1 ## 2 ## Copyright (c) 2017 The WebM project authors. All Rights Reserved. 3 ## 4 ## Use of this source code is governed by a BSD-style license 5 ## that can be found in the LICENSE file in the root of the source 6 ## tree. An additional intellectual property rights grant can be found 7 ## in the file PATENTS. All contributing project authors may 8 ## be found in the AUTHORS file in the root of the source tree. 9 ## 10 11 sub vp8_common_forward_decls() { 12 print <<EOF 13 /* 14 * VP8 15 */ 16 17 struct blockd; 18 struct macroblockd; 19 struct loop_filter_info; 20 21 /* Encoder forward decls */ 22 struct block; 23 struct macroblock; 24 struct variance_vtable; 25 union int_mv; 26 struct yv12_buffer_config; 27 EOF 28 } 29 forward_decls qw/vp8_common_forward_decls/; 30 31 # 32 # Dequant 33 # 34 add_proto qw/void vp8_dequantize_b/, "struct blockd*, short *dqc"; 35 specialize qw/vp8_dequantize_b mmx neon msa mmi/; 36 37 add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char *output, int stride"; 38 specialize qw/vp8_dequant_idct_add mmx neon dspr2 msa mmi/; 39 40 add_proto qw/void vp8_dequant_idct_add_y_block/, "short *q, short *dq, unsigned char *dst, int stride, char *eobs"; 41 specialize qw/vp8_dequant_idct_add_y_block sse2 neon dspr2 msa mmi/; 42 43 add_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"; 44 specialize qw/vp8_dequant_idct_add_uv_block sse2 neon dspr2 msa mmi/; 45 46 # 47 # Loopfilter 48 # 49 add_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"; 50 specialize qw/vp8_loop_filter_mbv sse2 neon dspr2 msa mmi/; 51 52 add_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"; 53 specialize qw/vp8_loop_filter_bv sse2 neon dspr2 msa mmi/; 54 55 add_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"; 56 specialize qw/vp8_loop_filter_mbh sse2 neon dspr2 msa mmi/; 57 58 add_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"; 59 specialize qw/vp8_loop_filter_bh sse2 neon dspr2 msa mmi/; 60 61 62 add_proto qw/void vp8_loop_filter_simple_mbv/, "unsigned char *y, int ystride, const unsigned char *blimit"; 63 specialize qw/vp8_loop_filter_simple_mbv sse2 neon msa mmi/; 64 $vp8_loop_filter_simple_mbv_c=vp8_loop_filter_simple_vertical_edge_c; 65 $vp8_loop_filter_simple_mbv_sse2=vp8_loop_filter_simple_vertical_edge_sse2; 66 $vp8_loop_filter_simple_mbv_neon=vp8_loop_filter_mbvs_neon; 67 $vp8_loop_filter_simple_mbv_msa=vp8_loop_filter_simple_vertical_edge_msa; 68 $vp8_loop_filter_simple_mbv_mmi=vp8_loop_filter_simple_vertical_edge_mmi; 69 70 add_proto qw/void vp8_loop_filter_simple_mbh/, "unsigned char *y, int ystride, const unsigned char *blimit"; 71 specialize qw/vp8_loop_filter_simple_mbh sse2 neon msa mmi/; 72 $vp8_loop_filter_simple_mbh_c=vp8_loop_filter_simple_horizontal_edge_c; 73 $vp8_loop_filter_simple_mbh_sse2=vp8_loop_filter_simple_horizontal_edge_sse2; 74 $vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon; 75 $vp8_loop_filter_simple_mbh_msa=vp8_loop_filter_simple_horizontal_edge_msa; 76 $vp8_loop_filter_simple_mbh_mmi=vp8_loop_filter_simple_horizontal_edge_mmi; 77 78 add_proto qw/void vp8_loop_filter_simple_bv/, "unsigned char *y, int ystride, const unsigned char *blimit"; 79 specialize qw/vp8_loop_filter_simple_bv sse2 neon msa mmi/; 80 $vp8_loop_filter_simple_bv_c=vp8_loop_filter_bvs_c; 81 $vp8_loop_filter_simple_bv_sse2=vp8_loop_filter_bvs_sse2; 82 $vp8_loop_filter_simple_bv_neon=vp8_loop_filter_bvs_neon; 