/* * 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. */ #include <assert.h> #include <immintrin.h> #include "config/aom_config.h" #include "config/av1_rtcd.h" #include "av1/common/av1_inv_txfm1d_cfg.h" #include "av1/common/idct.h" #include "av1/common/x86/av1_inv_txfm_ssse3.h" #include "av1/common/x86/highbd_txfm_utility_sse4.h" #include "aom_dsp/x86/txfm_common_avx2.h" // Note: // Total 32x4 registers to represent 32x32 block coefficients. // For high bit depth, each coefficient is 4-byte. // Each __m256i register holds 8 coefficients. // So each "row" we needs 4 register. Totally 32 rows // Register layout: // v0, v1, v2, v3, // v4, v5, v6, v7, // ... ... // v124, v125, v126, v127 static inline __m256i highbd_clamp_epi16_avx2(__m256i u, int bd) { … } static inline void round_shift_4x4_avx2(__m256i *in, int shift) { … } static inline void round_shift_8x8_avx2(__m256i *in, int shift) { … } static void highbd_clamp_epi32_avx2(__m256i *in, __m256i *out, const __m256i *clamp_lo, const __m256i *clamp_hi, int size) { … } static inline __m256i highbd_get_recon_16x8_avx2(const __m256i pred, __m256i res0, __m256i res1, const int bd) { … } static inline void highbd_write_buffer_16xn_avx2(__m256i *in, uint16_t *output, int stride, int flipud, int height, const int bd) { … } static inline __m256i highbd_get_recon_8x8_avx2(const __m256i pred, __m256i res, const int bd) { … } static inline void highbd_write_buffer_8xn_avx2(__m256i *in, uint16_t *output, int stride, int flipud, int height, const int bd) { … } static void neg_shift_avx2(const __m256i in0, const __m256i in1, __m256i *out0, __m256i *out1, const __m256i *clamp_lo, const __m256i *clamp_hi, int shift) { … } static void transpose_8x8_avx2(const __m256i *in, __m256i *out) { … } static void transpose_8x8_flip_avx2(const __m256i *in, __m256i *out) { … } static inline void load_buffer_32bit_input(const int32_t *in, int stride, __m256i *out, int out_size) { … } static inline __m256i half_btf_0_avx2(const __m256i *w0, const __m256i *n0, const __m256i *rounding, int bit) { … } static inline __m256i half_btf_avx2(const __m256i *w0, const __m256i *n0, const __m256i *w1, const __m256i *n1, const __m256i *rounding, int bit) { … } static void addsub_avx2(const __m256i in0, const __m256i in1, __m256i *out0, __m256i *out1, const __m256i *clamp_lo, const __m256i *clamp_hi) { … } static inline void idct32_stage4_avx2( __m256i *bf1, const __m256i *cospim8, const __m256i *cospi56, const __m256i *cospi8, const __m256i *cospim56, const __m256i *cospim40, const __m256i *cospi24, const __m256i *cospi40, const __m256i *cospim24, const __m256i *rounding, int bit) { … } static inline void idct32_stage5_avx2( __m256i *bf1, const __m256i *cospim16, const __m256i *cospi48, const __m256i *cospi16, const __m256i *cospim48, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rounding, int bit) { … } static inline void idct32_stage6_avx2( __m256i *bf1, const __m256i *cospim32, const __m256i *cospi32, const __m256i *cospim16, const __m256i *cospi48, const __m256i *cospi16, const __m256i *cospim48, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rounding, int bit) { … } static inline void idct32_stage7_avx2(__m256i *bf1, const __m256i *cospim32, const __m256i *cospi32, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rounding, int bit) { … } static inline void idct32_stage8_avx2(__m256i *bf1, const __m256i *cospim32, const __m256i *cospi32, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rounding, int bit) { … } static inline void idct32_stage9_avx2(__m256i *bf1, __m256i *out, const int do_cols, const int bd, const int out_shift, const __m256i *clamp_lo, const __m256i *clamp_hi) { … } static void idct32_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct32_low8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct32_low16_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct32_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct16_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct16_low8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct16_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void iadst16_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void iadst16_low8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void iadst16_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct8x8_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct8x8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void iadst8x8_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void iadst8x8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static inline void idct64_stage8_avx2( __m256i *u, const __m256i *cospim32, const __m256i *cospi32, const __m256i *cospim16, const __m256i *cospi48, const __m256i *cospi16, const __m256i *cospim48, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rnding, int bit) { … } static inline void idct64_stage9_avx2(__m256i *u, const __m256i *cospim32, const __m256i *cospi32, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rnding, int bit) { … } static inline void idct64_stage10_avx2(__m256i *u, const __m256i *cospim32, const __m256i *cospi32, const __m256i *clamp_lo, const __m256i *clamp_hi, const __m256i *rnding, int bit) { … } static inline void idct64_stage11_avx2(__m256i *u, __m256i *out, int do_cols, int bd, int out_shift, const __m256i *clamp_lo, const __m256i *clamp_hi) { … } static void idct64_low1_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct64_low8_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct64_low16_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } static void idct64_avx2(__m256i *in, __m256i *out, int bit, int do_cols, int bd, int out_shift) { … } transform_1d_avx2; static const transform_1d_avx2 highbd_txfm_all_1d_zeros_w8_arr[TX_SIZES][ITX_TYPES_1D][4] = …; static void highbd_inv_txfm2d_add_no_identity_avx2(const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, TX_SIZE tx_size, int eob, const int bd) { … } static void av1_highbd_inv_txfm2d_add_universe_avx2(const int32_t *input, uint8_t *output, int stride, TX_TYPE tx_type, TX_SIZE tx_size, int eob, const int bd) { … } void av1_highbd_inv_txfm_add_avx2(const tran_low_t *input, uint8_t *dest, int stride, const TxfmParam *txfm_param) { … }
Codebase Browser
chromium