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$vp8_clear_system_state_mmx=vpx_reset_mmx_state;
27
28#
29# Dequant
30#
31add_proto qw/void vp8_dequantize_b/, "struct blockd*, short *dqc";
32specialize qw/vp8_dequantize_b mmx media neon msa/;
33$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 msa/;
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 msa/;
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 msa/;
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 msa/;
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 msa/;
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 msa/;
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 msa/;
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 msa/;
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$vp8_loop_filter_simple_mbv_neon=vp8_loop_filter_mbvs_neon;
81$vp8_loop_filter_simple_mbv_msa=vp8_loop_filter_simple_vertical_edge_msa;
82
83add_proto qw/void vp8_loop_filter_simple_mbh/, "unsigned char *y, int ystride, const unsigned char *blimit";
84specialize qw/vp8_loop_filter_simple_mbh mmx sse2 media neon msa/;
85$vp8_loop_filter_simple_mbh_c=vp8_loop_filter_simple_horizontal_edge_c;
86$vp8_loop_filter_simple_mbh_mmx=vp8_loop_filter_simple_horizontal_edge_mmx;
87$vp8_loop_filter_simple_mbh_sse2=vp8_loop_filter_simple_horizontal_edge_sse2;
88$vp8_loop_filter_simple_mbh_media=vp8_loop_filter_simple_horizontal_edge_armv6;
89$vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon;
90$vp8_loop_filter_simple_mbh_msa=vp8_loop_filter_simple_horizontal_edge_msa;
91
92add_proto qw/void vp8_loop_filter_simple_bv/, "unsigned char *y, int ystride, const unsigned char *blimit";
93specialize qw/vp8_loop_filter_simple_bv mmx sse2 media neon msa/;
94$vp8_loop_filter_simple_bv_c=vp8_loop_filter_bvs_c;
95$vp8_loop_filter_simple_bv_mmx=vp8_loop_filter_bvs_mmx;
96$vp8_loop_filter_simple_bv_sse2=vp8_loop_filter_bvs_sse2;
97$vp8_loop_filter_simple_bv_media=vp8_loop_filter_bvs_armv6;
98$vp8_loop_filter_simple_bv_neon=vp8_loop_filter_bvs_neon;
99$vp8_loop_filter_simple_bv_msa=vp8_loop_filter_bvs_msa;
100
101add_proto qw/void vp8_loop_filter_simple_bh/, "unsigned char *y, int ystride, const unsigned char *blimit";
102specialize qw/vp8_loop_filter_simple_bh mmx sse2 media neon msa/;
103$vp8_loop_filter_simple_bh_c=vp8_loop_filter_bhs_c;
104$vp8_loop_filter_simple_bh_mmx=vp8_loop_filter_bhs_mmx;
105$vp8_loop_filter_simple_bh_sse2=vp8_loop_filter_bhs_sse2;
106$vp8_loop_filter_simple_bh_media=vp8_loop_filter_bhs_armv6;
107$vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon;
108$vp8_loop_filter_simple_bh_msa=vp8_loop_filter_bhs_msa;
109
110#
111# IDCT
112#
113#idct16
114add_proto qw/void vp8_short_idct4x4llm/, "short *input, unsigned char *pred, int pitch, unsigned char *dst, int dst_stride";
115specialize qw/vp8_short_idct4x4llm mmx media neon dspr2 msa/;
116$vp8_short_idct4x4llm_media=vp8_short_idct4x4llm_v6_dual;
117$vp8_short_idct4x4llm_dspr2=vp8_short_idct4x4llm_dspr2;
118
119#iwalsh1
120add_proto qw/void vp8_short_inv_walsh4x4_1/, "short *input, short *output";
121specialize qw/vp8_short_inv_walsh4x4_1 dspr2/;
122$vp8_short_inv_walsh4x4_1_dspr2=vp8_short_inv_walsh4x4_1_dspr2;
123# no asm yet
124
125#iwalsh16
126add_proto qw/void vp8_short_inv_walsh4x4/, "short *input, short *output";
127specialize qw/vp8_short_inv_walsh4x4 mmx sse2 media neon dspr2 msa/;
128$vp8_short_inv_walsh4x4_media=vp8_short_inv_walsh4x4_v6;
129$vp8_short_inv_walsh4x4_dspr2=vp8_short_inv_walsh4x4_dspr2;
130
131#idct1_scalar_add
132add_proto qw/void vp8_dc_only_idct_add/, "short input, unsigned char *pred, int pred_stride, unsigned char *dst, int dst_stride";
133specialize qw/vp8_dc_only_idct_add	mmx media neon dspr2 msa/;
134$vp8_dc_only_idct_add_media=vp8_dc_only_idct_add_v6;
135$vp8_dc_only_idct_add_dspr2=vp8_dc_only_idct_add_dspr2;
136
137#
138# RECON
139#
140add_proto qw/void vp8_copy_mem16x16/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
141specialize qw/vp8_copy_mem16x16 mmx sse2 media neon dspr2 msa/;
142$vp8_copy_mem16x16_media=vp8_copy_mem16x16_v6;
143$vp8_copy_mem16x16_dspr2=vp8_copy_mem16x16_dspr2;
144
145add_proto qw/void vp8_copy_mem8x8/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
146specialize qw/vp8_copy_mem8x8 mmx media neon dspr2 msa/;
147$vp8_copy_mem8x8_media=vp8_copy_mem8x8_v6;
