dsputil_init_neon.c
Go to the documentation of this file.
1 /*
2  * ARM NEON optimised DSP functions
3  * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
4  *
5  * This file is part of Libav.
6  *
7  * Libav is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * Libav is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with Libav; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 #include <stdint.h>
23 
24 #include "libavcodec/avcodec.h"
25 #include "libavcodec/dsputil.h"
26 #include "dsputil_arm.h"
27 
29 void ff_simple_idct_put_neon(uint8_t *dest, int line_size, DCTELEM *data);
30 void ff_simple_idct_add_neon(uint8_t *dest, int line_size, DCTELEM *data);
31 
33 void ff_clear_blocks_neon(DCTELEM *blocks);
34 
35 void ff_put_pixels16_neon(uint8_t *, const uint8_t *, int, int);
36 void ff_put_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int);
37 void ff_put_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int);
38 void ff_put_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int);
39 void ff_put_pixels8_neon(uint8_t *, const uint8_t *, int, int);
40 void ff_put_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int);
41 void ff_put_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int);
42 void ff_put_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int);
43 void ff_put_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
44 void ff_put_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
45 void ff_put_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
46 void ff_put_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
47 void ff_put_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
48 void ff_put_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
49 
50 void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int);
51 void ff_avg_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int);
52 void ff_avg_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int);
53 void ff_avg_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int);
54 void ff_avg_pixels8_neon(uint8_t *, const uint8_t *, int, int);
55 void ff_avg_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int);
56 void ff_avg_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int);
57 void ff_avg_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int);
58 void ff_avg_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
59 void ff_avg_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
60 void ff_avg_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
61 void ff_avg_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
62 void ff_avg_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
63 void ff_avg_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
64 
65 void ff_add_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
66 void ff_put_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
68 
85 
102 
119 
136 
137 void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
138 void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
139 void ff_put_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
140 
141 void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
142 void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
143 void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
144 
145 void ff_vector_fmul_window_neon(float *dst, const float *src0,
146  const float *src1, const float *win, int len);
147 void ff_butterflies_float_neon(float *v1, float *v2, int len);
148 float ff_scalarproduct_float_neon(const float *v1, const float *v2, int len);
149 void ff_vector_fmul_reverse_neon(float *dst, const float *src0,
150  const float *src1, int len);
151 void ff_vector_fmul_add_neon(float *dst, const float *src0, const float *src1,
152  const float *src2, int len);
153 
154 void ff_vector_clipf_neon(float *dst, const float *src, float min, float max,
155  int len);
156 void ff_vector_clip_int32_neon(int32_t *dst, const int32_t *src, int32_t min,
157  int32_t max, unsigned int len);
158 
159 void ff_vorbis_inverse_coupling_neon(float *mag, float *ang, int blocksize);
160 
161 int32_t ff_scalarproduct_int16_neon(const int16_t *v1, const int16_t *v2, int len);
162 int32_t ff_scalarproduct_and_madd_int16_neon(int16_t *v1, const int16_t *v2,
163  const int16_t *v3, int len, int mul);
164 
165 void ff_apply_window_int16_neon(int16_t *dst, const int16_t *src,
166  const int16_t *window, unsigned n);
167 
169 {
170  const int high_bit_depth = avctx->bits_per_raw_sample > 8;
171 
172  if (avctx->bits_per_raw_sample <= 8) {
173  if (avctx->idct_algo == FF_IDCT_AUTO ||
174  avctx->idct_algo == FF_IDCT_SIMPLENEON) {
179  }
180  }
181 
182  if (!high_bit_depth) {
185 
194 
203 
212 
221  }
222 
226 
227  if (CONFIG_H264_DECODER && !high_bit_depth) {
231 
235 
252 
269 
286 
303  }
304 
312 
315 
318 
320 }
void ff_avg_h264_qpel16_mc12_neon(uint8_t *, uint8_t *, int)
void(* vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len)
Definition: dsputil.h:358
void ff_put_h264_qpel16_mc03_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
op_pixels_func put_pixels_tab[4][4]
Halfpel motion compensation with rounding (a+b+1)>>1.
