2018-11-29 16:00:05 +00:00
|
|
|
/*****************************************************************************
|
|
|
|
* This file is part of Kvazaar HEVC encoder.
|
|
|
|
*
|
|
|
|
* Copyright (C) 2013-2015 Tampere University of Technology and others (see
|
|
|
|
* COPYING file).
|
|
|
|
*
|
|
|
|
* Kvazaar is free software: you can redistribute it and/or modify it under
|
|
|
|
* the terms of the GNU Lesser General Public License as published by the
|
|
|
|
* Free Software Foundation; either version 2.1 of the License, or (at your
|
|
|
|
* option) any later version.
|
|
|
|
*
|
|
|
|
* Kvazaar is distributed in the hope that it will be useful, but WITHOUT ANY
|
|
|
|
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
|
|
|
|
* FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for
|
|
|
|
* more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public License along
|
|
|
|
* with Kvazaar. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
****************************************************************************/
|
|
|
|
|
|
|
|
#include "strategyselector.h"
|
|
|
|
|
2019-02-04 17:34:30 +00:00
|
|
|
#if COMPILE_INTEL_AVX2
|
2018-12-18 16:13:50 +00:00
|
|
|
#include "avx2_common_functions.h"
|
2018-11-29 16:00:05 +00:00
|
|
|
#include "cabac.h"
|
|
|
|
#include "context.h"
|
|
|
|
#include "encode_coding_tree-avx2.h"
|
2019-02-04 12:55:41 +00:00
|
|
|
#include "encode_coding_tree.h"
|
2019-05-23 16:59:40 +00:00
|
|
|
#include "strategies/missing-intel-intrinsics.h"
|
2018-12-03 15:21:22 +00:00
|
|
|
#include <immintrin.h>
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-14 19:09:54 +00:00
|
|
|
/*
|
2019-02-01 14:26:21 +00:00
|
|
|
* NOTE: Unlike SSE/AVX comparisons that would return 11 or 00 for gt/lte,
|
|
|
|
* this'll use 1x and 0x as bit patterns (x: garbage). A couple extra
|
|
|
|
* instructions will get you 11 and 00 if you need to use this as a mask
|
|
|
|
* somewhere at some point, but we don't need this right now.
|
|
|
|
*
|
|
|
|
* I'd love to draw a logic circuit here to describe this, but I can't. Two
|
|
|
|
* 2-bit uints can be compared for greaterness by first comparing their high
|
|
|
|
* bits using AND-NOT; (x AND (NOT y)) == 1 if x > y. If A_hi > B_hi, A > B.
|
|
|
|
* If A_hi == B_hi AND A_lo > B_lo, A > B. Otherwise, A <= B. It's really
|
|
|
|
* simple when drawn on paper, but quite messy on a general-purpose ALU. But
|
|
|
|
* look, just five instructions!
|
2018-12-14 19:09:54 +00:00
|
|
|
*/
|
2019-02-01 14:26:21 +00:00
|
|
|
static INLINE uint32_t u32vec_cmpgt_epu2(uint32_t a, uint32_t b)
|
2018-12-14 19:09:54 +00:00
|
|
|
{
|
2018-12-18 18:42:09 +00:00
|
|
|
uint32_t a_gt_b = _andn_u32(b, a);
|
2018-12-14 19:09:54 +00:00
|
|
|
uint32_t a_ne_b = a ^ b;
|
|
|
|
uint32_t a_gt_b_sh = a_gt_b << 1;
|
2018-12-18 18:42:09 +00:00
|
|
|
uint32_t lobit_tiebrk_hi = _andn_u32(a_ne_b, a_gt_b_sh);
|
2019-02-01 14:26:21 +00:00
|
|
|
uint32_t res = a_gt_b | lobit_tiebrk_hi;
|
2018-12-14 19:09:54 +00:00
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
2018-12-20 08:51:44 +00:00
|
|
|
static INLINE uint32_t pack_16x16b_to_16x2b(__m256i src)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* For each 16-bit element in src:
|
2019-01-03 14:37:05 +00:00
|
|
|
* ABCD EFGH IJKL MNOP Original elements
|
|
|
|
* 0000 0000 0000 00XY Element clipped to [0, 3] using _mm256_min_epu16
|
2018-12-20 08:51:44 +00:00
|
|
|
* 0000 000X Y000 0000 Shift word to align LSBs across byte boundary
|
2019-01-03 14:37:05 +00:00
|
|
|
* 0000 0001 1000 0000 Comparison mask to be compared against
|
2018-12-20 08:51:44 +00:00
|
|
|
* XXXX XXXX YYYY YYYY Comparison result, for movemask
|
|
|
|
*/
|
2019-01-03 14:37:05 +00:00
|
|
|
const __m256i threes = _mm256_set1_epi16 (3);
|
2019-02-04 12:44:47 +00:00
|
|
|
const __m256i cmpmask = _mm256_slli_epi16 (threes, 7); // 0x0180 (avoid set1)
|
2018-12-20 08:51:44 +00:00
|
|
|
|
|
|
|
__m256i clipped = _mm256_min_epu16 (src, threes);
|
|
|
|
__m256i shifted = _mm256_slli_epi16 (clipped, 7);
|
|
|
|
__m256i cmpres = _mm256_cmpeq_epi8 (shifted, cmpmask);
|
|
|
|
uint32_t result = _mm256_movemask_epi8(cmpres);
|
|
|
|
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2018-12-12 19:27:09 +00:00
|
|
|
/**
|
|
|
|
* \brief Context derivation process of coeff_abs_significant_flag,
|
|
|
|
* parallelized to handle 16 coeffs at once
|
|
|
|
* \param pattern_sig_ctx pattern for current coefficient group
|
|
|
|
* \param scan_idx pixel scan type in use
|
|
|
|
* \param pos_xs column addresses of current scan positions
|
|
|
|
* \param pos_ys row addresses of current scan positions
|
|
|
|
* \param block_type log2 value of block size if square block, or 4 otherwise
|
|
|
|
* \param width width of the block
|
|
|
|
* \param texture_type texture type (TEXT_LUMA...)