148$vp8_copy_mem8x8_dspr2=vp8_copy_mem8x8_dspr2;
149
150add_proto qw/void vp8_copy_mem8x4/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
151specialize qw/vp8_copy_mem8x4 mmx media neon dspr2 msa/;
152$vp8_copy_mem8x4_media=vp8_copy_mem8x4_v6;
153$vp8_copy_mem8x4_dspr2=vp8_copy_mem8x4_dspr2;
154
155#
156# Postproc
157#
158if (vpx_config("CONFIG_POSTPROC") eq "yes") {
159    add_proto qw/void vp8_mbpost_proc_down/, "unsigned char *dst, int pitch, int rows, int cols,int flimit";
160    specialize qw/vp8_mbpost_proc_down mmx sse2 msa/;
161    $vp8_mbpost_proc_down_sse2=vp8_mbpost_proc_down_xmm;
162
163    add_proto qw/void vp8_mbpost_proc_across_ip/, "unsigned char *dst, int pitch, int rows, int cols,int flimit";
164    specialize qw/vp8_mbpost_proc_across_ip sse2 msa/;
165    $vp8_mbpost_proc_across_ip_sse2=vp8_mbpost_proc_across_ip_xmm;
166
167    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";
168    specialize qw/vp8_post_proc_down_and_across_mb_row sse2 msa/;
169
170    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";
171    specialize qw/vp8_plane_add_noise mmx sse2 msa/;
172    $vp8_plane_add_noise_sse2=vp8_plane_add_noise_wmt;
173
174    add_proto qw/void vp8_blend_mb_inner/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
175    # no asm yet
176
177    add_proto qw/void vp8_blend_mb_outer/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
178    # no asm yet
179
180    add_proto qw/void vp8_blend_b/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
181    # no asm yet
182
183    add_proto qw/void vp8_filter_by_weight16x16/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
184    specialize qw/vp8_filter_by_weight16x16 sse2 msa/;
185
186    add_proto qw/void vp8_filter_by_weight8x8/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
187    specialize qw/vp8_filter_by_weight8x8 sse2 msa/;
188
189    add_proto qw/void vp8_filter_by_weight4x4/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
190    # no asm yet
191}
192
193#
194# Subpixel
195#
196add_proto qw/void vp8_sixtap_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
197specialize qw/vp8_sixtap_predict16x16 mmx sse2 ssse3 media neon dspr2 msa/;
198$vp8_sixtap_predict16x16_media=vp8_sixtap_predict16x16_armv6;
199$vp8_sixtap_predict16x16_dspr2=vp8_sixtap_predict16x16_dspr2;
200
201add_proto qw/void vp8_sixtap_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
202specialize qw/vp8_sixtap_predict8x8 mmx sse2 ssse3 media neon dspr2 msa/;
203$vp8_sixtap_predict8x8_media=vp8_sixtap_predict8x8_armv6;
204$vp8_sixtap_predict8x8_dspr2=vp8_sixtap_predict8x8_dspr2;
205
206add_proto qw/void vp8_sixtap_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
207specialize qw/vp8_sixtap_predict8x4 mmx sse2 ssse3 media neon dspr2 msa/;
208$vp8_sixtap_predict8x4_media=vp8_sixtap_predict8x4_armv6;
209$vp8_sixtap_predict8x4_dspr2=vp8_sixtap_predict8x4_dspr2;
210
211add_proto qw/void vp8_sixtap_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
212#TODO(johannkoenig): fix the neon version https://code.google.com/p/webm/issues/detail?id=817
213specialize qw/vp8_sixtap_predict4x4 mmx ssse3 media dspr2 msa/;
214$vp8_sixtap_predict4x4_media=vp8_sixtap_predict4x4_armv6;
215$vp8_sixtap_predict4x4_dspr2=vp8_sixtap_predict4x4_dspr2;
216
217add_proto qw/void vp8_bilinear_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
218specialize qw/vp8_bilinear_predict16x16 mmx sse2 ssse3 media neon msa/;
219$vp8_bilinear_predict16x16_media=vp8_bilinear_predict16x16_armv6;
220
221add_proto qw/void vp8_bilinear_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
222specialize qw/vp8_bilinear_predict8x8 mmx sse2 ssse3 media neon msa/;
223$vp8_bilinear_predict8x8_media=vp8_bilinear_predict8x8_armv6;
224
225add_proto qw/void vp8_bilinear_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
226specialize qw/vp8_bilinear_predict8x4 mmx media neon msa/;
227$vp8_bilinear_predict8x4_media=vp8_bilinear_predict8x4_armv6;
228
229add_proto qw/void vp8_bilinear_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
230#TODO(johannkoenig): fix the neon version https://code.google.com/p/webm/issues/detail?id=892
231specialize qw/vp8_bilinear_predict4x4 mmx media msa/;
232$vp8_bilinear_predict4x4_media=vp8_bilinear_predict4x4_armv6;
233
234#
235# Encoder functions below this point.
