av1/common/av1_rtcd_defs.pl

   1 sub av1_common_forward_decls() {
   2 print <<EOF
   3 /*
   4  * AV1
   5  */
   6
   7 #include "aom/aom_integer.h"
   8 #include "aom_dsp/txfm_common.h"
   9 #include "av1/common/common.h"
  10 #include "av1/common/enums.h"
  11 #include "av1/common/quant_common.h"
  12 #include "av1/common/filter.h"
  13 #include "av1/common/convolve.h"
  14 #include "av1/common/av1_txfm.h"
  15 #include "av1/common/odintrin.h"
  16
  17 #if CONFIG_LOOP_RESTORATION
  18 #include "av1/common/restoration.h"
  19 #endif
  20
  21 #if CONFIG_DAALA_TX
  22 #include "av1/common/daala_inv_txfm.h"
  23 #endif
  24
  25 struct macroblockd;
  26
  27 /* Encoder forward decls */
  28 struct macroblock;
  29 struct txfm_param;
  30 struct aom_variance_vtable;
  31 struct search_site_config;
  32 struct mv;
  33 union int_mv;
  34 struct yv12_buffer_config;
  35 EOF
  36 }
  37 forward_decls qw/av1_common_forward_decls/;
  38
  39 # functions that are 64 bit only.
  40 $mmx_x86_64 = $sse2_x86_64 = $ssse3_x86_64 = $avx_x86_64 = $avx2_x86_64 = '';
  41 if ($opts{arch} eq "x86_64") {
  42   $mmx_x86_64 = 'mmx';
  43   $sse2_x86_64 = 'sse2';
  44   $ssse3_x86_64 = 'ssse3';
  45   $avx_x86_64 = 'avx';
  46   $avx2_x86_64 = 'avx2';
  47 }
  48
  49 #
  50 # 10/12-tap convolution filters
  51 #
  52 add_proto qw/void av1_lowbd_convolve_init/, "void";
  53 specialize qw/av1_lowbd_convolve_init ssse3/;
  54
  55 add_proto qw/void av1_convolve_horiz/, "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, ConvolveParams *conv_params";
  56 specialize qw/av1_convolve_horiz ssse3/;
  57
  58 add_proto qw/void av1_convolve_vert/, "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, ConvolveParams *conv_params";
  59 specialize qw/av1_convolve_vert ssse3/;
  60
  61 if (aom_config("CONFIG_HORZONLY_FRAME_SUPERRES") eq "yes") {
  62   add_proto qw/void av1_convolve_horiz_rs/, "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int w, int h, const int16_t *x_filters, int interp_taps, const int x0_qn, const int x_step_qn";
  63   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
  64     add_proto qw/void av1_highbd_convolve_horiz_rs/, "const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const int16_t *x_filters, int interp_taps, const int x0_qn, const int x_step_qn, int bd";
  65   }
  66 }
  67
  68 if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
  69   add_proto qw/void av1_highbd_convolve_init/, "void";
  70   specialize qw/av1_highbd_convolve_init sse4_1/;
  71   add_proto qw/void av1_highbd_convolve_horiz/, "const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, int avg, int bd";
  72   specialize qw/av1_highbd_convolve_horiz sse4_1/;
  73   add_proto qw/void av1_highbd_convolve_vert/, "const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w, int h, const InterpFilterParams fp, const int subpel_x_q4, int x_step_q4, int avg, int bd";
  74   specialize qw/av1_highbd_convolve_vert sse4_1/;
  75 }
  76
  77 #
  78 # txb
  79 #
  80 if (aom_config("CONFIG_LV_MAP") eq "yes") {
  81   add_proto qw/void av1_get_br_level_counts/, "const uint8_t *const levels, const int width, const int height, uint8_t *const level_counts";
  82   specialize qw/av1_get_br_level_counts sse2/;
  83 }
  84
  85 #
  86 # Inverse dct
  87 #
  88 add_proto qw/void av1_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
  89 if (aom_config("CONFIG_DAALA_TX4") ne "yes") {
  90   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
  91     specialize qw/av1_iht4x4_16_add sse2/;
  92   } else {
  93     specialize qw/av1_iht4x4_16_add sse2 neon/;
  94   }
  95 }
  96
  97 if (aom_config("CONFIG_DAALA_TX") ne "yes") {
  98   add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
  99   specialize qw/av1_iht4x8_32_add sse2/;
 100
 101   add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 102   specialize qw/av1_iht8x4_32_add sse2/;
 103
 104   add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 105   specialize qw/av1_iht8x16_128_add sse2/;
 106
 107   add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 108   specialize qw/av1_iht16x8_128_add sse2/;
 109
 110   add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 111   specialize qw/av1_iht16x32_512_add sse2/;
 112
 113   add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 114   specialize qw/av1_iht32x16_512_add sse2/;
 115
 116   add_proto qw/void av1_iht4x16_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 117
 118   add_proto qw/void av1_iht16x4_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 119
 120   add_proto qw/void av1_iht8x32_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 121