|
|
|
|
* \returns ctx_inc for current scan position
|
|
|
|
*/
|
|
|
|
static INLINE __m256i kvz_context_get_sig_ctx_inc_16x16b(int32_t pattern_sig_ctx, uint32_t scan_idx, __m256i pos_xs,
|
|
|
|
__m256i pos_ys, int32_t block_type, int8_t texture_type)
|
|
|
|
{
|
|
|
|
const __m256i zero = _mm256_set1_epi8(0);
|
|
|
|
const __m256i ff = _mm256_set1_epi8(0xff);
|
|
|
|
|
|
|
|
const __m256i ones = _mm256_set1_epi16(1);
|
|
|
|
const __m256i twos = _mm256_set1_epi16(2);
|
|
|
|
const __m256i threes = _mm256_set1_epi16(3);
|
|
|
|
|
|
|
|
const __m256i ctx_ind_map[3] = {
|
|
|
|
_mm256_setr_epi16(
|
|
|
|
0, 2, 1, 6,
|
|
|
|
3, 4, 7, 6,
|
|
|
|
4, 5, 7, 8,
|
|
|
|
5, 8, 8, 8
|
|
|
|
),
|
|
|
|
_mm256_setr_epi16(
|
|
|
|
0, 1, 4, 5,
|
|
|
|
2, 3, 4, 5,
|
|
|
|
6, 6, 8, 8,
|
|
|
|
7, 7, 8, 8
|
|
|
|
),
|
|
|
|
_mm256_setr_epi16(
|
|
|
|
0, 2, 6, 7,
|
|
|
|
1, 3, 6, 7,
|
|
|
|
4, 4, 8, 8,
|
|
|
|
5, 5, 8, 8
|
|
|
|
),
|
|
|
|
};
|
|
|
|
|
|
|
|
int16_t offset;
|
|
|
|
if (block_type == 3)
|
|
|
|
if (scan_idx == SCAN_DIAG)
|
|
|
|
offset = 9;
|
|
|
|
else
|
|
|
|
offset = 15;
|
|
|
|
else
|
|
|
|
if (texture_type == 0)
|
|
|
|
offset = 21;
|
|
|
|
else
|
|
|
|
offset = 12;
|
|
|
|
|
|
|
|
__m256i offsets = _mm256_set1_epi16(offset);
|
|
|
|
|
|
|
|
// This will only ever be compared to 0, 1 and 2, so it's fine to cast down
|
|
|
|
// to 16b (and it should never be above 3 anyways)
|
|
|
|
__m256i pattern_sig_ctxs = _mm256_set1_epi16((int16_t)(MIN(0xffff, pattern_sig_ctx)));
|
|
|
|
__m256i pattern_sig_ctxs_eq_zero = _mm256_cmpeq_epi16(pattern_sig_ctxs, zero);
|
|
|
|
__m256i pattern_sig_ctxs_eq_one = _mm256_cmpeq_epi16(pattern_sig_ctxs, ones);
|
|
|
|
__m256i pattern_sig_ctxs_eq_two = _mm256_cmpeq_epi16(pattern_sig_ctxs, twos);
|
|
|
|
|
|
|
|
__m256i pattern_sig_ctxs_eq_1or2 = _mm256_or_si256 (pattern_sig_ctxs_eq_one,
|
|
|
|
pattern_sig_ctxs_eq_two);
|
|
|
|
__m256i pattern_sig_ctxs_lt3 = _mm256_or_si256 (pattern_sig_ctxs_eq_1or2,
|
|
|
|
pattern_sig_ctxs_eq_zero);
|
|
|
|
__m256i pattern_sig_ctxs_other = _mm256_xor_si256(pattern_sig_ctxs_lt3,
|
|
|
|
ff);
|
|
|
|
__m256i x_plus_y = _mm256_add_epi16 (pos_xs, pos_ys);
|
|
|
|
__m256i x_plus_y_zero = _mm256_cmpeq_epi16(x_plus_y, zero); // All these should be 0, preempts block_type_two rule
|
|
|
|
|
|
|
|
__m256i texture_types = _mm256_set1_epi16((int16_t)texture_type);
|
|
|
|
|
|
|
|
__m256i block_types = _mm256_set1_epi16((int16_t)block_type);
|
|
|
|
__m256i block_type_two = _mm256_cmpeq_epi16(block_types, twos); // All these should be ctx_ind_map[4 * pos_y + pos_x];
|
|
|
|
__m256i bt2_vals = ctx_ind_map[scan_idx];
|
|
|
|
__m256i bt2_vals_masked = _mm256_and_si256(bt2_vals, block_type_two);
|
|
|
|
|
|
|
|
__m256i pos_xs_in_subset = _mm256_and_si256(pos_xs, threes);
|
|
|
|
__m256i pos_ys_in_subset = _mm256_and_si256(pos_ys, threes);
|
|
|
|
|
|
|
|
__m256i cg_pos_xs = _mm256_srli_epi16(pos_xs, 2);
|
