/* * Copyright (c) 2017, 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 <emmintrin.h> #include <smmintrin.h> #include "config/av1_rtcd.h" #include "av1/common/warped_motion.h" /* This is a modified version of 'av1_warped_filter' from warped_motion.c: * Each coefficient is stored in 8 bits instead of 16 bits * The coefficients are rearranged in the column order 0, 2, 4, 6, 1, 3, 5, 7 This is done in order to avoid overflow: Since the tap with the largest coefficient could be any of taps 2, 3, 4 or 5, we can't use the summation order ((0 + 1) + (4 + 5)) + ((2 + 3) + (6 + 7)) used in the regular convolve functions. Instead, we use the summation order ((0 + 2) + (4 + 6)) + ((1 + 3) + (5 + 7)). The rearrangement of coefficients in this table is so that we can get the coefficients into the correct order more quickly. */ /* clang-format off */ DECLARE_ALIGNED(8, const int8_t, av1_filter_8bit[WARPEDPIXEL_PREC_SHIFTS * 3 + 1][8]) = …; /* clang-format on */ // Shuffle masks: we want to convert a sequence of bytes 0, 1, 2, ..., 15 // in an SSE register into two sequences: // 0, 2, 2, 4, ..., 12, 12, 14, <don't care> // 1, 3, 3, 5, ..., 13, 13, 15, <don't care> DECLARE_ALIGNED(16, static const uint8_t, even_mask[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, odd_mask[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_alpha0_mask01[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_alpha0_mask23[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_alpha0_mask45[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_alpha0_mask67[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_gamma0_mask0[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_gamma0_mask1[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_gamma0_mask2[16]) = …; DECLARE_ALIGNED(16, static const uint8_t, shuffle_gamma0_mask3[16]) = …; static inline void filter_src_pixels(__m128i src, __m128i *tmp, __m128i *coeff, const int offset_bits_horiz, const int reduce_bits_horiz, int k) { … } static inline void prepare_horizontal_filter_coeff(int alpha, int sx, __m128i *coeff) { … } static inline void prepare_horizontal_filter_coeff_alpha0(int sx, __m128i *coeff) { … } static inline void horizontal_filter(__m128i src, __m128i *tmp, int sx, int alpha, int k, const int offset_bits_horiz, const int reduce_bits_horiz) { … } static inline void warp_horizontal_filter(const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, int32_t sx4, int alpha, int beta, int p_height, int height, int i, const int offset_bits_horiz, const int reduce_bits_horiz) { … } static inline void warp_horizontal_filter_alpha0( const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, int32_t sx4, int alpha, int beta, int p_height, int height, int i, const int offset_bits_horiz, const int reduce_bits_horiz) { … } static inline void warp_horizontal_filter_beta0( const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, int32_t sx4, int alpha, int beta, int p_height, int height, int i, const int offset_bits_horiz, const int reduce_bits_horiz) { … } static inline void warp_horizontal_filter_alpha0_beta0( const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, int32_t sx4, int alpha, int beta, int p_height, int height, int i, const int offset_bits_horiz, const int reduce_bits_horiz) { … } static inline void unpack_weights_and_set_round_const( ConvolveParams *conv_params, const int round_bits, const int offset_bits, __m128i *res_sub_const, __m128i *round_bits_const, __m128i *wt) { … } static inline void prepare_vertical_filter_coeffs(int gamma, int sy, __m128i *coeffs) { … } static inline void prepare_vertical_filter_coeffs_gamma0(int sy, __m128i *coeffs) { … } static inline void filter_src_pixels_vertical(__m128i *tmp, __m128i *coeffs, __m128i *res_lo, __m128i *res_hi, int k) { … } static inline void store_vertical_filter_output( __m128i *res_lo, __m128i *res_hi, const __m128i *res_add_const, const __m128i *wt, const __m128i *res_sub_const, __m128i *round_bits_const, uint8_t *pred, ConvolveParams *conv_params, int i, int j, int k, const int reduce_bits_vert, int p_stride, int p_width, const int round_bits) { … } static inline void warp_vertical_filter( uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, int16_t delta, int p_height, int p_stride, int p_width, int i, int j, int sy4, const int reduce_bits_vert, const __m128i *res_add_const, const int round_bits, const int offset_bits) { … } static inline void warp_vertical_filter_gamma0( uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, int16_t delta, int p_height, int p_stride, int p_width, int i, int j, int sy4, const int reduce_bits_vert, const __m128i *res_add_const, const int round_bits, const int offset_bits) { … } static inline void warp_vertical_filter_delta0( uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, int16_t delta, int p_height, int p_stride, int p_width, int i, int j, int sy4, const int reduce_bits_vert, const __m128i *res_add_const, const int round_bits, const int offset_bits) { … } static inline void warp_vertical_filter_gamma0_delta0( uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, int16_t delta, int p_height, int p_stride, int p_width, int i, int j, int sy4, const int reduce_bits_vert, const __m128i *res_add_const, const int round_bits, const int offset_bits) { … } static inline void prepare_warp_vertical_filter( uint8_t *pred, __m128i *tmp, ConvolveParams *conv_params, int16_t gamma, int16_t delta, int p_height, int p_stride, int p_width, int i, int j, int sy4, const int reduce_bits_vert, const __m128i *res_add_const, const int round_bits, const int offset_bits) { … } static inline void prepare_warp_horizontal_filter( const uint8_t *ref, __m128i *tmp, int stride, int32_t ix4, int32_t iy4, int32_t sx4, int alpha, int beta, int p_height, int height, int i, const int offset_bits_horiz, const int reduce_bits_horiz) { … } void av1_warp_affine_sse4_1(const int32_t *mat, const uint8_t *ref, int width, int height, int stride, uint8_t *pred, int p_col, int p_row, int p_width, int p_height, int p_stride, int subsampling_x, int subsampling_y, ConvolveParams *conv_params, int16_t alpha, int16_t beta, int16_t gamma, int16_t delta) { … }
Codebase Browser
chromium