 122   add_proto qw/void av1_iht32x8_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 123
 124   add_proto qw/void av1_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 125   if (aom_config("CONFIG_DAALA_TX8") ne "yes") {
 126     if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 127       specialize qw/av1_iht8x8_64_add sse2/;
 128    } else {
 129        specialize qw/av1_iht8x8_64_add sse2 neon/;
 130     }
 131   }
 132
 133   add_proto qw/void av1_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 134
 135   if (aom_config("CONFIG_DAALA_TX16") ne "yes") {
 136     if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 137       specialize qw/av1_iht16x16_256_add sse2 avx2/;
 138     } else {
 139       specialize qw/av1_iht16x16_256_add sse2 avx2/;
 140     }
 141   }
 142
 143   add_proto qw/void av1_iht32x32_1024_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 144
 145   if (aom_config("CONFIG_HIGHBITDEPTH") ne "yes") {
 146   }
 147
 148   add_proto qw/void av1_iht32x32_1024_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 149
 150   if (aom_config("CONFIG_TX64X64") eq "yes") {
 151     add_proto qw/void av1_iht64x64_4096_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 152     add_proto qw/void av1_iht32x64_2048_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 153     add_proto qw/void av1_iht64x32_2048_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 154     add_proto qw/void av1_iht16x64_1024_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 155     add_proto qw/void av1_iht64x16_1024_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 156   }
 157 }
 158
 159 if (aom_config("CONFIG_NEW_QUANT") eq "yes") {
 160   add_proto qw/void quantize_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 161
 162   add_proto qw/void quantize_fp_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 163
 164   add_proto qw/void quantize_32x32_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 165
 166   add_proto qw/void quantize_32x32_fp_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 167
 168   if (aom_config("CONFIG_TX64X64") eq "yes") {
 169     add_proto qw/void quantize_64x64_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 170
 171     add_proto qw/void quantize_64x64_fp_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 172   }
 173 }
 174
 175 # FILTER_INTRA predictor functions
 176 if (aom_config("CONFIG_FILTER_INTRA") eq "yes") {
 177   add_proto qw/void av1_dc_filter_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left";
 178   add_proto qw/void av1_v_filter_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left";
 179   add_proto qw/void av1_h_filter_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left";
 180   add_proto qw/void av1_d117_filter_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left";
 181   add_proto qw/void av1_d153_filter_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left";
 182   add_proto qw/void av1_paeth_filter_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left";
 183   # High bitdepth functions
 184   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 185     add_proto qw/void av1_highbd_dc_filter_predictor/, "uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int bd";
 186     add_proto qw/void av1_highbd_v_filter_predictor/, "uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int bd";
 187     add_proto qw/void av1_highbd_h_filter_predictor/, "uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int bd";
 188     add_proto qw/void av1_highbd_d117_filter_predictor/, "uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int bd";
 189     add_proto qw/void av1_highbd_d153_filter_predictor/, "uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int bd";
 190     add_proto qw/void av1_highbd_paeth_filter_predictor/, "uint16_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint16_t *above, const uint16_t *left, int bd";
 191   }
 192 }
 193
 194 # High bitdepth functions
 195 if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 196   #
 197   # Sub Pixel Filters
 198   #
 199   add_proto qw/void av1_highbd_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 200
 201   add_proto qw/void av1_highbd_convolve_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 202
 203   add_proto qw/void av1_highbd_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 204   specialize qw/av1_highbd_convolve8/, "$sse2_x86_64";
 205
 206   add_proto qw/void av1_highbd_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 207   specialize qw/av1_highbd_convolve8_horiz/, "$sse2_x86_64";
 208
 209   add_proto qw/void av1_highbd_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 210   specialize qw/av1_highbd_convolve8_vert/, "$sse2_x86_64";