|
|
|
__m256i cg_pos_ys = _mm256_srli_epi16(pos_ys, 2);
|
|
|
|
__m256i cg_pos_xysums = _mm256_add_epi16 (cg_pos_xs, cg_pos_ys);
|
|
|
|
|
|
|
|
__m256i pos_xy_sums_in_subset = _mm256_add_epi16(pos_xs_in_subset, pos_ys_in_subset);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if (pattern_sig_ctx == 0) {
|
|
|
|
* switch (pos_x_in_subset + pos_y_in_subset) {
|
|
|
|
* case 0:
|
|
|
|
* cnt = 2;
|
|
|
|
* break;
|
|
|
|
* case 1:
|
|
|
|
* case 2:
|
|
|
|
* cnt = 1;
|
|
|
|
* break;
|
|
|
|
* default:
|
|
|
|
* cnt = 0;
|
|
|
|
* }
|
|
|
|
* }
|
|
|
|
*
|
|
|
|
* Equivalent to:
|
|
|
|
*
|
|
|
|
* if (pattern_sig_ctx == 0) {
|
|
|
|
* subamt = cnt <= 1 ? 1 : 0;
|
|
|
|
* pxyis_max3 = min(3, pos_x_in_subset + pos_y_in_subset);
|
|
|
|
* cnt = (3 - pxyis_max3) - subamt;
|
|
|
|
* }
|
|
|
|
*/
|
|
|
|
__m256i pxyis_lte_1 = _mm256_cmpgt_epi16(twos, pos_xy_sums_in_subset);
|
|
|
|
__m256i subamts = _mm256_and_si256 (pxyis_lte_1, ones);
|
|
|
|
__m256i pxyis_max3 = _mm256_min_epu16 (pos_xy_sums_in_subset, threes);
|
|
|
|
__m256i cnts_tmp = _mm256_sub_epi16 (threes, pxyis_max3);
|
|
|
|
__m256i cnts_sig_ctx_0 = _mm256_sub_epi16 (cnts_tmp, subamts);
|
|
|
|
__m256i cnts_sc0_masked = _mm256_and_si256 (cnts_sig_ctx_0, pattern_sig_ctxs_eq_zero);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if (pattern_sig_ctx == 1 || pattern_sig_ctx == 2) {
|
|
|
|
* if (pattern_sig_ctx == 1)
|
|
|
|
* subtrahend = pos_y_in_subset;
|
|
|
|
* else
|
|
|
|
* subtrahend = pos_x_in_subset;
|
|
|
|
* cnt = 2 - min(2, subtrahend);
|
|
|
|
* }
|
|
|
|
*/
|
|
|
|
__m256i pos_operands_ctx_1or2 = _mm256_blendv_epi8(pos_ys_in_subset,
|
|
|
|
pos_xs_in_subset,
|
|
|
|
pattern_sig_ctxs_eq_two);
|
|
|
|
|
|
|
|
__m256i pos_operands_max2 = _mm256_min_epu16 (pos_operands_ctx_1or2, twos);
|
|
|
|
__m256i cnts_sig_ctx_1or2 = _mm256_sub_epi16 (twos, pos_operands_max2);
|
|
|
|
__m256i cnts_sc12_masked = _mm256_and_si256 (cnts_sig_ctx_1or2, pattern_sig_ctxs_eq_1or2);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if (pattern_sig_ctx > 2)
|
|
|
|
* cnt = 2;
|
|
|
|
*/
|
|
|
|
__m256i cnts_scother_masked = _mm256_and_si256(twos, pattern_sig_ctxs_other);
|
|
|
|
|
|
|
|
// Select correct count
|
|
|
|
__m256i cnts_sc012_masked = _mm256_or_si256 (cnts_sc0_masked, cnts_sc12_masked);
|
|
|
|
__m256i cnts = _mm256_or_si256 (cnts_scother_masked, cnts_sc012_masked);
|
|
|
|
|
|
|
|
// Compute final values
|
|
|
|
__m256i textype_eq_0 = _mm256_cmpeq_epi16(texture_types, zero);
|
|
|
|
__m256i cg_pos_sums_gt_0 = _mm256_cmpgt_epi16(cg_pos_xysums, zero);
|
|
|
|
__m256i tmpcond = _mm256_and_si256 (textype_eq_0, cg_pos_sums_gt_0);
|
|
|
|
__m256i tmp = _mm256_and_si256 (tmpcond, threes);
|
|
|
|
__m256i tmp_with_offsets = _mm256_add_epi16 (tmp, offsets);
|
|
|
|
__m256i rv_noshortcirc = _mm256_add_epi16 (cnts, tmp_with_offsets);
|
|
|
|
|
|
|
|
// Ol' sprite mask method works here!
