/* * 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. */ #ifndef AOM_AOM_DSP_X86_LPF_COMMON_SSE2_H_ #define AOM_AOM_DSP_X86_LPF_COMMON_SSE2_H_ #include <emmintrin.h> // SSE2 #include "config/aom_config.h" #define mm_storelu(dst, v) … #define mm_storehu(dst, v) … static inline void highbd_transpose6x6_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3, __m128i *d4, __m128i *d5) { … } static inline void highbd_transpose4x8_8x4_low_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3) { … } static inline void highbd_transpose4x8_8x4_high_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } // here in and out pointers (x and d) should be different! we don't store their // values inside static inline void highbd_transpose4x8_8x4_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } static inline void highbd_transpose8x8_low_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3) { … } static inline void highbd_transpose8x8_high_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } // here in and out pointers (x and d) should be different! we don't store their // values inside static inline void highbd_transpose8x8_sse2( __m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } // here in and out pointers (x and d arrays) should be different! we don't store // their values inside static inline void highbd_transpose8x16_sse2( __m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } // Low bit depth functions static inline void transpose4x8_8x4_low_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3) { … } static inline void transpose4x8_8x4_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } static inline void transpose8x8_low_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3) { … } static inline void transpose8x8_sse2(__m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d0d1, __m128i *d2d3, __m128i *d4d5, __m128i *d6d7) { … } static inline void transpose16x8_8x16_sse2( __m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *x8, __m128i *x9, __m128i *x10, __m128i *x11, __m128i *x12, __m128i *x13, __m128i *x14, __m128i *x15, __m128i *d0, __m128i *d1, __m128i *d2, __m128i *d3, __m128i *d4, __m128i *d5, __m128i *d6, __m128i *d7) { … } static inline void transpose8x16_16x8_sse2( __m128i *x0, __m128i *x1, __m128i *x2, __m128i *x3, __m128i *x4, __m128i *x5, __m128i *x6, __m128i *x7, __m128i *d0d1, __m128i *d2d3, __m128i *d4d5, __m128i *d6d7, __m128i *d8d9, __m128i *d10d11, __m128i *d12d13, __m128i *d14d15) { … } static inline void transpose_16x8(unsigned char *in0, unsigned char *in1, int in_p, unsigned char *out, int out_p) { … } static inline void transpose_16x8_to_8x16(unsigned char *src, int in_p, unsigned char *dst, int out_p) { … } static inline void transpose_8xn(unsigned char *src[], int in_p, unsigned char *dst[], int out_p, int num_8x8_to_transpose) { … } #endif // AOM_AOM_DSP_X86_LPF_COMMON_SSE2_H_
Codebase Browser
chromium