1sub vp8_common_forward_decls() {
2print <<EOF
3/*
4 * VP8
5 */
6
7struct blockd;
8struct macroblockd;
9struct loop_filter_info;
10
11/* Encoder forward decls */
12struct block;
13struct macroblock;
14struct variance_vtable;
15union int_mv;
16struct yv12_buffer_config;
17EOF
18}
19forward_decls qw/vp8_common_forward_decls/;
20
21#
22# system state
23#
24add_proto qw/void vp8_clear_system_state/, "";
25specialize qw/vp8_clear_system_state mmx/;
26$vp8_clear_system_state_mmx=vpx_reset_mmx_state;
27
28#
29# Dequant
30#
31add_proto qw/void vp8_dequantize_b/, "struct blockd*, short *dqc";
32specialize qw/vp8_dequantize_b mmx media neon/;
33$vp8_dequantize_b_media=vp8_dequantize_b_v6;
34
35add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char *output, int stride";
36specialize qw/vp8_dequant_idct_add mmx media neon dspr2/;
37$vp8_dequant_idct_add_media=vp8_dequant_idct_add_v6;
38$vp8_dequant_idct_add_dspr2=vp8_dequant_idct_add_dspr2;
39
40add_proto qw/void vp8_dequant_idct_add_y_block/, "short *q, short *dq, unsigned char *dst, int stride, char *eobs";
41specialize qw/vp8_dequant_idct_add_y_block mmx sse2 media neon dspr2/;
42$vp8_dequant_idct_add_y_block_media=vp8_dequant_idct_add_y_block_v6;
43$vp8_dequant_idct_add_y_block_dspr2=vp8_dequant_idct_add_y_block_dspr2;
44
45add_proto qw/void vp8_dequant_idct_add_uv_block/, "short *q, short *dq, unsigned char *dst_u, unsigned char *dst_v, int stride, char *eobs";
46specialize qw/vp8_dequant_idct_add_uv_block mmx sse2 media neon dspr2/;
47$vp8_dequant_idct_add_uv_block_media=vp8_dequant_idct_add_uv_block_v6;
48$vp8_dequant_idct_add_y_block_dspr2=vp8_dequant_idct_add_y_block_dspr2;
49
50#
51# Loopfilter
52#
53add_proto qw/void vp8_loop_filter_mbv/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
54specialize qw/vp8_loop_filter_mbv mmx sse2 media neon dspr2/;
55$vp8_loop_filter_mbv_media=vp8_loop_filter_mbv_armv6;
56$vp8_loop_filter_mbv_dspr2=vp8_loop_filter_mbv_dspr2;
57
58add_proto qw/void vp8_loop_filter_bv/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
59specialize qw/vp8_loop_filter_bv mmx sse2 media neon dspr2/;
60$vp8_loop_filter_bv_media=vp8_loop_filter_bv_armv6;
61$vp8_loop_filter_bv_dspr2=vp8_loop_filter_bv_dspr2;
62
63add_proto qw/void vp8_loop_filter_mbh/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
64specialize qw/vp8_loop_filter_mbh mmx sse2 media neon dspr2/;
65$vp8_loop_filter_mbh_media=vp8_loop_filter_mbh_armv6;
66$vp8_loop_filter_mbh_dspr2=vp8_loop_filter_mbh_dspr2;
67
68add_proto qw/void vp8_loop_filter_bh/, "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi";
69specialize qw/vp8_loop_filter_bh mmx sse2 media neon dspr2/;
70$vp8_loop_filter_bh_media=vp8_loop_filter_bh_armv6;
71$vp8_loop_filter_bh_dspr2=vp8_loop_filter_bh_dspr2;
72
73
74add_proto qw/void vp8_loop_filter_simple_mbv/, "unsigned char *y, int ystride, const unsigned char *blimit";
75specialize qw/vp8_loop_filter_simple_mbv mmx sse2 media neon/;
76$vp8_loop_filter_simple_mbv_c=vp8_loop_filter_simple_vertical_edge_c;
77$vp8_loop_filter_simple_mbv_mmx=vp8_loop_filter_simple_vertical_edge_mmx;
78$vp8_loop_filter_simple_mbv_sse2=vp8_loop_filter_simple_vertical_edge_sse2;
79$vp8_loop_filter_simple_mbv_media=vp8_loop_filter_simple_vertical_edge_armv6;
80$vp8_loop_filter_simple_mbv_neon=vp8_loop_filter_mbvs_neon;
81
82add_proto qw/void vp8_loop_filter_simple_mbh/, "unsigned char *y, int ystride, const unsigned char *blimit";
83specialize qw/vp8_loop_filter_simple_mbh mmx sse2 media neon/;
84$vp8_loop_filter_simple_mbh_c=vp8_loop_filter_simple_horizontal_edge_c;
85$vp8_loop_filter_simple_mbh_mmx=vp8_loop_filter_simple_horizontal_edge_mmx;
86$vp8_loop_filter_simple_mbh_sse2=vp8_loop_filter_simple_horizontal_edge_sse2;