83 $vp8_loop_filter_simple_bv_msa=vp8_loop_filter_bvs_msa; 84 $vp8_loop_filter_simple_bv_mmi=vp8_loop_filter_bvs_mmi; 85 86 add_proto qw/void vp8_loop_filter_simple_bh/, "unsigned char *y, int ystride, const unsigned char *blimit"; 87 specialize qw/vp8_loop_filter_simple_bh sse2 neon msa mmi/; 88 $vp8_loop_filter_simple_bh_c=vp8_loop_filter_bhs_c; 89 $vp8_loop_filter_simple_bh_sse2=vp8_loop_filter_bhs_sse2; 90 $vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon; 91 $vp8_loop_filter_simple_bh_msa=vp8_loop_filter_bhs_msa; 92 $vp8_loop_filter_simple_bh_mmi=vp8_loop_filter_bhs_mmi; 93 94 # 95 # IDCT 96 # 97 #idct16 98 add_proto qw/void vp8_short_idct4x4llm/, "short *input, unsigned char *pred, int pitch, unsigned char *dst, int dst_stride"; 99 specialize qw/vp8_short_idct4x4llm mmx neon dspr2 msa mmi/; 100 101 #iwalsh1 102 add_proto qw/void vp8_short_inv_walsh4x4_1/, "short *input, short *output"; 103 specialize qw/vp8_short_inv_walsh4x4_1 dspr2/; 104 105 #iwalsh16 106 add_proto qw/void vp8_short_inv_walsh4x4/, "short *input, short *output"; 107 specialize qw/vp8_short_inv_walsh4x4 sse2 neon dspr2 msa mmi/; 108 109 #idct1_scalar_add 110 add_proto qw/void vp8_dc_only_idct_add/, "short input, unsigned char *pred, int pred_stride, unsigned char *dst, int dst_stride"; 111 specialize qw/vp8_dc_only_idct_add mmx neon dspr2 msa mmi/; 112 113 # 114 # RECON 115 # 116 add_proto qw/void vp8_copy_mem16x16/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"; 117 specialize qw/vp8_copy_mem16x16 sse2 neon dspr2 msa mmi/; 118 119 add_proto qw/void vp8_copy_mem8x8/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"; 120 specialize qw/vp8_copy_mem8x8 mmx neon dspr2 msa mmi/; 121 122 add_proto qw/void vp8_copy_mem8x4/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"; 123 specialize qw/vp8_copy_mem8x4 mmx neon dspr2 msa mmi/; 124 125 # 126 # Postproc 127 # 128 if (vpx_config("CONFIG_POSTPROC") eq "yes") { 129 130 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"; 131 132 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"; 133 134 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"; 135 136 add_proto qw/void vp8_filter_by_weight16x16/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight"; 137 specialize qw/vp8_filter_by_weight16x16 sse2 msa/; 138 139 add_proto qw/void vp8_filter_by_weight8x8/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight"; 140 specialize qw/vp8_filter_by_weight8x8 sse2 msa/; 141 142 add_proto qw/void vp8_filter_by_weight4x4/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight"; 143 } 144 145 # 146 # Subpixel 147 # 148 add_proto qw/void vp8_sixtap_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 149 specialize qw/vp8_sixtap_predict16x16 sse2 ssse3 neon dspr2 msa mmi/; 150 151 add_proto qw/void vp8_sixtap_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 152 specialize qw/vp8_sixtap_predict8x8 sse2 ssse3 neon dspr2 msa mmi/; 153 154 add_proto qw/void vp8_sixtap_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 155 specialize qw/vp8_sixtap_predict8x4 sse2 ssse3 neon dspr2 msa mmi/; 156 157 add_proto qw/void vp8_sixtap_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 158 specialize qw/vp8_sixtap_predict4x4 mmx ssse3 neon dspr2 msa mmi/; 159 160 add_proto qw/void vp8_bilinear_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 161 specialize qw/vp8_bilinear_predict16x16 