|
|
|
|
__m256i rv1 = _mm256_andnot_si256(block_type_two, rv_noshortcirc);
|
|
|
|
__m256i rv2 = _mm256_or_si256 (rv1, bt2_vals_masked);
|
|
|
|
__m256i rv = _mm256_andnot_si256(x_plus_y_zero, rv2);
|
|
|
|
return rv;
|
|
|
|
}
|
|
|
|
|
2018-11-29 16:00:05 +00:00
|
|
|
void kvz_encode_coeff_nxn_avx2(encoder_state_t * const state,
|
|
|
|
cabac_data_t * const cabac,
|
|
|
|
const coeff_t *coeff,
|
|
|
|
uint8_t width,
|
|
|
|
uint8_t type,
|
|
|
|
int8_t scan_mode,
|
|
|
|
int8_t tr_skip)
|
|
|
|
{
|
|
|
|
const encoder_control_t * const encoder = state->encoder_control;
|
|
|
|
int c1 = 1;
|
|
|
|
uint8_t last_coeff_x = 0;
|
|
|
|
uint8_t last_coeff_y = 0;
|
|
|
|
int32_t i;
|
2019-02-04 13:48:53 +00:00
|
|
|
uint32_t sig_coeffgroup_nzs[8 * 8] = { 0 };
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
int8_t be_valid = encoder->cfg.signhide_enable;
|
|
|
|
int32_t scan_pos_sig;
|
|
|
|
uint32_t go_rice_param = 0;
|
2018-12-10 21:47:07 +00:00
|
|
|
uint32_t ctx_sig;
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
// CONSTANTS
|
|
|
|
const uint32_t num_blk_side = width >> TR_MIN_LOG2_SIZE;
|
|
|
|
const uint32_t log2_block_size = kvz_g_convert_to_bit[width] + 2;
|
|
|
|
const uint32_t *scan =
|
|
|
|
kvz_g_sig_last_scan[scan_mode][log2_block_size - 1];
|
|
|
|
const uint32_t *scan_cg = g_sig_last_scan_cg[log2_block_size - 2][scan_mode];
|
2018-12-10 16:29:32 +00:00
|
|
|
const uint32_t num_blocks = num_blk_side * num_blk_side;
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-03 15:21:22 +00:00
|
|
|
const __m256i zero = _mm256_set1_epi8(0);
|
2018-12-12 13:33:49 +00:00
|
|
|
const __m256i ones = _mm256_set1_epi16(1);
|
|
|
|
const __m256i twos = _mm256_set1_epi16(2);
|
2018-12-03 15:21:22 +00:00
|
|
|
|
2018-11-29 16:00:05 +00:00
|
|
|
// Init base contexts according to block type
|
|
|
|
cabac_ctx_t *base_coeff_group_ctx = &(cabac->ctx.cu_sig_coeff_group_model[type]);
|
|
|
|
cabac_ctx_t *baseCtx = (type == 0) ? &(cabac->ctx.cu_sig_model_luma[0]) :
|
|
|
|
&(cabac->ctx.cu_sig_model_chroma[0]);
|
|
|
|
|
|
|
|
// Scan all coeff groups to find out which of them have coeffs.
|
2019-02-04 13:48:53 +00:00
|
|
|
// Populate sig_coeffgroup_nzs with that info.
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-03 15:21:22 +00:00
|
|
|
// NOTE: Modified the functionality a bit, sig_coeffgroup_flag used to be
|
|
|
|
// 1 if true and 0 if false, now it's "undefined but nonzero" if true and
|
|
|
|
// 0 if false (not actually undefined, it's a bitmask representing the
|
|
|
|
// significant coefficients' position in the group which in itself could
|
|
|
|
// be useful information)
|
2018-12-12 10:54:44 +00:00
|
|
|
int32_t scan_cg_last = -1;
|
2018-12-10 16:29:32 +00:00
|
|
|
|
2018-12-12 10:54:44 +00:00
|
|
|
for (int32_t i = 0; i < num_blocks; i++) {
|
2018-12-10 16:29:32 +00:00
|
|
|
const uint32_t cg_id = scan_cg[i];
|
|
|
|
const uint32_t n_xbits = log2_block_size - 2; // How many lowest bits of scan_cg represent X coord
|
|
|
|
const uint32_t cg_x = cg_id & ((1 << n_xbits) - 1);
|
|
|
|
const uint32_t cg_y = cg_id >> n_xbits;
|
|
|
|
|
|
|
|
const uint32_t cg_pos = cg_y * width * 4 + cg_x * 4;
|
|
|
|
const uint32_t cg_pos_y = (cg_pos >> log2_block_size) >> TR_MIN_LOG2_SIZE;
|
|
|
|
const uint32_t cg_pos_x = (cg_pos & (width - 1)) >> TR_MIN_LOG2_SIZE;
|
2019-02-04 13:48:53 +00:00
|
|
|
const uint32_t idx = cg_pos_x + cg_pos_y * num_blk_side;
|
2018-12-10 16:29:32 +00:00
|
|
|
|
2018-12-18 18:42:09 +00:00
|
|
|
__m128d coeffs_d_upper = _mm_setzero_pd();
|
|
|
|
__m128d coeffs_d_lower = _mm_setzero_pd();
|
2018-12-10 16:29:32 +00:00
|
|
|
__m128i coeffs_upper;
|
|
|
|
__m128i coeffs_lower;
|
|
|
|
__m256i cur_coeffs;
|
|
|
|
|
|
|
|
coeffs_d_upper = _mm_loadl_pd(coeffs_d_upper, (double *)(coeff + cg_pos + 0 * width));
|
|
|
|
coeffs_d_upper = _mm_loadh_pd(coeffs_d_upper, (double *)(coeff + cg_pos + 1 * width));
|
|
|
|
coeffs_d_lower = _mm_loadl_pd(coeffs_d_lower, (double *)(coeff + cg_pos + 2 * width));
|
|
|
|
coeffs_d_lower = _mm_loadh_pd(coeffs_d_lower, (double *)(coeff + cg_pos + 3 * width));
|
|
|
|
|
|
|
|
coeffs_upper = _mm_castpd_si128(coeffs_d_upper);
|
|
|
|
coeffs_lower = _mm_castpd_si128(coeffs_d_lower);
|
|
|
|
|
|
|
|
cur_coeffs = _mm256_insertf128_si256(_mm256_castsi128_si256(coeffs_upper),
|
|
|
|
coeffs_lower,
|
|
|
|
1);
|
|
|
|
|
|
|
|
__m256i coeffs_zero = _mm256_cmpeq_epi16(cur_coeffs, zero);
|
|
|
|
|
|
|
|
uint32_t nz_coeffs_2b = ~((uint32_t)_mm256_movemask_epi8(coeffs_zero));
|
2019-02-04 13:48:53 +00:00
|
|
|
sig_coeffgroup_nzs[idx] = nz_coeffs_2b;
|
2018-12-10 16:29:32 +00:00
|
|
|
|
2018-12-12 10:54:44 +00:00
|
|
|
if (nz_coeffs_2b)
|
2018-12-10 16:29:32 +00:00
|
|
|
scan_cg_last = i;
|
2018-11-29 16:00:05 +00:00
|
|
|
}
|
|
|
|
// Rest of the code assumes at least one non-zero coeff.