87$vp8_loop_filter_simple_mbh_media=vp8_loop_filter_simple_horizontal_edge_armv6;
88$vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon;
89
90add_proto qw/void vp8_loop_filter_simple_bv/, "unsigned char *y, int ystride, const unsigned char *blimit";
91specialize qw/vp8_loop_filter_simple_bv mmx sse2 media neon/;
92$vp8_loop_filter_simple_bv_c=vp8_loop_filter_bvs_c;
93$vp8_loop_filter_simple_bv_mmx=vp8_loop_filter_bvs_mmx;
94$vp8_loop_filter_simple_bv_sse2=vp8_loop_filter_bvs_sse2;
95$vp8_loop_filter_simple_bv_media=vp8_loop_filter_bvs_armv6;
96$vp8_loop_filter_simple_bv_neon=vp8_loop_filter_bvs_neon;
97
98add_proto qw/void vp8_loop_filter_simple_bh/, "unsigned char *y, int ystride, const unsigned char *blimit";
99specialize qw/vp8_loop_filter_simple_bh mmx sse2 media neon/;
100$vp8_loop_filter_simple_bh_c=vp8_loop_filter_bhs_c;
101$vp8_loop_filter_simple_bh_mmx=vp8_loop_filter_bhs_mmx;
102$vp8_loop_filter_simple_bh_sse2=vp8_loop_filter_bhs_sse2;
103$vp8_loop_filter_simple_bh_media=vp8_loop_filter_bhs_armv6;
104$vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon;
105
106#
107# IDCT
108#
109#idct16
110add_proto qw/void vp8_short_idct4x4llm/, "short *input, unsigned char *pred, int pitch, unsigned char *dst, int dst_stride";
111specialize qw/vp8_short_idct4x4llm mmx media neon dspr2/;
112$vp8_short_idct4x4llm_media=vp8_short_idct4x4llm_v6_dual;
113$vp8_short_idct4x4llm_dspr2=vp8_short_idct4x4llm_dspr2;
114
115#iwalsh1
116add_proto qw/void vp8_short_inv_walsh4x4_1/, "short *input, short *output";
117specialize qw/vp8_short_inv_walsh4x4_1 dspr2/;
118$vp8_short_inv_walsh4x4_1_dspr2=vp8_short_inv_walsh4x4_1_dspr2;
119# no asm yet
120
121#iwalsh16
122add_proto qw/void vp8_short_inv_walsh4x4/, "short *input, short *output";
123specialize qw/vp8_short_inv_walsh4x4 mmx sse2 media neon dspr2/;
124$vp8_short_inv_walsh4x4_media=vp8_short_inv_walsh4x4_v6;
125$vp8_short_inv_walsh4x4_dspr2=vp8_short_inv_walsh4x4_dspr2;
126
127#idct1_scalar_add
128add_proto qw/void vp8_dc_only_idct_add/, "short input, unsigned char *pred, int pred_stride, unsigned char *dst, int dst_stride";
129specialize qw/vp8_dc_only_idct_add	mmx media neon dspr2/;
130$vp8_dc_only_idct_add_media=vp8_dc_only_idct_add_v6;
131$vp8_dc_only_idct_add_dspr2=vp8_dc_only_idct_add_dspr2;
132
133#
134# RECON
135#
136add_proto qw/void vp8_copy_mem16x16/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
137specialize qw/vp8_copy_mem16x16 mmx sse2 media neon dspr2/;
138$vp8_copy_mem16x16_media=vp8_copy_mem16x16_v6;
139$vp8_copy_mem16x16_dspr2=vp8_copy_mem16x16_dspr2;
140
141add_proto qw/void vp8_copy_mem8x8/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
142specialize qw/vp8_copy_mem8x8 mmx media neon dspr2/;
143$vp8_copy_mem8x8_media=vp8_copy_mem8x8_v6;
144$vp8_copy_mem8x8_dspr2=vp8_copy_mem8x8_dspr2;
145
146add_proto qw/void vp8_copy_mem8x4/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
147specialize qw/vp8_copy_mem8x4 mmx media neon dspr2/;
148$vp8_copy_mem8x4_media=vp8_copy_mem8x4_v6;
149$vp8_copy_mem8x4_dspr2=vp8_copy_mem8x4_dspr2;
150
151add_proto qw/void vp8_build_intra_predictors_mby_s/, "struct macroblockd *x, unsigned char * yabove_row, unsigned char * yleft, int left_stride, unsigned char * ypred_ptr, int y_stride";
152specialize qw/vp8_build_intra_predictors_mby_s sse2 ssse3/;
153#TODO: fix assembly for neon
154
155add_proto qw/void vp8_build_intra_predictors_mbuv_s/, "struct macroblockd *x, unsigned char * uabove_row, unsigned char * vabove_row,  unsigned char *uleft, unsigned char *vleft, int left_stride, unsigned char * upred_ptr, unsigned char * vpred_ptr, int pred_stride";
156specialize qw/vp8_build_intra_predictors_mbuv_s sse2 ssse3/;
157