 211
 212   add_proto qw/void av1_highbd_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 213   specialize qw/av1_highbd_convolve8_avg/, "$sse2_x86_64";
 214
 215   add_proto qw/void av1_highbd_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 216   specialize qw/av1_highbd_convolve8_avg_horiz/, "$sse2_x86_64";
 217
 218   add_proto qw/void av1_highbd_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 219   specialize qw/av1_highbd_convolve8_avg_vert/, "$sse2_x86_64";
 220
 221   #
 222   # dct
 223   #
 224   add_proto qw/void av1_highbd_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 225
 226   add_proto qw/void av1_highbd_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 227
 228   add_proto qw/void av1_highbd_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 229
 230   add_proto qw/void av1_highbd_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 231
 232   add_proto qw/void av1_highbd_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 233
 234   add_proto qw/void av1_highbd_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 235
 236   add_proto qw/void av1_highbd_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 237
 238   add_proto qw/void av1_highbd_iht4x16_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 239
 240   add_proto qw/void av1_highbd_iht16x4_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 241
 242   add_proto qw/void av1_highbd_iht8x32_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 243
 244   add_proto qw/void av1_highbd_iht32x8_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 245
 246   add_proto qw/void av1_highbd_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, const struct txfm_param *param";
 247
 248   add_proto qw/void av1_highbd_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, const struct txfm_param *param";
 249 }
 250
 251 #inv txfm
 252 add_proto qw/void av1_inv_txfm2d_add_4x8/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 253 add_proto qw/void av1_inv_txfm2d_add_8x4/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 254 add_proto qw/void av1_inv_txfm2d_add_8x16/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 255 add_proto qw/void av1_inv_txfm2d_add_16x8/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 256 add_proto qw/void av1_inv_txfm2d_add_16x32/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 257 add_proto qw/void av1_inv_txfm2d_add_32x16/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 258 add_proto qw/void av1_inv_txfm2d_add_4x4/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 259 if (aom_config("CONFIG_DAALA_TX4") ne "yes") {
 260   specialize qw/av1_inv_txfm2d_add_4x4 sse4_1/;
 261 }
 262 add_proto qw/void av1_inv_txfm2d_add_8x8/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 263 if (aom_config("CONFIG_DAALA_TX8") ne "yes") {
 264   specialize qw/av1_inv_txfm2d_add_8x8 sse4_1/;
 265 }
 266 add_proto qw/void av1_inv_txfm2d_add_16x16/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 267 if (aom_config("CONFIG_DAALA_TX16") ne "yes") {
 268   specialize qw/av1_inv_txfm2d_add_16x16 sse4_1/;
 269 }
 270 add_proto qw/void av1_inv_txfm2d_add_32x32/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 271 if (aom_config("CONFIG_DAALA_TX32") ne "yes") {
 272   specialize qw/av1_inv_txfm2d_add_32x32 avx2/;
 273 }
 274 if (aom_config("CONFIG_TX64X64") eq "yes") {
 275   add_proto qw/void av1_inv_txfm2d_add_64x64/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 276   add_proto qw/void av1_inv_txfm2d_add_32x64/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 277   add_proto qw/void av1_inv_txfm2d_add_64x32/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 278   add_proto qw/void av1_inv_txfm2d_add_16x64/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 279   add_proto qw/void av1_inv_txfm2d_add_64x16/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 280 }
 281 add_proto qw/void av1_inv_txfm2d_add_4x16/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 282 add_proto qw/void av1_inv_txfm2d_add_16x4/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 283 add_proto qw/void av1_inv_txfm2d_add_8x32/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 284 add_proto qw/void av1_inv_txfm2d_add_32x8/, "const int32_t *input, uint16_t *output, int stride, TX_TYPE tx_type, int bd";
 285
 286 #
 287 # Encoder functions below this point.
 288 #
 289 if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 290
 291   # ENCODEMB INVOKE
 292
 293   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 294     # the transform coefficients are held in 32-bit
 295     # values, so the assembler code for  av1_block_error can no longer be used.