Definition: dsputil.h:259
void ff_avg_h264_qpel16_mc33_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int)
void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int)
void ff_put_h264_qpel8_mc22_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int)
void ff_vorbis_inverse_coupling_neon(float *mag, float *ang, int blocksize)
void ff_avg_h264_qpel8_mc13_neon(uint8_t *, uint8_t *, int)
void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
void ff_avg_h264_qpel8_mc20_neon(uint8_t *, uint8_t *, int)
void(* idct_add)(uint8_t *dest, int line_size, DCTELEM *block)
block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
Definition: dsputil.h:411
void(* vorbis_inverse_coupling)(float *mag, float *ang, int blocksize)
Definition: dsputil.h:352
void ff_avg_h264_qpel16_mc31_neon(uint8_t *, uint8_t *, int)
void ff_butterflies_float_neon(float *v1, float *v2, int len)
#define FF_PARTTRANS_IDCT_PERM
Definition: dsputil.h:431
void ff_avg_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_h264_qpel16_mc21_neon(uint8_t *, uint8_t *, int)
void(* apply_window_int16)(int16_t *output, const int16_t *input, const int16_t *window, unsigned int len)
Apply symmetric window in 16-bit fixed-point.
Definition: dsputil.h:470
int bits_per_raw_sample
Bits per sample/pixel of internal libavcodec pixel/sample format.
Definition: avcodec.h:2711
void ff_avg_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_clear_blocks_neon(DCTELEM *blocks)
float ff_scalarproduct_float_neon(const float *v1, const float *v2, int len)
void ff_put_h264_qpel8_mc33_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_vector_fmul_add_neon(float *dst, const float *src0, const float *src1, const float *src2, int len)
void ff_put_h264_qpel8_mc30_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc03_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel16_mc13_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
int32_t ff_scalarproduct_int16_neon(const int16_t *v1, const int16_t *v2, int len)
void ff_put_h264_qpel16_mc01_neon(uint8_t *, uint8_t *, int)
void ff_vector_clip_int32_neon(int32_t *dst, const int32_t *src, int32_t min, int32_t max, unsigned int len)
void ff_avg_pixels8_neon(uint8_t *, const uint8_t *, int, int)
void ff_avg_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int)
uint8_t
void ff_put_h264_qpel16_mc22_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_h264_qpel8_mc12_neon(uint8_t *, uint8_t *, int)
h264_chroma_mc_func avg_h264_chroma_pixels_tab[3]
Definition: dsputil.h:322
void ff_put_h264_qpel16_mc23_neon(uint8_t *, uint8_t *, int)
void ff_vector_fmul_reverse_neon(float *dst, const float *src0, const float *src1, int len)
void ff_avg_h264_qpel8_mc22_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc32_neon(uint8_t *, uint8_t *, int)
op_pixels_func avg_pixels_tab[4][4]
Halfpel motion compensation with rounding (a+b+1)>>1.
Definition: dsputil.h:271
const char data[16]
Definition: mxf.c:66
void ff_put_h264_qpel16_mc12_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc22_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel8_mc31_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc01_neon(uint8_t *, uint8_t *, int)
void ff_vector_clipf_neon(float *dst, const float *src, float min, float max, int len)
void ff_put_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void(* vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len)
Definition: dsputil.h:356
void(* add_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
Definition: dsputil.h:205
void ff_put_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int)
void ff_put_h264_qpel16_mc33_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc11_neon(uint8_t *, uint8_t *, int)
int32_t(* scalarproduct_int16)(const int16_t *v1, const int16_t *v2, int len)
Calculate scalar product of two vectors.
Definition: dsputil.h:450
void(* put_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
Definition: dsputil.h:203
void ff_avg_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_pixels16_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_h264_qpel8_mc32_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc13_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel8_mc21_neon(uint8_t *, uint8_t *, int)
void(* idct_put)(uint8_t *dest, int line_size, DCTELEM *block)
block -> idct -> clip to unsigned 8 bit -> dest.
Definition: dsputil.h:405
void ff_put_signed_pixels_clamped_neon(const DCTELEM *, uint8_t *, int)
void ff_put_h264_qpel16_mc02_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels_clamped_neon(const DCTELEM *, uint8_t *, int)
void ff_put_h264_qpel8_mc13_neon(uint8_t *, uint8_t *, int)
op_pixels_func put_no_rnd_pixels_tab[4][4]
Halfpel motion compensation with no rounding (a+b)>>1.
Definition: dsputil.h:283
void(* clear_block)(DCTELEM *block)
Definition: dsputil.h:218
void(* idct)(DCTELEM *block)
Definition: dsputil.h:398
h264_chroma_mc_func put_h264_chroma_pixels_tab[3]
h264 Chroma MC
Definition: dsputil.h:321
void ff_avg_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel8_mc00_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel16_mc30_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int)
void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int)
void ff_avg_h264_qpel8_mc30_neon(uint8_t *, uint8_t *, int)
static DCTELEM block[64]
Definition: dct-test.c:169
void ff_avg_h264_qpel16_mc21_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_avg_h264_qpel8_mc01_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel8_mc12_neon(uint8_t *, uint8_t *, int)
int idct_algo
IDCT algorithm, see FF_IDCT_* below.