236#
237if (vpx_config("CONFIG_VP8_ENCODER") eq "yes") {
238
239#
240# Block copy
241#
242if ($opts{arch} =~ /x86/) {
243    add_proto qw/void vp8_copy32xn/, "const unsigned char *src_ptr, int source_stride, unsigned char *dst_ptr, int dst_stride, int n";
244    specialize qw/vp8_copy32xn sse2 sse3/;
245}
246
247#
248# Forward DCT
249#
250add_proto qw/void vp8_short_fdct4x4/, "short *input, short *output, int pitch";
251specialize qw/vp8_short_fdct4x4 mmx sse2 media neon msa/;
252$vp8_short_fdct4x4_media=vp8_short_fdct4x4_armv6;
253
254add_proto qw/void vp8_short_fdct8x4/, "short *input, short *output, int pitch";
255specialize qw/vp8_short_fdct8x4 mmx sse2 media neon msa/;
256$vp8_short_fdct8x4_media=vp8_short_fdct8x4_armv6;
257
258add_proto qw/void vp8_short_walsh4x4/, "short *input, short *output, int pitch";
259specialize qw/vp8_short_walsh4x4 sse2 media neon msa/;
260$vp8_short_walsh4x4_media=vp8_short_walsh4x4_armv6;
261
262#
263# Quantizer
264#
265add_proto qw/void vp8_regular_quantize_b/, "struct block *, struct blockd *";
266specialize qw/vp8_regular_quantize_b sse2 sse4_1 msa/;
267
268add_proto qw/void vp8_fast_quantize_b/, "struct block *, struct blockd *";
269specialize qw/vp8_fast_quantize_b sse2 ssse3 neon msa/;
270
271#
272# Block subtraction
273#
274add_proto qw/int vp8_block_error/, "short *coeff, short *dqcoeff";
275specialize qw/vp8_block_error mmx sse2 msa/;
276$vp8_block_error_sse2=vp8_block_error_xmm;
277
278add_proto qw/int vp8_mbblock_error/, "struct macroblock *mb, int dc";
279specialize qw/vp8_mbblock_error mmx sse2 msa/;
280$vp8_mbblock_error_sse2=vp8_mbblock_error_xmm;
281
282add_proto qw/int vp8_mbuverror/, "struct macroblock *mb";
283specialize qw/vp8_mbuverror mmx sse2 msa/;
284$vp8_mbuverror_sse2=vp8_mbuverror_xmm;
285
286#
287# Motion search
288#
289add_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";
290specialize qw/vp8_full_search_sad sse3 sse4_1/;
291$vp8_full_search_sad_sse3=vp8_full_search_sadx3;
292$vp8_full_search_sad_sse4_1=vp8_full_search_sadx8;
293
294add_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";
295specialize qw/vp8_refining_search_sad sse3/;
296$vp8_refining_search_sad_sse3=vp8_refining_search_sadx4;
297
298add_proto qw/int vp8_diamond_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, union int_mv *best_mv, int search_param, int sad_per_bit, int *num00, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
299$vp8_diamond_search_sad_sse3=vp8_diamond_search_sadx4;
300
301#
302# Alt-ref Noise Reduction (ARNR)
303#
304if (vpx_config("CONFIG_REALTIME_ONLY") ne "yes") {
305    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";
306    specialize qw/vp8_temporal_filter_apply sse2 msa/;
307}
308
309#
310# Denoiser filter
311#
312if (vpx_config("CONFIG_TEMPORAL_DENOISING") eq "yes") {
313    add_proto qw/int vp8_denoiser_filter/, "unsigned char *mc_running_avg_y, int mc_avg_y_stride, unsigned char *running_avg_y, int avg_y_stride, unsigned char *sig, int sig_stride, unsigned int motion_magnitude, int increase_denoising";
314    specialize qw/vp8_denoiser_filter sse2 neon msa/;
315    add_proto qw/int vp8_denoiser_filter_uv/, "unsigned char *mc_running_avg, int mc_avg_stride, unsigned char *running_avg, int avg_stride, unsigned char *sig, int sig_stride, unsigned int motion_magnitude, int increase_denoising";
316    specialize qw/vp8_denoiser_filter_uv sse2 neon msa/;
317}
318
319# End of encoder only functions
320}
3211;
322