/* * 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. */ #include "av1/common/x86/av1_txfm_sse2.h" #include "av1/encoder/av1_fwd_txfm1d_cfg.h" #include "av1/encoder/x86/av1_fwd_txfm_sse2.h" // TODO(linfengz): refine fdct4x8 and fadst4x8 optimization (if possible). static void fdct4x4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fdct8x4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fdct4x8_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fdct8x16_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } void av1_fdct8x32_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } void av1_fdct8x64_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fadst4x4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fadst4x8_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fadst8x4_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static void fadst8x16_new_sse2(const __m128i *input, __m128i *output, int8_t cos_bit) { … } static const transform_1d_sse2 col_txfm4x4_arr[TX_TYPES] = …; static const transform_1d_sse2 row_txfm4x4_arr[TX_TYPES] = …; static const transform_1d_sse2 col_txfm4x8_arr[TX_TYPES] = …; static const transform_1d_sse2 row_txfm8x4_arr[TX_TYPES] = …; static const transform_1d_sse2 col_txfm8x4_arr[TX_TYPES] = …; static const transform_1d_sse2 row_txfm4x8_arr[TX_TYPES] = …; static const transform_1d_sse2 col_txfm8x8_arr[TX_TYPES] = …; static const transform_1d_sse2 row_txfm8x8_arr[TX_TYPES] = …; static const transform_1d_sse2 col_txfm8x16_arr[TX_TYPES] = …; static const transform_1d_sse2 row_txfm8x16_arr[TX_TYPES] = …; static const transform_1d_sse2 row_txfm8x32_arr[TX_TYPES] = …; void av1_lowbd_fwd_txfm2d_4x4_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_4x8_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_4x16_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_8x4_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_8x8_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_8x16_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_8x32_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_16x4_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_16x8_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_16x16_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_16x32_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_32x8_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_32x16_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_32x32_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_64x16_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } void av1_lowbd_fwd_txfm2d_16x64_sse2(const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd) { … } // Include top-level function only for 32-bit x86, to support Valgrind. // For normal use, we require SSE4.1, so av1_lowbd_fwd_txfm_sse4_1 will be used // instead of this function. However, 32-bit Valgrind does not support SSE4.1, // so we include a fallback to SSE2 to improve performance #if AOM_ARCH_X86 static FwdTxfm2dFunc fwd_txfm2d_func_ls[TX_SIZES_ALL] = { av1_lowbd_fwd_txfm2d_4x4_sse2, // 4x4 transform av1_lowbd_fwd_txfm2d_8x8_sse2, // 8x8 transform av1_lowbd_fwd_txfm2d_16x16_sse2, // 16x16 transform av1_lowbd_fwd_txfm2d_32x32_sse2, // 32x32 transform NULL, // 64x64 transform av1_lowbd_fwd_txfm2d_4x8_sse2, // 4x8 transform av1_lowbd_fwd_txfm2d_8x4_sse2, // 8x4 transform av1_lowbd_fwd_txfm2d_8x16_sse2, // 8x16 transform av1_lowbd_fwd_txfm2d_16x8_sse2, // 16x8 transform av1_lowbd_fwd_txfm2d_16x32_sse2, // 16x32 transform av1_lowbd_fwd_txfm2d_32x16_sse2, // 32x16 transform NULL, // 32x64 transform NULL, // 64x32 transform av1_lowbd_fwd_txfm2d_4x16_sse2, // 4x16 transform av1_lowbd_fwd_txfm2d_16x4_sse2, // 16x4 transform av1_lowbd_fwd_txfm2d_8x32_sse2, // 8x32 transform av1_lowbd_fwd_txfm2d_32x8_sse2, // 32x8 transform av1_lowbd_fwd_txfm2d_16x64_sse2, // 16x64 transform av1_lowbd_fwd_txfm2d_64x16_sse2, // 64x16 transform }; void av1_lowbd_fwd_txfm_sse2(const int16_t *src_diff, tran_low_t *coeff, int diff_stride, TxfmParam *txfm_param) { FwdTxfm2dFunc fwd_txfm2d_func = fwd_txfm2d_func_ls[txfm_param->tx_size]; if ((fwd_txfm2d_func == NULL) || (txfm_param->lossless && txfm_param->tx_size == TX_4X4)) av1_lowbd_fwd_txfm_c(src_diff, coeff, diff_stride, txfm_param); else fwd_txfm2d_func(src_diff, coeff, diff_stride, txfm_param->tx_type, txfm_param->bd); } #endif // AOM_ARCH_X86
Codebase Browser
chromium