158add_proto qw/void vp8_intra4x4_predict/, "unsigned char *Above, unsigned char *yleft, int left_stride, int b_mode, unsigned char *dst, int dst_stride, unsigned char top_left";
159specialize qw/vp8_intra4x4_predict media/;
160$vp8_intra4x4_predict_media=vp8_intra4x4_predict_armv6;
161
162#
163# Postproc
164#
165if (vpx_config("CONFIG_POSTPROC") eq "yes") {
166    add_proto qw/void vp8_mbpost_proc_down/, "unsigned char *dst, int pitch, int rows, int cols,int flimit";
167    specialize qw/vp8_mbpost_proc_down mmx sse2/;
168    $vp8_mbpost_proc_down_sse2=vp8_mbpost_proc_down_xmm;
169
170    add_proto qw/void vp8_mbpost_proc_across_ip/, "unsigned char *dst, int pitch, int rows, int cols,int flimit";
171    specialize qw/vp8_mbpost_proc_across_ip sse2/;
172    $vp8_mbpost_proc_across_ip_sse2=vp8_mbpost_proc_across_ip_xmm;
173
174    add_proto qw/void vp8_post_proc_down_and_across_mb_row/, "unsigned char *src, unsigned char *dst, int src_pitch, int dst_pitch, int cols, unsigned char *flimits, int size";
175    specialize qw/vp8_post_proc_down_and_across_mb_row sse2/;
176
177    add_proto qw/void vp8_plane_add_noise/, "unsigned char *s, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int w, unsigned int h, int pitch";
178    specialize qw/vp8_plane_add_noise mmx sse2/;
179    $vp8_plane_add_noise_sse2=vp8_plane_add_noise_wmt;
180
181    add_proto qw/void vp8_blend_mb_inner/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
182    # no asm yet
183
184    add_proto qw/void vp8_blend_mb_outer/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
185    # no asm yet
186
187    add_proto qw/void vp8_blend_b/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
188    # no asm yet
189
190    add_proto qw/void vp8_filter_by_weight16x16/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
191    specialize qw/vp8_filter_by_weight16x16 sse2/;
192
193    add_proto qw/void vp8_filter_by_weight8x8/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
194    specialize qw/vp8_filter_by_weight8x8 sse2/;
195
196    add_proto qw/void vp8_filter_by_weight4x4/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
197    # no asm yet
198}
199
200#
201# Subpixel
202#
203add_proto qw/void vp8_sixtap_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
204specialize qw/vp8_sixtap_predict16x16 mmx sse2 ssse3 media neon dspr2/;
205$vp8_sixtap_predict16x16_media=vp8_sixtap_predict16x16_armv6;
206$vp8_sixtap_predict16x16_dspr2=vp8_sixtap_predict16x16_dspr2;
207
208add_proto qw/void vp8_sixtap_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
209specialize qw/vp8_sixtap_predict8x8 mmx sse2 ssse3 media neon dspr2/;
210$vp8_sixtap_predict8x8_media=vp8_sixtap_predict8x8_armv6;
211$vp8_sixtap_predict8x8_dspr2=vp8_sixtap_predict8x8_dspr2;
212
213add_proto qw/void vp8_sixtap_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
214specialize qw/vp8_sixtap_predict8x4 mmx sse2 ssse3 media neon dspr2/;
215$vp8_sixtap_predict8x4_media=vp8_sixtap_predict8x4_armv6;
216$vp8_sixtap_predict8x4_dspr2=vp8_sixtap_predict8x4_dspr2;
217
218add_proto qw/void vp8_sixtap_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
219specialize qw/vp8_sixtap_predict4x4 mmx ssse3 media neon dspr2/;
220$vp8_sixtap_predict4x4_media=vp8_sixtap_predict4x4_armv6;
221$vp8_sixtap_predict4x4_dspr2=vp8_sixtap_predict4x4_dspr2;
222
223add_proto qw/void vp8_bilinear_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
224specialize qw/vp8_bilinear_predict16x16 mmx sse2 ssse3 media neon/;
225$vp8_bilinear_predict16x16_media=vp8_bilinear_predict16x16_armv6;
226
227add_proto qw/void vp8_bilinear_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
228specialize qw/vp8_bilinear_predict8x8 mmx sse2 ssse3 media neon/;
229$vp8_bilinear_predict8x8_media=vp8_bilinear_predict8x8_armv6;
230
