1 /*****************************************************************************
2 * mc-c.c: arm motion compensation
3 *****************************************************************************
4 * Copyright (C) 2009-2011 x264 project
6 * Authors: David Conrad <lessen42@gmail.com>
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
22 * This program is also available under a commercial proprietary license.
23 * For more information, contact us at licensing@x264.com.
24 *****************************************************************************/
26 #include "common/common.h"
29 void x264_prefetch_ref_arm( uint8_t *, int, int );
30 void x264_prefetch_fenc_arm( uint8_t *, int, uint8_t *, int, int );
32 void *x264_memcpy_aligned_neon( void * dst, const void * src, size_t n );
33 void x264_memzero_aligned_neon( void *dst, int n );
35 void x264_pixel_avg_16x16_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
36 void x264_pixel_avg_16x8_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
37 void x264_pixel_avg_8x16_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
38 void x264_pixel_avg_8x8_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
39 void x264_pixel_avg_8x4_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
40 void x264_pixel_avg_4x8_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
41 void x264_pixel_avg_4x4_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
42 void x264_pixel_avg_4x2_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int, int );
44 void x264_pixel_avg2_w4_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int );
45 void x264_pixel_avg2_w8_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int );
46 void x264_pixel_avg2_w16_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int );
47 void x264_pixel_avg2_w20_neon( uint8_t *, int, uint8_t *, int, uint8_t *, int );
49 #define MC_WEIGHT(func)\
50 void x264_mc_weight_w20##func##_neon( uint8_t *, int, uint8_t *, int, const x264_weight_t *, int );\
51 void x264_mc_weight_w16##func##_neon( uint8_t *, int, uint8_t *, int, const x264_weight_t *, int );\
52 void x264_mc_weight_w8##func##_neon( uint8_t *, int, uint8_t *, int, const x264_weight_t *, int );\
53 void x264_mc_weight_w4##func##_neon( uint8_t *, int, uint8_t *, int, const x264_weight_t *, int );\
55 static void (* const x264_mc##func##_wtab_neon[6])( uint8_t *, int, uint8_t *, int, const x264_weight_t *, int ) =\
57 x264_mc_weight_w4##func##_neon,\
58 x264_mc_weight_w4##func##_neon,\
59 x264_mc_weight_w8##func##_neon,\
60 x264_mc_weight_w16##func##_neon,\
61 x264_mc_weight_w16##func##_neon,\
62 x264_mc_weight_w20##func##_neon,\
70 void x264_mc_copy_w4_neon( uint8_t *, int, uint8_t *, int, int );
71 void x264_mc_copy_w8_neon( uint8_t *, int, uint8_t *, int, int );
72 void x264_mc_copy_w16_neon( uint8_t *, int, uint8_t *, int, int );
73 void x264_mc_copy_w16_aligned_neon( uint8_t *, int, uint8_t *, int, int );
75 void x264_mc_chroma_neon( uint8_t *, int, uint8_t *, int, int, int, int, int );
76 void x264_frame_init_lowres_core_neon( uint8_t *, uint8_t *, uint8_t *, uint8_t *, uint8_t *, int, int, int, int);
78 void x264_hpel_filter_v_neon( uint8_t *, uint8_t *, int16_t *, int, int );
79 void x264_hpel_filter_c_neon( uint8_t *, int16_t *, int );
80 void x264_hpel_filter_h_neon( uint8_t *, uint8_t *, int );
83 static void x264_weight_cache_neon( x264_t *h, x264_weight_t *w )
85 if( w->i_scale == 1<<w->i_denom )
89 w->weightfn = x264_mc_offsetsub_wtab_neon;
90 w->cachea[0] = -w->i_offset;
94 w->weightfn = x264_mc_offsetadd_wtab_neon;
95 w->cachea[0] = w->i_offset;
98 else if( !w->i_denom )
99 w->weightfn = x264_mc_nodenom_wtab_neon;
101 w->weightfn = x264_mc_wtab_neon;
104 static void (* const x264_pixel_avg_wtab_neon[6])( uint8_t *, int, uint8_t *, int, uint8_t *, int ) =
107 x264_pixel_avg2_w4_neon,
108 x264_pixel_avg2_w8_neon,
109 x264_pixel_avg2_w16_neon, // no slower than w12, so no point in a separate function
110 x264_pixel_avg2_w16_neon,
111 x264_pixel_avg2_w20_neon,
