chromium/third_party/libaom/source/libaom/aom_dsp/x86/synonyms_avx2.h

/*
 * Copyright (c) 2018, 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_AOM_DSP_X86_SYNONYMS_AVX2_H_
#define AOM_AOM_DSP_X86_SYNONYMS_AVX2_H_

#include <immintrin.h>

#include "config/aom_config.h"

#include "aom/aom_integer.h"

/**
 * Various reusable shorthands for x86 SIMD intrinsics.
 *
 * Intrinsics prefixed with xx_ operate on or return 128bit XMM registers.
 * Intrinsics prefixed with yy_ operate on or return 256bit YMM registers.
 */

// Loads and stores to do away with the tedium of casting the address
// to the right type.
static inline __m256i yy_load_256(const void *a) { … }

static inline __m256i yy_loadu_256(const void *a) { … }

static inline void yy_store_256(void *const a, const __m256i v) { … }

static inline void yy_storeu_256(void *const a, const __m256i v) { … }

// Fill an AVX register using an interleaved pair of values, ie. set the
// 16 channels to {a, b} repeated 8 times, using the same channel ordering
// as when a register is stored to / loaded from memory.
//
// This is useful for rearranging filter kernels for use with the _mm_madd_epi16
// instruction
static inline __m256i yy_set2_epi16(int16_t a, int16_t b) { … }

// Some compilers don't have _mm256_set_m128i defined in immintrin.h. We
// therefore define an equivalent function using a different intrinsic.
// ([ hi ], [ lo ]) -> [ hi ][ lo ]
static inline __m256i yy_set_m128i(__m128i hi, __m128i lo) { … }

// This behaves similarly to _mm256_set_epi64x(), but avoids undefined
// sanitizer warnings when loading values from unaligned buffers using
// `*(int64_t *)val`.
static inline __m256i yy_loadu_4x64(const void *e3, const void *e2,
                                    const void *e1, const void *e0) { … }

static inline __m256i yy_loadu2_128(const void *hi, const void *lo) { … }

static inline void yy_storeu2_128(void *hi, void *lo, const __m256i a) { … }

static inline __m256i yy_roundn_epu16(__m256i v_val_w, int bits) { … }
#endif  // AOM_AOM_DSP_X86_SYNONYMS_AVX2_H_