231add_proto qw/void vp8_bilinear_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
232specialize qw/vp8_bilinear_predict8x4 mmx media neon/;
233$vp8_bilinear_predict8x4_media=vp8_bilinear_predict8x4_armv6;
234
235add_proto qw/void vp8_bilinear_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
236specialize qw/vp8_bilinear_predict4x4 mmx media neon/;
237$vp8_bilinear_predict4x4_media=vp8_bilinear_predict4x4_armv6;
238
239#
240# Whole-pixel Variance
241#
242add_proto qw/unsigned int vp8_variance4x4/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
243specialize qw/vp8_variance4x4 mmx sse2/;
244$vp8_variance4x4_sse2=vp8_variance4x4_wmt;
245
246add_proto qw/unsigned int vp8_variance8x8/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
247specialize qw/vp8_variance8x8 mmx sse2 media neon/;
248$vp8_variance8x8_sse2=vp8_variance8x8_wmt;
249$vp8_variance8x8_media=vp8_variance8x8_armv6;
250
251add_proto qw/unsigned int vp8_variance8x16/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
252specialize qw/vp8_variance8x16 mmx sse2 neon/;
253$vp8_variance8x16_sse2=vp8_variance8x16_wmt;
254
255add_proto qw/unsigned int vp8_variance16x8/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
256specialize qw/vp8_variance16x8 mmx sse2 neon/;
257$vp8_variance16x8_sse2=vp8_variance16x8_wmt;
258
259add_proto qw/unsigned int vp8_variance16x16/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
260specialize qw/vp8_variance16x16 mmx sse2 media neon/;
261$vp8_variance16x16_sse2=vp8_variance16x16_wmt;
262$vp8_variance16x16_media=vp8_variance16x16_armv6;
263
264#
265# Sub-pixel Variance
266#
267add_proto qw/unsigned int vp8_sub_pixel_variance4x4/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
268specialize qw/vp8_sub_pixel_variance4x4 mmx sse2/;
269$vp8_sub_pixel_variance4x4_sse2=vp8_sub_pixel_variance4x4_wmt;
270
271add_proto qw/unsigned int vp8_sub_pixel_variance8x8/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
272specialize qw/vp8_sub_pixel_variance8x8 mmx sse2 media neon/;
273$vp8_sub_pixel_variance8x8_sse2=vp8_sub_pixel_variance8x8_wmt;
274$vp8_sub_pixel_variance8x8_media=vp8_sub_pixel_variance8x8_armv6;
275
276add_proto qw/unsigned int vp8_sub_pixel_variance8x16/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
277specialize qw/vp8_sub_pixel_variance8x16 mmx sse2/;
278$vp8_sub_pixel_variance8x16_sse2=vp8_sub_pixel_variance8x16_wmt;
279
280add_proto qw/unsigned int vp8_sub_pixel_variance16x8/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
281specialize qw/vp8_sub_pixel_variance16x8 mmx sse2 ssse3/;
282$vp8_sub_pixel_variance16x8_sse2=vp8_sub_pixel_variance16x8_wmt;
283
284add_proto qw/unsigned int vp8_sub_pixel_variance16x16/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
285specialize qw/vp8_sub_pixel_variance16x16 mmx sse2 ssse3 media neon/;
286$vp8_sub_pixel_variance16x16_sse2=vp8_sub_pixel_variance16x16_wmt;
287$vp8_sub_pixel_variance16x16_media=vp8_sub_pixel_variance16x16_armv6;
288
289add_proto qw/unsigned int vp8_variance_halfpixvar16x16_h/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
290specialize qw/vp8_variance_halfpixvar16x16_h mmx sse2 media neon/;
291$vp8_variance_halfpixvar16x16_h_sse2=vp8_variance_halfpixvar16x16_h_wmt;
292$vp8_variance_halfpixvar16x16_h_media=vp8_variance_halfpixvar16x16_h_armv6;
293
294add_proto qw/unsigned int vp8_variance_halfpixvar16x16_v/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
295specialize qw/vp8_variance_halfpixvar16x16_v mmx sse2 media neon/;
296$vp8_variance_halfpixvar16x16_v_sse2=vp8_variance_halfpixvar16x16_v_wmt;
297$vp8_variance_halfpixvar16x16_v_media=vp8_variance_halfpixvar16x16_v_armv6;
298
299add_proto qw/unsigned int vp8_variance_halfpixvar16x16_hv/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