|
2018-12-12 10:54:44 +00:00
|
|
|
assert(scan_cg_last >= 0);
|
2018-12-10 16:29:32 +00:00
|
|
|
|
|
|
|
ALIGNED(64) int16_t coeff_reord[LCU_WIDTH * LCU_WIDTH];
|
2018-12-14 19:09:54 +00:00
|
|
|
uint32_t pos_last, scan_pos_last;
|
|
|
|
|
|
|
|
{
|
|
|
|
__m256i coeffs_r;
|
|
|
|
for (int32_t i = 0; i <= scan_cg_last; i++) {
|
|
|
|
int32_t subpos = i * 16;
|
2018-12-18 17:34:35 +00:00
|
|
|
scanord_read_vector(&coeff, scan, scan_mode, subpos, width, &coeffs_r, 1);
|
2018-12-14 19:09:54 +00:00
|
|
|
_mm256_store_si256((__m256i *)(coeff_reord + subpos), coeffs_r);
|
|
|
|
}
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2019-02-04 13:48:53 +00:00
|
|
|
// Find the last coeff by going backwards in scan order. With cmpeq_epi16
|
|
|
|
// and movemask, we can generate a dword with 16 2-bit masks that are 11
|
|
|
|
// for zero words in the coeff vector, and 00 for nonzero words. By
|
|
|
|
// inverting the bits and counting leading zeros, we can determine the
|
|
|
|
// number of zero bytes in the vector counting from high to low memory
|
|
|
|
// addresses; subtract that from 31 and divide by 2 to get the offset of
|
|
|
|
// the last nonzero word.
|
2018-12-14 19:09:54 +00:00
|
|
|
uint32_t baseaddr = scan_cg_last * 16;
|
|
|
|
__m256i cur_coeffs_zeros = _mm256_cmpeq_epi16(coeffs_r, zero);
|
|
|
|
uint32_t nz_bytes = ~(_mm256_movemask_epi8(cur_coeffs_zeros));
|
|
|
|
scan_pos_last = baseaddr + ((31 - _lzcnt_u32(nz_bytes)) >> 1);
|
|
|
|
pos_last = scan[scan_pos_last];
|
|
|
|
}
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
// transform skip flag
|
|
|
|
if(width == 4 && encoder->cfg.trskip_enable) {
|
|
|
|
cabac->cur_ctx = (type == 0) ? &(cabac->ctx.transform_skip_model_luma) : &(cabac->ctx.transform_skip_model_chroma);
|
|
|
|
CABAC_BIN(cabac, tr_skip, "transform_skip_flag");
|
|
|
|
}
|
|
|
|
|
|
|
|
last_coeff_x = pos_last & (width - 1);
|
|
|
|
last_coeff_y = (uint8_t)(pos_last >> log2_block_size);
|
|
|
|
|
|
|
|
// Code last_coeff_x and last_coeff_y
|
2019-02-04 12:55:41 +00:00
|
|
|
kvz_encode_last_significant_xy(cabac,
|
|
|
|
last_coeff_x,
|
|
|
|
last_coeff_y,
|
|
|
|
width,
|
|
|
|
width,
|
|
|
|
type,
|
|
|
|
scan_mode);
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-03 19:50:57 +00:00
|
|
|
scan_pos_sig = scan_pos_last;
|
|
|
|
|
2018-12-12 13:33:49 +00:00
|
|
|
ALIGNED(64) uint16_t abs_coeff[16];
|
|
|
|
ALIGNED(32) uint16_t abs_coeff_buf_sb[16];
|
|
|
|
ALIGNED(32) int16_t pos_ys_buf[16];
|
|
|
|
ALIGNED(32) int16_t pos_xs_buf[16];
|
2018-12-12 19:27:09 +00:00
|
|
|
ALIGNED(32) int16_t ctx_sig_buf[16];
|
2018-12-12 13:33:49 +00:00
|
|
|
|
2018-12-03 19:50:57 +00:00
|
|
|
abs_coeff[0] = abs(coeff[pos_last]);
|
|
|
|
uint32_t coeff_signs = (coeff[pos_last] < 0);
|
|
|
|
int32_t num_non_zero = 1;
|
|
|
|
int32_t last_nz_pos_in_cg = scan_pos_sig;
|
|
|
|
int32_t first_nz_pos_in_cg = scan_pos_sig;
|
|
|
|
scan_pos_sig--;
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
// significant_coeff_flag
|
|
|
|
for (i = scan_cg_last; i >= 0; i--) {
|
|
|
|
int32_t sub_pos = i << 4; // LOG2_SCAN_SET_SIZE;
|
|
|
|
int32_t cg_blk_pos = scan_cg[i];
|
|
|
|
int32_t cg_pos_y = cg_blk_pos / num_blk_side;
|
|
|
|
int32_t cg_pos_x = cg_blk_pos - (cg_pos_y * num_blk_side);
|
|
|
|
|
|
|
|
go_rice_param = 0;
|
|
|
|
|
|
|
|
if (i == scan_cg_last || i == 0) {
|
2019-02-04 13:48:53 +00:00
|
|
|
sig_coeffgroup_nzs[cg_blk_pos] = 1;
|
2018-11-29 16:00:05 +00:00
|
|
|
} else {
|
2019-02-04 13:48:53 +00:00
|
|
|
uint32_t sig_coeff_group = (sig_coeffgroup_nzs[cg_blk_pos] != 0);
|
|
|
|
uint32_t ctx_sig = kvz_context_get_sig_coeff_group(sig_coeffgroup_nzs, cg_pos_x,
|
