pixels_needed = txhpx + (n_bottomleft_px >= 0 ? txwpx : 0); i = 0; if (n_left_px > 0) { for (; i < n_left_px; i++) left_col[i] = left_ref[i * ref_stride]; if (n_bottomleft_px > 0) { assert(i == txhpx); for (; i < txhpx + n_bottomleft_px; i++) left_col[i] = left_ref[i * ref_stride]; } if (i < num_left_pixels_needed) aom_memset16(&left_col[i], left_col[i - 1], num_left_pixels_needed - i); } else if (n_top_px > 0) { aom_memset16(left_col, above_ref[0], num_left_pixels_needed); } } // NEED_ABOVE if (need_above) { const int num_top_pixels_needed = txwpx + (n_topright_px >= 0 ? txhpx : 0); if (n_top_px > 0) { memcpy(above_row, above_ref, n_top_px * sizeof(above_ref[0])); i = n_top_px; if (n_topright_px > 0) { assert(n_top_px == txwpx); memcpy(above_row + txwpx, above_ref + txwpx, n_topright_px * sizeof(above_ref[0])); i += n_topright_px; } if (i < num_top_pixels_needed) aom_memset16(&above_row[i], above_row[i - 1], num_top_pixels_needed - i); } else if (n_left_px > 0) { aom_memset16(above_row, left_ref[0], num_top_pixels_needed); } } if (need_above_left) { if (n_top_px > 0 && n_left_px > 0) { above_row[-1] = above_ref[-1]; } else if (n_top_px > 0) { above_row[-1] = above_ref[0]; } else if (n_left_px > 0) { above_row[-1] = left_ref[0]; } else { above_row[-1] = base; } left_col[-1] = above_row[-1]; } if (use_filter_intra) { highbd_filter_intra_predictor(dst, dst_stride, tx_size, above_row, left_col, filter_intra_mode, bit_depth); return; } assert(is_dr_mode); int upsample_above = 0; int upsample_left = 0; if (!disable_edge_filter) { const int need_right = p_angle < 90; const int need_bottom = p_angle > 180; if (p_angle != 90 && p_angle != 180) { assert(need_above_left); const int ab_le = 1; if (need_above && need_left && (txwpx + txhpx >= 24)) { highbd_filter_intra_edge_corner(above_row, left_col); } if (need_above && n_top_px > 0) { const int strength = intra_edge_filter_strength( txwpx, txhpx, p_angle - 90, intra_edge_filter_type); const int n_px = n_top_px + ab_le + (need_right ? txhpx : 0); av1_highbd_filter_intra_edge(above_row - ab_le, n_px, strength); } if (need_left && n_left_px > 0) { const int strength = intra_edge_filter_strength( txhpx, txwpx, p_angle - 180, intra_edge_filter_type); const int n_px = n_left_px + ab_le + (need_bottom ? txwpx : 0); av1_highbd_filter_intra_edge(left_col - ab_le, n_px, strength); } } upsample_above = av1_use_intra_edge_upsample(txwpx, txhpx, p_angle - 90, intra_edge_filter_type); if (need_above && upsample_above) { const int n_px = txwpx + (need_right ? txhpx : 0); av1_highbd_upsample_intra_edge(above_row, n_px, bit_depth); } upsample_left = av1_use_intra_edge_upsample(txhpx, txwpx, p_angle - 180, intra_edge_filter_type); if (need_left && upsample_left) { const int n_px = txhpx + (need_bottom ? txwpx : 0); av1_highbd_upsample_intra_edge(left_col, n_px, bit_depth); } } highbd_dr_predictor(dst, dst_stride, tx_size, above_row, left_col, upsample_above, upsample_left, p_angle, bit_depth); } // For HBD encode/decode, this function generates the pred data of a given // block for non-directional intra prediction modes (i.e., DC, SMOOTH, SMOOTH_H, // SMOOTH_V and PAETH). static void