300specialize qw/vp8_variance_halfpixvar16x16_hv mmx sse2 media neon/;
301$vp8_variance_halfpixvar16x16_hv_sse2=vp8_variance_halfpixvar16x16_hv_wmt;
302$vp8_variance_halfpixvar16x16_hv_media=vp8_variance_halfpixvar16x16_hv_armv6;
303
304#
305# Single block SAD
306#
307add_proto qw/unsigned int vp8_sad4x4/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
308specialize qw/vp8_sad4x4 mmx sse2 neon/;
309$vp8_sad4x4_sse2=vp8_sad4x4_wmt;
310
311add_proto qw/unsigned int vp8_sad8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
312specialize qw/vp8_sad8x8 mmx sse2 neon/;
313$vp8_sad8x8_sse2=vp8_sad8x8_wmt;
314
315add_proto qw/unsigned int vp8_sad8x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
316specialize qw/vp8_sad8x16 mmx sse2 neon/;
317$vp8_sad8x16_sse2=vp8_sad8x16_wmt;
318
319add_proto qw/unsigned int vp8_sad16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
320specialize qw/vp8_sad16x8 mmx sse2 neon/;
321$vp8_sad16x8_sse2=vp8_sad16x8_wmt;
322
323add_proto qw/unsigned int vp8_sad16x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
324specialize qw/vp8_sad16x16 mmx sse2 sse3 media neon/;
325$vp8_sad16x16_sse2=vp8_sad16x16_wmt;
326$vp8_sad16x16_media=vp8_sad16x16_armv6;
327
328#
329# Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
330#
331add_proto qw/void vp8_sad4x4x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
332specialize qw/vp8_sad4x4x3 sse3/;
333
334add_proto qw/void vp8_sad8x8x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
335specialize qw/vp8_sad8x8x3 sse3/;
336
337add_proto qw/void vp8_sad8x16x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
338specialize qw/vp8_sad8x16x3 sse3/;
339
340add_proto qw/void vp8_sad16x8x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
341specialize qw/vp8_sad16x8x3 sse3 ssse3/;
342
343add_proto qw/void vp8_sad16x16x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
344specialize qw/vp8_sad16x16x3 sse3 ssse3/;
345
346# Note the only difference in the following prototypes is that they return into
347# an array of short
348add_proto qw/void vp8_sad4x4x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
349specialize qw/vp8_sad4x4x8 sse4_1/;
350$vp8_sad4x4x8_sse4_1=vp8_sad4x4x8_sse4;
351
352add_proto qw/void vp8_sad8x8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
353specialize qw/vp8_sad8x8x8 sse4_1/;
354$vp8_sad8x8x8_sse4_1=vp8_sad8x8x8_sse4;
355
356add_proto qw/void vp8_sad8x16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
357specialize qw/vp8_sad8x16x8 sse4_1/;
358$vp8_sad8x16x8_sse4_1=vp8_sad8x16x8_sse4;
359
360add_proto qw/void vp8_sad16x8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
361specialize qw/vp8_sad16x8x8 sse4_1/;
362$vp8_sad16x8x8_sse4_1=vp8_sad16x8x8_sse4;
363
364add_proto qw/void vp8_sad16x16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
365specialize qw/vp8_sad16x16x8 sse4_1/;
366$vp8_sad16x16x8_sse4_1=vp8_sad16x16x8_sse4;
367
368#
369# Multi-block SAD, comparing a reference to N independent blocks
370#
371add_proto qw/void vp8_sad4x4x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array";
372specialize qw/vp8_sad4x4x4d sse3/;
373
374add_proto qw/void vp8_sad8x8x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array";
375specialize qw/vp8_sad8x8x4d sse3/;
376
377add_proto qw/void vp8_sad8x16x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array";
378specialize qw/vp8_sad8x16x4d sse3/;
379
380add_proto qw/void vp8_sad16x8x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array";
381specialize qw/vp8_sad16x8x4d sse3/;
382
383add_proto qw/void vp8_sad16x16x4d/, "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array";
384specialize qw/vp8_sad16x16x4d sse3/;
385
386#
387# Encoder functions below this point.