2018-11-29 16:00:05 +00:00
|
|
|
cg_pos_y, width);
|
|
|
|
cabac->cur_ctx = &base_coeff_group_ctx[ctx_sig];
|
|
|
|
CABAC_BIN(cabac, sig_coeff_group, "coded_sub_block_flag");
|
|
|
|
}
|
|
|
|
|
2019-02-04 13:48:53 +00:00
|
|
|
if (sig_coeffgroup_nzs[cg_blk_pos]) {
|
|
|
|
int32_t pattern_sig_ctx = kvz_context_calc_pattern_sig_ctx(sig_coeffgroup_nzs,
|
2018-11-29 16:00:05 +00:00
|
|
|
cg_pos_x, cg_pos_y, width);
|
|
|
|
|
2019-02-04 13:48:53 +00:00
|
|
|
// A mask with the first 16-bit word unmasked (bits set ie. 0xffff)
|
2018-12-10 21:47:07 +00:00
|
|
|
const __m256i coeff_pos_zero = _mm256_castsi128_si256(_mm_cvtsi32_si128(0xffff));
|
2019-02-04 13:48:53 +00:00
|
|
|
|
2018-12-10 21:47:07 +00:00
|
|
|
const __m128i log2_block_size_128 = _mm_cvtsi32_si128(log2_block_size);
|
2018-12-10 16:29:32 +00:00
|
|
|
|
2018-12-10 17:21:27 +00:00
|
|
|
__m256i coeffs = _mm256_load_si256((__m256i *)(coeff_reord + sub_pos));
|
2018-12-10 16:29:32 +00:00
|
|
|
__m256i sigs_inv = _mm256_cmpeq_epi16(coeffs, zero);
|
|
|
|
__m256i is = _mm256_set1_epi16(i);
|
|
|
|
__m256i is_zero = _mm256_cmpeq_epi16(is, zero);
|
2019-02-04 14:05:43 +00:00
|
|
|
__m256i coeffs_negative = _mm256_cmpgt_epi16(zero, coeffs);
|
2018-12-10 21:47:07 +00:00
|
|
|
|
|
|
|
__m256i masked_coeffs = _mm256_andnot_si256(sigs_inv, coeffs);
|
|
|
|
__m256i abs_coeffs = _mm256_abs_epi16(masked_coeffs);
|
|
|
|
|
|
|
|
// TODO: obtain 16-bit block positions, maybe? :P
|
|
|
|
__m256i blk_poses_hi = _mm256_loadu_si256((__m256i *)(scan + sub_pos + 8));
|
|
|
|
__m256i blk_poses_lo = _mm256_loadu_si256((__m256i *)(scan + sub_pos + 0));
|
|
|
|
__m256i blk_poses_tmp = _mm256_packs_epi32(blk_poses_lo, blk_poses_hi);
|
|
|
|
__m256i blk_poses = _mm256_permute4x64_epi64(blk_poses_tmp, _MM_SHUFFLE(3, 1, 2, 0));
|
|
|
|
|
|
|
|
__m256i pos_ys = _mm256_srl_epi16(blk_poses, log2_block_size_128);
|
|
|
|
__m256i pos_xs = _mm256_sub_epi16(blk_poses, _mm256_sll_epi16(pos_ys, log2_block_size_128));
|
|
|
|
|
|
|
|
_mm256_store_si256((__m256i *)pos_ys_buf, pos_ys);
|
|
|
|
_mm256_store_si256((__m256i *)pos_xs_buf, pos_xs);
|
|
|
|
|
|
|
|
__m256i encode_sig_coeff_flags_inv = _mm256_andnot_si256(is_zero, coeff_pos_zero);
|
|
|
|
|
|
|
|
get_first_last_nz_int16(masked_coeffs, &first_nz_pos_in_cg, &last_nz_pos_in_cg);
|
2018-12-12 13:33:49 +00:00
|
|
|
_mm256_store_si256((__m256i *)abs_coeff_buf_sb, abs_coeffs);
|
2018-12-10 21:47:07 +00:00
|
|
|
|
2018-12-12 19:27:09 +00:00
|
|
|
__m256i ctx_sigs = kvz_context_get_sig_ctx_inc_16x16b(pattern_sig_ctx, scan_mode, pos_xs, pos_ys,
|
|
|
|
log2_block_size, type);
|
|
|
|
|
|
|
|
_mm256_store_si256((__m256i *)ctx_sig_buf, ctx_sigs);
|
|
|
|
|
2018-12-10 21:47:07 +00:00
|
|
|
uint32_t esc_flags = ~(_mm256_movemask_epi8(encode_sig_coeff_flags_inv));
|
|
|
|
uint32_t sigs = ~(_mm256_movemask_epi8(sigs_inv));
|
2019-02-04 14:05:43 +00:00
|
|
|
uint32_t coeff_sign_buf = _mm256_movemask_epi8(coeffs_negative);
|
2018-12-10 16:29:32 +00:00
|
|
|
|
2018-11-29 16:00:05 +00:00
|
|
|
for (; scan_pos_sig >= sub_pos; scan_pos_sig--) {
|
2018-12-10 21:47:07 +00:00
|
|
|
uint32_t id = scan_pos_sig - sub_pos;
|
2019-02-04 14:05:43 +00:00
|
|
|
uint32_t shift = (id << 1) + 1;
|
2018-12-10 21:47:07 +00:00
|
|
|
|
2019-02-04 14:05:43 +00:00
|
|
|
uint32_t curr_sig = (sigs >> shift) & 1;
|
|
|
|
uint32_t curr_esc_flag = (esc_flags >> shift) & 1;
|
|
|
|
uint32_t curr_coeff_sign = (coeff_sign_buf >> shift) & 1;
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-10 21:47:07 +00:00
|
|
|
if (curr_esc_flag | num_non_zero) {
|
2018-12-12 19:27:09 +00:00
|
|
|
ctx_sig = ctx_sig_buf[id];
|
2018-11-29 16:00:05 +00:00
|
|