highbd_build_non_directional_intra_predictors( const uint8_t *ref8, int ref_stride, uint8_t *dst8, int dst_stride, PREDICTION_MODE mode, TX_SIZE tx_size, int n_top_px, int n_left_px, int bit_depth) { int i = 0; uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); const uint16_t *const ref = CONVERT_TO_SHORTPTR(ref8); const int txwpx = tx_size_wide[tx_size]; const int txhpx = tx_size_high[tx_size]; int need_left = extend_modes[mode] & NEED_LEFT; int need_above = extend_modes[mode] & NEED_ABOVE; int need_above_left = extend_modes[mode] & NEED_ABOVELEFT; const uint16_t *above_ref = ref - ref_stride; const uint16_t *left_ref = ref - 1; const int base = 128 << (bit_depth - 8); assert(n_top_px >= 0); assert(n_left_px >= 0); assert(mode == DC_PRED || mode == SMOOTH_PRED || mode == SMOOTH_V_PRED || mode == SMOOTH_H_PRED || mode == PAETH_PRED); if ((!need_above && n_left_px == 0) || (!need_left && n_top_px == 0)) { int val = 0; if (need_left) { val = (n_top_px > 0) ? above_ref[0] : base + 1; } else { val = (n_left_px > 0) ? left_ref[0] : base - 1; } for (i = 0; i < txhpx; ++i) { aom_memset16(dst, val, txwpx); dst += dst_stride; } return; } DECLARE_ALIGNED(16, uint16_t, left_data[NUM_INTRA_NEIGHBOUR_PIXELS]); DECLARE_ALIGNED(16, uint16_t, above_data[NUM_INTRA_NEIGHBOUR_PIXELS]); uint16_t *const above_row = above_data + 16; uint16_t *const left_col = left_data + 16; if (need_left) { aom_memset16(left_data, base + 1, NUM_INTRA_NEIGHBOUR_PIXELS); if (n_left_px > 0) { for (i = 0; i < n_left_px; i++) left_col[i] = left_ref[i * ref_stride]; if (i < txhpx) aom_memset16(&left_col[i], left_col[i - 1], txhpx - i); } else if (n_top_px > 0) { aom_memset16(left_col, above_ref[0], txhpx); } } if (need_above) { aom_memset16(above_data, base - 1, NUM_INTRA_NEIGHBOUR_PIXELS); if (n_top_px > 0) { memcpy(above_row, above_ref, n_top_px * sizeof(above_ref[0])); i = n_top_px; if (i < txwpx) aom_memset16(&above_row[i], above_row[i - 1], (txwpx - i)); } else if (n_left_px > 0) { aom_memset16(above_row, left_ref[0], txwpx); } } if (need_above_left) { if (n_top_px > 0 && n_left_px > 0) { above_row[-1] = above_ref[-1]; } else if (n_top_px > 0) { above_row[-1] = above_ref[0]; } else if (n_left_px > 0) { above_row[-1] = left_ref[0]; } else { above_row[-1] = base; } left_col[-1] = above_row[-1]; } if (mode == DC_PRED) { dc_pred_high[n_left_px > 0][n_top_px > 0][tx_size]( dst, dst_stride, above_row, left_col, bit_depth); } else { pred_high[mode][tx_size](dst, dst_stride, above_row, left_col, bit_depth); } } #endif // CONFIG_AV1_HIGHBITDEPTH static inline BLOCK_SIZE scale_chroma_bsize(BLOCK_SIZE bsize, int subsampling_x, int subsampling_y) { … } void av1_predict_intra_block(const MACROBLOCKD *xd, BLOCK_SIZE sb_size, int enable_intra_edge_filter, int wpx, int hpx, TX_SIZE tx_size, PREDICTION_MODE mode, int angle_delta, int use_palette, FILTER_INTRA_MODE filter_intra_mode, const uint8_t *ref, int ref_stride, uint8_t *dst, int dst_stride, int col_off, int row_off, int plane) { … } void av1_predict_intra_block_facade(const AV1_COMMON *cm, MACROBLOCKD *xd, int plane, int blk_col, int blk_row, TX_SIZE tx_size) { … } void av1_init_intra_predictors(void) { … }
Codebase Browser
chromium