388#
389if (vpx_config("CONFIG_VP8_ENCODER") eq "yes") {
390
391#
392# Sum of squares (vector)
393#
394add_proto qw/unsigned int vp8_get_mb_ss/, "const short *";
395specialize qw/vp8_get_mb_ss mmx sse2/;
396
397#
398# SSE (Sum Squared Error)
399#
400add_proto qw/unsigned int vp8_sub_pixel_mse16x16/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
401specialize qw/vp8_sub_pixel_mse16x16 mmx sse2/;
402$vp8_sub_pixel_mse16x16_sse2=vp8_sub_pixel_mse16x16_wmt;
403
404add_proto qw/unsigned int vp8_mse16x16/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
405specialize qw/vp8_mse16x16 mmx sse2 media neon/;
406$vp8_mse16x16_sse2=vp8_mse16x16_wmt;
407$vp8_mse16x16_media=vp8_mse16x16_armv6;
408
409add_proto qw/unsigned int vp8_get4x4sse_cs/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride";
410specialize qw/vp8_get4x4sse_cs mmx neon/;
411
412#
413# Block copy
414#
415if ($opts{arch} =~ /x86/) {
416    add_proto qw/void vp8_copy32xn/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int n";
417    specialize qw/vp8_copy32xn sse2 sse3/;
418}
419
420#
421# Structured Similarity (SSIM)
422#
423if (vpx_config("CONFIG_INTERNAL_STATS") eq "yes") {
424    $opts{arch} eq "x86_64" and $sse2_on_x86_64 = "sse2";
425
426    add_proto qw/void vp8_ssim_parms_8x8/, "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr";
427    specialize qw/vp8_ssim_parms_8x8/, "$sse2_on_x86_64";
428
429    add_proto qw/void vp8_ssim_parms_16x16/, "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr";
430    specialize qw/vp8_ssim_parms_16x16/, "$sse2_on_x86_64";
431}
432
433#
434# Forward DCT
435#
436add_proto qw/void vp8_short_fdct4x4/, "short *input, short *output, int pitch";
437specialize qw/vp8_short_fdct4x4 mmx sse2 media neon/;
438$vp8_short_fdct4x4_media=vp8_short_fdct4x4_armv6;
439
440add_proto qw/void vp8_short_fdct8x4/, "short *input, short *output, int pitch";
441specialize qw/vp8_short_fdct8x4 mmx sse2 media neon/;
442$vp8_short_fdct8x4_media=vp8_short_fdct8x4_armv6;
443
444add_proto qw/void vp8_short_walsh4x4/, "short *input, short *output, int pitch";
445specialize qw/vp8_short_walsh4x4 sse2 media neon/;
446$vp8_short_walsh4x4_media=vp8_short_walsh4x4_armv6;
447
448#
449# Quantizer
450#
451add_proto qw/void vp8_regular_quantize_b/, "struct block *, struct blockd *";
452specialize qw/vp8_regular_quantize_b sse2/;
453# TODO(johann) Update sse4 implementation and re-enable
454#$vp8_regular_quantize_b_sse4_1=vp8_regular_quantize_b_sse4;
455
456add_proto qw/void vp8_fast_quantize_b/, "struct block *, struct blockd *";
457specialize qw/vp8_fast_quantize_b sse2 ssse3 media neon/;
458$vp8_fast_quantize_b_media=vp8_fast_quantize_b_armv6;
459
460add_proto qw/void vp8_regular_quantize_b_pair/, "struct block *b1, struct block *b2, struct blockd *d1, struct blockd *d2";
461# no asm yet
462
463add_proto qw/void vp8_fast_quantize_b_pair/, "struct block *b1, struct block *b2, struct blockd *d1, struct blockd *d2";
464specialize qw/vp8_fast_quantize_b_pair neon/;
465
466add_proto qw/void vp8_quantize_mb/, "struct macroblock *";
467specialize qw/vp8_quantize_mb neon/;
468
469add_proto qw/void vp8_quantize_mby/, "struct macroblock *";