|
cabac->cur_ctx = &baseCtx[ctx_sig];
|
2018-12-10 21:47:07 +00:00
|
|
|
CABAC_BIN(cabac, curr_sig, "sig_coeff_flag");
|
2018-11-29 16:00:05 +00:00
|
|
|
}
|
|
|
|
|
2018-12-10 21:47:07 +00:00
|
|
|
if (curr_sig) {
|
2018-12-14 19:09:54 +00:00
|
|
|
abs_coeff[num_non_zero] = abs_coeff_buf_sb[id];
|
2018-12-10 21:47:07 +00:00
|
|
|
coeff_signs = 2 * coeff_signs + curr_coeff_sign;
|
2018-11-29 16:00:05 +00:00
|
|
|
num_non_zero++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
scan_pos_sig = sub_pos - 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (num_non_zero > 0) {
|
|
|
|
bool sign_hidden = last_nz_pos_in_cg - first_nz_pos_in_cg >= 4 /* SBH_THRESHOLD */
|
|
|
|
&& !encoder->cfg.lossless;
|
|
|
|
uint32_t ctx_set = (i > 0 && type == 0) ? 2 : 0;
|
|
|
|
cabac_ctx_t *base_ctx_mod;
|
2018-12-14 19:09:54 +00:00
|
|
|
int32_t num_c1_flag, first_c2_flag_idx, idx;
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-14 19:09:54 +00:00
|
|
|
__m256i abs_coeffs = _mm256_load_si256((__m256i *)abs_coeff);
|
2018-12-12 13:33:49 +00:00
|
|
|
__m256i coeffs_gt1 = _mm256_cmpgt_epi16(abs_coeffs, ones);
|
|
|
|
__m256i coeffs_gt2 = _mm256_cmpgt_epi16(abs_coeffs, twos);
|
|
|
|
uint32_t coeffs_gt1_bits = _mm256_movemask_epi8(coeffs_gt1);
|
|
|
|
uint32_t coeffs_gt2_bits = _mm256_movemask_epi8(coeffs_gt2);
|
|
|
|
|
2018-11-29 16:00:05 +00:00
|
|
|
if (c1 == 0) {
|
|
|
|
ctx_set++;
|
|
|
|
}
|
|
|
|
|
|
|
|
base_ctx_mod = (type == 0) ? &(cabac->ctx.cu_one_model_luma[4 * ctx_set]) :
|
|
|
|
&(cabac->ctx.cu_one_model_chroma[4 * ctx_set]);
|
|
|
|
num_c1_flag = MIN(num_non_zero, C1FLAG_NUMBER);
|
|
|
|
first_c2_flag_idx = -1;
|
|
|
|
|
2018-12-14 13:45:46 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* c1s_pattern is 16 base-4 numbers: 3, 3, 3, ... , 3, 2 (c1 will never
|
|
|
|
* be less than 0 or greater than 3, so two bits per iter are enough).
|
|
|
|
* It's essentially the values that c1 will be for the next iteration as
|
|
|
|
* long as we have not encountered any >1 symbols. Count how long run of
|
|
|
|
* such symbols there is in the beginning of this CG, and zero all c1's
|
|
|
|
* that are located at or after the first >1 symbol.
|
|
|
|
*/
|
|
|
|
const uint32_t c1s_pattern = 0xfffffffe;
|
|
|
|
uint32_t n_nongt1_bits = _tzcnt_u32(coeffs_gt1_bits);
|
2018-12-14 19:09:54 +00:00
|
|
|
uint32_t c1s_nextiter = _bzhi_u32(c1s_pattern, n_nongt1_bits);
|
|
|
|
first_c2_flag_idx = n_nongt1_bits >> 1;
|
2018-12-14 13:45:46 +00:00
|
|
|
|
|
|
|
c1 = 1;
|
2018-11-29 16:00:05 +00:00
|
|
|
for (idx = 0; idx < num_c1_flag; idx++) {
|
2019-02-04 14:05:43 +00:00
|
|
|
uint32_t shift = idx << 1;
|
|
|
|
uint32_t symbol = (coeffs_gt1_bits >> shift) & 1;
|
2018-12-12 13:33:49 +00:00
|
|
|
|
2018-11-29 16:00:05 +00:00
|
|
|
cabac->cur_ctx = &base_ctx_mod[c1];
|
|
|
|
CABAC_BIN(cabac, symbol, "coeff_abs_level_greater1_flag");
|
|
|
|
|
2019-02-04 14:05:43 +00:00
|
|
|
c1 = (c1s_nextiter >> shift) & 3;
|
2018-11-29 16:00:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (c1 == 0) {
|
|
|
|
base_ctx_mod = (type == 0) ? &(cabac->ctx.cu_abs_model_luma[ctx_set]) :
|
|
|
|
&(cabac->ctx.cu_abs_model_chroma[ctx_set]);
|
|
|
|
|
|
|
|
if (first_c2_flag_idx != -1) {
|
2019-02-04 14:05:43 +00:00
|
|
|
uint32_t shift = (first_c2_flag_idx << 1) + 1;
|
|
|
|
uint8_t symbol = (coeffs_gt2_bits >> shift) & 1;
|
2018-12-12 13:33:49 +00:00
|
|
|
cabac->cur_ctx = &base_ctx_mod[0];
|
|
|
|
|
2018-11-29 16:00:05 +00:00
|
|
|
CABAC_BIN(cabac, symbol, "coeff_abs_level_greater2_flag");
|
|
|
|
}
|
|
|
|
}
|
2018-12-05 19:25:08 +00:00
|
|
|
int32_t shiftamt = (be_valid && sign_hidden) ? 1 : 0;
|
|
|
|
