/* * 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. */ #ifndef AOM_AV1_COMMON_CDEF_BLOCK_SIMD_H_ #define AOM_AV1_COMMON_CDEF_BLOCK_SIMD_H_ #include "config/aom_config.h" #include "config/av1_rtcd.h" #include "av1/common/cdef_block.h" /* partial A is a 16-bit vector of the form: [x8 x7 x6 x5 x4 x3 x2 x1] and partial B has the form: [0 y1 y2 y3 y4 y5 y6 y7]. This function computes (x1^2+y1^2)*C1 + (x2^2+y2^2)*C2 + ... (x7^2+y2^7)*C7 + (x8^2+0^2)*C8 where the C1..C8 constants are in const1 and const2. */ static inline v128 fold_mul_and_sum(v128 partiala, v128 partialb, v128 const1, v128 const2) { … } static inline v128 hsum4(v128 x0, v128 x1, v128 x2, v128 x3) { … } /* Computes cost for directions 0, 5, 6 and 7. We can call this function again to compute the remaining directions. */ static inline v128 compute_directions(v128 lines[8], int32_t tmp_cost1[4]) { … } /* transpose and reverse the order of the lines -- equivalent to a 90-degree counter-clockwise rotation of the pixels. */ static inline void array_reverse_transpose_8x8(v128 *in, v128 *res) { … } int SIMD_FUNC(cdef_find_dir)(const uint16_t *img, int stride, int32_t *var, int coeff_shift) { … } // Work around compiler out of memory issues with Win32 builds. This issue has // been observed with Visual Studio 2017, 2019, and 2022 (version 17.10.3). #if defined(_MSC_VER) && defined(_M_IX86) #define CDEF_INLINE … #else #define CDEF_INLINE … #endif // sign(a-b) * min(abs(a-b), max(0, threshold - (abs(a-b) >> adjdamp))) CDEF_INLINE v256 constrain16(v256 a, v256 b, unsigned int threshold, unsigned int adjdamp) { … } SIMD_INLINE v256 get_max_primary(const int is_lowbd, v256 *tap, v256 max, v256 cdef_large_value_mask) { … } SIMD_INLINE v256 get_max_secondary(const int is_lowbd, v256 *tap, v256 max, v256 cdef_large_value_mask) { … } // MSVC takes far too much time optimizing these. // https://bugs.chromium.org/p/aomedia/issues/detail?id=3395 #if defined(_MSC_VER) && !defined(__clang__) #pragma optimize("", off) #endif CDEF_INLINE void filter_block_4x4(const int is_lowbd, void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int height, int enable_primary, int enable_secondary) { … } CDEF_INLINE void filter_block_8x8(const int is_lowbd, void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int height, int enable_primary, int enable_secondary) { … } #if defined(_MSC_VER) && !defined(__clang__) #pragma optimize("", on) #endif SIMD_INLINE void copy_block_4xh(const int is_lowbd, void *dest, int dstride, const uint16_t *in, int height) { … } SIMD_INLINE void copy_block_8xh(const int is_lowbd, void *dest, int dstride, const uint16_t *in, int height) { … } void SIMD_FUNC(cdef_filter_8_0)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_8_1)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_8_2)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_8_3)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_16_0)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_16_1)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_16_2)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } void SIMD_FUNC(cdef_filter_16_3)(void *dest, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int coeff_shift, int block_width, int block_height) { … } #if CONFIG_AV1_HIGHBITDEPTH void SIMD_FUNC(cdef_copy_rect8_16bit_to_16bit)(uint16_t *dst, int dstride, const uint16_t *src, int sstride, int width, int height) { int i, j; for (i = 0; i < height; i++) { for (j = 0; j < (width & ~0x7); j += 8) { v128 row = v128_load_unaligned(&src[i * sstride + j]); v128_store_unaligned(&dst[i * dstride + j], row); } for (; j < width; j++) { dst[i * dstride + j] = src[i * sstride + j]; } } } #endif // CONFIG_AV1_HIGHBITDEPTH #undef CDEF_INLINE #endif // AOM_AV1_COMMON_CDEF_BLOCK_SIMD_H_
Codebase Browser
chromium