470specialize qw/vp8_quantize_mby neon/;
471
472add_proto qw/void vp8_quantize_mbuv/, "struct macroblock *";
473specialize qw/vp8_quantize_mbuv neon/;
474
475#
476# Block subtraction
477#
478add_proto qw/int vp8_block_error/, "short *coeff, short *dqcoeff";
479specialize qw/vp8_block_error mmx sse2/;
480$vp8_block_error_sse2=vp8_block_error_xmm;
481
482add_proto qw/int vp8_mbblock_error/, "struct macroblock *mb, int dc";
483specialize qw/vp8_mbblock_error mmx sse2/;
484$vp8_mbblock_error_sse2=vp8_mbblock_error_xmm;
485
486add_proto qw/int vp8_mbuverror/, "struct macroblock *mb";
487specialize qw/vp8_mbuverror mmx sse2/;
488$vp8_mbuverror_sse2=vp8_mbuverror_xmm;
489
490add_proto qw/void vp8_subtract_b/, "struct block *be, struct blockd *bd, int pitch";
491specialize qw/vp8_subtract_b mmx sse2 media neon/;
492$vp8_subtract_b_media=vp8_subtract_b_armv6;
493
494add_proto qw/void vp8_subtract_mby/, "short *diff, unsigned char *src, int src_stride, unsigned char *pred, int pred_stride";
495specialize qw/vp8_subtract_mby mmx sse2 media neon/;
496$vp8_subtract_mby_media=vp8_subtract_mby_armv6;
497
498add_proto qw/void vp8_subtract_mbuv/, "short *diff, unsigned char *usrc, unsigned char *vsrc, int src_stride, unsigned char *upred, unsigned char *vpred, int pred_stride";
499specialize qw/vp8_subtract_mbuv mmx sse2 media neon/;
500$vp8_subtract_mbuv_media=vp8_subtract_mbuv_armv6;
501
502#
503# Motion search
504#
505add_proto qw/int vp8_full_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
506specialize qw/vp8_full_search_sad sse3 sse4_1/;
507$vp8_full_search_sad_sse3=vp8_full_search_sadx3;
508$vp8_full_search_sad_sse4_1=vp8_full_search_sadx8;
509
510add_proto qw/int vp8_refining_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
511specialize qw/vp8_refining_search_sad sse3/;
512$vp8_refining_search_sad_sse3=vp8_refining_search_sadx4;
513
514add_proto qw/int vp8_diamond_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, union int_mv *best_mv, int search_param, int sad_per_bit, int *num00, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
515$vp8_diamond_search_sad_sse3=vp8_diamond_search_sadx4;
516
517#
518# Alt-ref Noise Reduction (ARNR)
519#
520if (vpx_config("CONFIG_REALTIME_ONLY") ne "yes") {
521    add_proto qw/void vp8_temporal_filter_apply/, "unsigned char *frame1, unsigned int stride, unsigned char *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, unsigned short *count";
522    specialize qw/vp8_temporal_filter_apply sse2/;
523}
524
525#
526# Pick Loopfilter
527#
528add_proto qw/void vp8_yv12_copy_partial_frame/, "struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc";
529specialize qw/vp8_yv12_copy_partial_frame neon/;
530
531#
532# Denoiser filter
533#
534if (vpx_config("CONFIG_TEMPORAL_DENOISING") eq "yes") {
535    add_proto qw/int vp8_denoiser_filter/, "struct yv12_buffer_config* mc_running_avg, struct yv12_buffer_config* running_avg, struct macroblock* signal, unsigned int motion_magnitude2, int y_offset, int uv_offset";
536    specialize qw/vp8_denoiser_filter sse2 neon/;
537}
538
539# End of encoder only functions
540}
5411;
542