int32_t nnz = num_non_zero - shiftamt;
|
|
|
|
coeff_signs >>= shiftamt;
|
|
|
|
if (!cabac->only_count) {
|
|
|
|
if (encoder->cfg.crypto_features & KVZ_CRYPTO_TRANSF_COEFF_SIGNS) {
|
|
|
|
coeff_signs ^= kvz_crypto_get_key(state->crypto_hdl, nnz);
|
|
|
|
}
|
2018-11-29 16:00:05 +00:00
|
|
|
}
|
2018-12-05 19:25:08 +00:00
|
|
|
CABAC_BINS_EP(cabac, coeff_signs, nnz, "coeff_sign_flag");
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
if (c1 == 0 || num_non_zero > C1FLAG_NUMBER) {
|
2018-12-14 19:09:54 +00:00
|
|
|
|
|
|
|
const __m256i ones = _mm256_set1_epi16(1);
|
|
|
|
|
|
|
|
__m256i abs_coeffs_gt1 = _mm256_cmpgt_epi16 (abs_coeffs, ones);
|
|
|
|
uint32_t acgt1_bits = _mm256_movemask_epi8(abs_coeffs_gt1);
|
|
|
|
uint32_t first_acgt1_bpos = _tzcnt_u32(acgt1_bits);
|
|
|
|
|
2018-12-20 08:51:44 +00:00
|
|
|
uint32_t abs_coeffs_base4 = pack_16x16b_to_16x2b(abs_coeffs);
|
2018-12-14 19:09:54 +00:00
|
|
|
|
|
|
|
const uint32_t ones_base4 = 0x55555555;
|
|
|
|
const uint32_t twos_base4 = 0xaaaaaaaa;
|
|
|
|
|
|
|
|
const uint32_t c1flag_number_mask_inv = 0xffffffff << (C1FLAG_NUMBER << 1);
|
|
|
|
const uint32_t c1flag_number_mask = ~c1flag_number_mask_inv;
|
|
|
|
|
|
|
|
// The addition will not overflow between 2-bit atoms because
|
|
|
|
// first_coeff2s will only be 1 or 0, and the other addend is 2
|
|
|
|
uint32_t first_coeff2s = _bzhi_u32(ones_base4, first_acgt1_bpos + 2);
|
|
|
|
uint32_t base_levels = first_coeff2s + twos_base4;
|
|
|
|
|
|
|
|
base_levels &= c1flag_number_mask;
|
|
|
|
base_levels |= (ones_base4 & c1flag_number_mask_inv);
|
|
|
|
|
2019-02-01 14:26:21 +00:00
|
|
|
uint32_t encode_decisions = u32vec_cmpgt_epu2(base_levels, abs_coeffs_base4);
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
for (idx = 0; idx < num_non_zero; idx++) {
|
|
|
|
|
2019-02-04 14:05:43 +00:00
|
|
|
uint32_t shift = idx << 1;
|
|
|
|
uint32_t dont_encode_curr = (encode_decisions >> shift);
|
|
|
|
int16_t base_level = (base_levels >> shift) & 3;
|
2018-12-14 19:09:54 +00:00
|
|
|
|
|
|
|
uint16_t curr_abs_coeff = abs_coeff[idx];
|
|
|
|
|
2019-02-01 14:26:21 +00:00
|
|
|
if (!(dont_encode_curr & 2)) {
|
2018-12-14 19:09:54 +00:00
|
|
|
uint16_t level_diff = curr_abs_coeff - base_level;
|
2018-12-05 19:25:08 +00:00
|
|
|
if (!cabac->only_count && (encoder->cfg.crypto_features & KVZ_CRYPTO_TRANSF_COEFFS)) {
|
2018-12-14 19:09:54 +00:00
|
|
|
kvz_cabac_write_coeff_remain_encry(state, cabac, level_diff, go_rice_param, base_level);
|
2018-12-05 19:25:08 +00:00
|
|
|
} else {
|
2018-12-14 19:09:54 +00:00
|
|
|
kvz_cabac_write_coeff_remain(cabac, level_diff, go_rice_param);
|
2018-12-05 19:25:08 +00:00
|
|
|
}
|
2018-11-29 16:00:05 +00:00
|
|
|
|
2018-12-14 19:09:54 +00:00
|
|
|
if (curr_abs_coeff > 3 * (1 << go_rice_param)) {
|
2018-11-29 16:00:05 +00:00
|
|
|
go_rice_param = MIN(go_rice_param + 1, 4);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2018-12-03 19:50:57 +00:00
|
|
|
last_nz_pos_in_cg = -1;
|
|
|
|
first_nz_pos_in_cg = 16;
|
|
|
|
num_non_zero = 0;
|
|
|
|
coeff_signs = 0;
|
2018-11-29 16:00:05 +00:00
|
|
|
}
|
|
|
|
}
|
2019-02-04 17:52:24 +00:00
|
|
|
#endif // COMPILE_INTEL_AVX2
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
int kvz_strategy_register_encode_avx2(void* opaque, uint8_t bitdepth)
|
|
|
|
{
|
|
|
|
bool success = true;
|
|
|
|
|
2019-02-04 17:52:24 +00:00
|
|
|
#if COMPILE_INTEL_AVX2
|
2018-11-29 16:00:05 +00:00
|
|
|
success &= kvz_strategyselector_register(opaque, "encode_coeff_nxn", "avx2", 40, &kvz_encode_coeff_nxn_avx2);
|
2019-02-04 17:52:24 +00:00
|
|
|
#endif
|
2018-11-29 16:00:05 +00:00
|
|
|
|
|
|
|
return success;
|
|
|
|
}
|