sse2 ssse3 neon msa/; 162 163 add_proto qw/void vp8_bilinear_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 164 specialize qw/vp8_bilinear_predict8x8 sse2 ssse3 neon msa/; 165 166 add_proto qw/void vp8_bilinear_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 167 specialize qw/vp8_bilinear_predict8x4 mmx neon msa/; 168 169 add_proto qw/void vp8_bilinear_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"; 170 specialize qw/vp8_bilinear_predict4x4 mmx neon msa/; 171 172 # 173 # Encoder functions below this point. 174 # 175 if (vpx_config("CONFIG_VP8_ENCODER") eq "yes") { 176 177 # 178 # Block copy 179 # 180 if ($opts{arch} =~ /x86/) { 181 add_proto qw/void vp8_copy32xn/, "const unsigned char *src_ptr, int source_stride, unsigned char *dst_ptr, int dst_stride, int n"; 182 specialize qw/vp8_copy32xn sse2 sse3/; 183 } 184 185 # 186 # Forward DCT 187 # 188 add_proto qw/void vp8_short_fdct4x4/, "short *input, short *output, int pitch"; 189 specialize qw/vp8_short_fdct4x4 sse2 neon msa mmi/; 190 191 add_proto qw/void vp8_short_fdct8x4/, "short *input, short *output, int pitch"; 192 specialize qw/vp8_short_fdct8x4 sse2 neon msa mmi/; 193 194 add_proto qw/void vp8_short_walsh4x4/, "short *input, short *output, int pitch"; 195 specialize qw/vp8_short_walsh4x4 sse2 neon msa mmi/; 196 197 # 198 # Quantizer 199 # 200 add_proto qw/void vp8_regular_quantize_b/, "struct block *, struct blockd *"; 201 specialize qw/vp8_regular_quantize_b sse2 sse4_1 msa mmi/; 202 203 add_proto qw/void vp8_fast_quantize_b/, "struct block *, struct blockd *"; 204 specialize qw/vp8_fast_quantize_b sse2 ssse3 neon msa mmi/; 205 206 # 207 # Block subtraction 208 # 209 add_proto qw/int vp8_block_error/, "short *coeff, short *dqcoeff"; 210 specialize qw/vp8_block_error sse2 msa/; 211 212 add_proto qw/int vp8_mbblock_error/, "struct macroblock *mb, int dc"; 213 specialize qw/vp8_mbblock_error sse2 msa/; 214 215 add_proto qw/int vp8_mbuverror/, "struct macroblock *mb"; 216 specialize qw/vp8_mbuverror sse2 msa/; 217 218 # 219 # Motion search 220 # 221 add_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"; 222 specialize qw/vp8_full_search_sad sse3 sse4_1/; 223 $vp8_full_search_sad_sse3=vp8_full_search_sadx3; 224 $vp8_full_search_sad_sse4_1=vp8_full_search_sadx8; 225 226 add_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"; 227 specialize qw/vp8_refining_search_sad sse2 msa/; 228 $vp8_refining_search_sad_sse2=vp8_refining_search_sadx4; 229 $vp8_refining_search_sad_msa=vp8_refining_search_sadx4; 230 231 add_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"; 232 specialize qw/vp8_diamond_search_sad sse2 msa/; 233 $vp8_diamond_search_sad_sse2=vp8_diamond_search_sadx4; 234 $vp8_diamond_search_sad_msa=vp8_diamond_search_sadx4; 235 236 # 237 # Alt-ref Noise Reduction (ARNR) 238 # 239 if (vpx_config("CONFIG_REALTIME_ONLY") ne "yes") { 240 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"; 241 specialize qw/vp8_temporal_filter_apply sse2 msa/; 242 } 243 244 # 245 # Denoiser filter 246 # 247 if (vpx_config("CONFIG_TEMPORAL_DENOISING") eq "yes") { 248 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"; 249 specialize qw/vp8_denoiser_filter sse2 neon msa/; 250 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"; 251 specialize qw/vp8_denoiser_filter_uv sse2 neon msa/; 252 } 253 254 # End of encoder only functions 255 } 256 1; 257