114 static void (* const x264_mc_copy_wtab_neon[5])( uint8_t *, int, uint8_t *, int, int ) =
117 x264_mc_copy_w4_neon,
118 x264_mc_copy_w8_neon,
120 x264_mc_copy_w16_neon,
123 static const uint8_t hpel_ref0[16] = {0,1,1,1,0,1,1,1,2,3,3,3,0,1,1,1};
124 static const uint8_t hpel_ref1[16] = {0,0,0,0,2,2,3,2,2,2,3,2,2,2,3,2};
126 static void mc_luma_neon( uint8_t *dst, int i_dst_stride,
127 uint8_t *src[4], int i_src_stride,
129 int i_width, int i_height, const x264_weight_t *weight )
131 int qpel_idx = ((mvy&3)<<2) + (mvx&3);
132 int offset = (mvy>>2)*i_src_stride + (mvx>>2);
133 uint8_t *src1 = src[hpel_ref0[qpel_idx]] + offset;
134 if ( (mvy&3) == 3 ) // explict if() to force conditional add
135 src1 += i_src_stride;
137 if( qpel_idx & 5 ) /* qpel interpolation needed */
139 uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
140 x264_pixel_avg_wtab_neon[i_width>>2](
141 dst, i_dst_stride, src1, i_src_stride,
143 if( weight->weightfn )
144 weight->weightfn[i_width>>2]( dst, i_dst_stride, dst, i_dst_stride, weight, i_height );
146 else if( weight->weightfn )
147 weight->weightfn[i_width>>2]( dst, i_dst_stride, src1, i_src_stride, weight, i_height );
149 x264_mc_copy_wtab_neon[i_width>>2]( dst, i_dst_stride, src1, i_src_stride, i_height );
152 static uint8_t *get_ref_neon( uint8_t *dst, int *i_dst_stride,
153 uint8_t *src[4], int i_src_stride,
155 int i_width, int i_height, const x264_weight_t *weight )
157 int qpel_idx = ((mvy&3)<<2) + (mvx&3);
158 int offset = (mvy>>2)*i_src_stride + (mvx>>2);
159 uint8_t *src1 = src[hpel_ref0[qpel_idx]] + offset;
160 if ( (mvy&3) == 3 ) // explict if() to force conditional add
161 src1 += i_src_stride;
163 if( qpel_idx & 5 ) /* qpel interpolation needed */
165 uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
166 x264_pixel_avg_wtab_neon[i_width>>2](
167 dst, *i_dst_stride, src1, i_src_stride,
169 if( weight->weightfn )
170 weight->weightfn[i_width>>2]( dst, *i_dst_stride, dst, *i_dst_stride, weight, i_height );
173 else if( weight->weightfn )
175 weight->weightfn[i_width>>2]( dst, *i_dst_stride, src1, i_src_stride, weight, i_height );
180 *i_dst_stride = i_src_stride;
185 static void hpel_filter_neon( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src,
186 int stride, int width, int height, int16_t *buf )
188 int realign = (intptr_t)src & 15;
196 x264_hpel_filter_v_neon( dstv, src, buf+8, stride, width );
197 x264_hpel_filter_c_neon( dstc, buf+8, width );
198 x264_hpel_filter_h_neon( dsth, src, width );
205 #endif // !HIGH_BIT_DEPTH
207 void x264_mc_init_arm( int cpu, x264_mc_functions_t *pf )
209 if( !(cpu&X264_CPU_ARMV6) )
213 pf->prefetch_fenc = x264_prefetch_fenc_arm;
214 pf->prefetch_ref = x264_prefetch_ref_arm;
215 #endif // !HIGH_BIT_DEPTH
217 if( !(cpu&X264_CPU_NEON) )
221 pf->copy_16x16_unaligned = x264_mc_copy_w16_neon;
222 pf->copy[PIXEL_16x16] = x264_mc_copy_w16_aligned_neon;
223 pf->copy[PIXEL_8x8] = x264_mc_copy_w8_neon;
224 pf->copy[PIXEL_4x4] = x264_mc_copy_w4_neon;
226 pf->avg[PIXEL_16x16] = x264_pixel_avg_16x16_neon;
227 pf->avg[PIXEL_16x8] = x264_pixel_avg_16x8_neon;
228 pf->avg[PIXEL_8x16] = x264_pixel_avg_8x16_neon;
229 pf->avg[PIXEL_8x8] = x264_pixel_avg_8x8_neon;
230 pf->avg[PIXEL_8x4] = x264_pixel_avg_8x4_neon;
231 pf->avg[PIXEL_4x8] = x264_pixel_avg_4x8_neon;
232 pf->avg[PIXEL_4x4] = x264_pixel_avg_4x4_neon;
233 pf->avg[PIXEL_4x2] = x264_pixel_avg_4x2_neon;
235 pf->weight = x264_mc_wtab_neon;
236 pf->offsetadd = x264_mc_offsetadd_wtab_neon;
237 pf->offsetsub = x264_mc_offsetsub_wtab_neon;
238 pf->weight_cache = x264_weight_cache_neon;
240 // pf->mc_chroma = x264_mc_chroma_neon;
241 pf->mc_luma = mc_luma_neon;
242 pf->get_ref = get_ref_neon;
243 pf->hpel_filter = hpel_filter_neon;
244 pf->frame_init_lowres_core = x264_frame_init_lowres_core_neon;
245 #endif // !HIGH_BIT_DEPTH
247 // Apple's gcc stupidly cannot align stack variables, and ALIGNED_ARRAY can't work on structs
249 pf->memcpy_aligned = x264_memcpy_aligned_neon;
251 pf->memzero_aligned = x264_memzero_aligned_neon;