Definition: avcodec.h:2661
int32_t
void ff_avg_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int)
int32_t ff_scalarproduct_and_madd_int16_neon(int16_t *v1, const int16_t *v2, const int16_t *v3, int len, int mul)
void ff_put_h264_qpel8_mc10_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel8_mc01_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel8_mc20_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_apply_window_int16_neon(int16_t *dst, const int16_t *src, const int16_t *window, unsigned n)
void ff_put_h264_qpel8_mc02_neon(uint8_t *, uint8_t *, int)
op_pixels_func avg_no_rnd_pixels_tab[4][4]
Halfpel motion compensation with no rounding (a+b)>>1.
Definition: dsputil.h:295
external API header
void ff_avg_h264_qpel16_mc02_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int)
void ff_avg_h264_qpel8_mc32_neon(uint8_t *, uint8_t *, int)
int idct_permutation_type
Definition: dsputil.h:426
void(* vector_clipf)(float *dst, const float *src, float min, float max, int len)
Definition: dsputil.h:360
main external API structure.
Definition: avcodec.h:1339
void ff_avg_h264_qpel8_mc00_neon(uint8_t *, uint8_t *, int)
void(* butterflies_float)(float *restrict v1, float *restrict v2, int len)
Calculate the sum and difference of two vectors of floats.
Definition: dsputil.h:374
void ff_simple_idct_add_neon(uint8_t *dest, int line_size, DCTELEM *data)
void ff_avg_h264_qpel8_mc03_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int)
void ff_simple_idct_put_neon(uint8_t *dest, int line_size, DCTELEM *data)
void ff_put_h264_qpel8_mc11_neon(uint8_t *, uint8_t *, int)
short DCTELEM
Definition: dsputil.h:39
void ff_clear_block_neon(DCTELEM *block)
void(* vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len)
Definition: dsputil.h:354
void ff_put_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
qpel_mc_func put_h264_qpel_pixels_tab[4][16]
Definition: dsputil.h:324
#define CONFIG_H264_DECODER
Definition: config.h:418
void(* clear_blocks)(DCTELEM *blocks)
Definition: dsputil.h:219
void ff_add_pixels_clamped_neon(const DCTELEM *, uint8_t *, int)
void(* vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min, int32_t max, unsigned int len)
Clip each element in an array of int32_t to a given minimum and maximum value.
Definition: dsputil.h:486
qpel_mc_func avg_h264_qpel_pixels_tab[4][16]
Definition: dsputil.h:325
void(* put_signed_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
Definition: dsputil.h:204
void ff_avg_h264_qpel8_mc11_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int)
void ff_avg_h264_qpel8_mc10_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels8_neon(uint8_t *, const uint8_t *, int, int)
void ff_avg_h264_qpel8_mc23_neon(uint8_t *, uint8_t *, int)
int32_t(* scalarproduct_and_madd_int16)(int16_t *v1, const int16_t *v2, const int16_t *v3, int len, int mul)
Calculate scalar product of v1 and v2, and v1[i] += v3[i] * mul.
Definition: dsputil.h:457
void ff_avg_h264_qpel16_mc23_neon(uint8_t *, uint8_t *, int)
void ff_put_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int)
DSP utils.
float(* scalarproduct_float)(const float *v1, const float *v2, int len)
Calculate the scalar product of two vectors of floats.
Definition: dsputil.h:367
void ff_put_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int)
int len
void ff_put_h264_qpel16_mc32_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel16_mc31_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel8_mc21_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel8_mc02_neon(uint8_t *, uint8_t *, int)
void ff_avg_h264_qpel16_mc30_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_h264_qpel16_mc11_neon(uint8_t *, uint8_t *, int)
void ff_simple_idct_neon(DCTELEM *data)
void ff_put_h264_qpel8_mc31_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel8_mc23_neon(uint8_t *, uint8_t *, int)
float min
void ff_avg_h264_qpel8_mc33_neon(uint8_t *, uint8_t *, int)
void ff_put_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int)
void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int)
void ff_avg_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int)
void ff_put_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int)
DSPContext.
Definition: dsputil.h:194
void ff_vector_fmul_window_neon(float *dst, const float *src0, const float *src1, const float *win, int len)
void ff_put_h264_qpel8_mc03_neon(uint8_t *, uint8_t *, int)
#define CONFIG_VORBIS_DECODER
Definition: config.h:590