 296     add_proto qw/int64_t av1_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz";
 297     specialize qw/av1_block_error avx2/;
 298
 299     add_proto qw/void av1_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 300     specialize qw/av1_quantize_fp sse2 avx2/;
 301
 302     add_proto qw/void av1_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 303     specialize qw/av1_quantize_fp_32x32 avx2/;
 304
 305     if (aom_config("CONFIG_TX64X64") eq "yes") {
 306       add_proto qw/void av1_quantize_fp_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 307     }
 308   } else {
 309     add_proto qw/int64_t av1_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz";
 310     specialize qw/av1_block_error sse2 avx2 msa/;
 311
 312     add_proto qw/int64_t av1_block_error_fp/, "const int16_t *coeff, const int16_t *dqcoeff, int block_size";
 313     specialize qw/av1_block_error_fp neon sse2/;
 314
 315     add_proto qw/void av1_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 316     specialize qw/av1_quantize_fp neon sse2 avx2/, "$ssse3_x86_64";
 317
 318     add_proto qw/void av1_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 319     specialize qw/av1_quantize_fp_32x32 avx2/, "$ssse3_x86_64";
 320
 321     if (aom_config("CONFIG_TX64X64") eq "yes") {
 322       add_proto qw/void av1_quantize_fp_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 323     }
 324
 325   }
 326
 327   # fdct functions
 328
 329   add_proto qw/void av1_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 330   if (aom_config("CONFIG_DAALA_TX4") ne "yes") {
 331     specialize qw/av1_fht4x4 sse2/;
 332   }
 333
 334   add_proto qw/void av1_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
 335
 336   add_proto qw/void av1_fht8x8/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 337   if (aom_config("CONFIG_DAALA_TX8") ne "yes") {
 338     specialize qw/av1_fht8x8 sse2/;
 339   }
 340
 341   add_proto qw/void av1_fht16x16/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 342   if (aom_config("CONFIG_DAALA_TX16") ne "yes") {
 343     specialize qw/av1_fht16x16 sse2 avx2/;
 344   }
 345
 346   add_proto qw/void av1_fht32x32/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 347   if (aom_config("CONFIG_DAALA_TX32") ne "yes") {
 348     specialize qw/av1_fht32x32 sse2 avx2/;
 349   }
 350
 351   if (aom_config("CONFIG_TX64X64") eq "yes") {
 352     add_proto qw/void av1_fht64x64/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 353     add_proto qw/void av1_fht32x64/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 354     add_proto qw/void av1_fht64x32/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 355     add_proto qw/void av1_fht16x64/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 356     add_proto qw/void av1_fht64x16/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 357   }
 358
 359   add_proto qw/void av1_fht4x8/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 360   specialize qw/av1_fht4x8 sse2/;
 361
 362   add_proto qw/void av1_fht8x4/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 363   specialize qw/av1_fht8x4 sse2/;
 364
 365   add_proto qw/void av1_fht8x16/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 366   specialize qw/av1_fht8x16 sse2/;
 367
 368   add_proto qw/void av1_fht16x8/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 369   specialize qw/av1_fht16x8 sse2/;
 370
 371   add_proto qw/void av1_fht16x32/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 372   specialize qw/av1_fht16x32 sse2/;
 373
 374   add_proto qw/void av1_fht32x16/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 375   specialize qw/av1_fht32x16 sse2/;
 376
 377   add_proto qw/void av1_fht4x16/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 378
 379   add_proto qw/void av1_fht16x4/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 380
 381   add_proto qw/void av1_fht8x32/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 382
 383   add_proto qw/void av1_fht32x8/, "const int16_t *input, tran_low_t *output, int stride, struct txfm_param *param";
 384
 385
 386   add_proto qw/void av1_fwd_idtx/, "const int16_t *src_diff, tran_low_t *coeff, int stride, int bsx, int bsy, TX_TYPE tx_type";
 387
 388   #fwd txfm
 389   add_proto qw/void av1_fwd_txfm2d_4x8/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 390   add_proto qw/void av1_fwd_txfm2d_8x4/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 391   add_proto qw/void av1_fwd_txfm2d_8x16/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 392   add_proto qw/void av1_fwd_txfm2d_16x8/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 393   add_proto qw/void av1_fwd_txfm2d_16x32/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 394   add_proto qw/void av1_fwd_txfm2d_32x16/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 395   add_proto qw/void av1_fwd_txfm2d_4x16/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 396   add_proto qw/void av1_fwd_txfm2d_16x4/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 397   add_proto qw/void av1_fwd_txfm2d_8x32/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 398   add_proto qw/void av1_fwd_txfm2d_32x8/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 399   add_proto qw/void av1_fwd_txfm2d_4x4/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 400   if (aom_config("CONFIG_DAALA_TX4") ne "yes") {
 401     specialize qw/av1_fwd_txfm2d_4x4 sse4_1/;
 402   }
 403   add_proto qw/void av1_fwd_txfm2d_8x8/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 404   if (aom_config("CONFIG_DAALA_TX8") ne "yes") {
 405     specialize qw/av1_fwd_txfm2d_8x8 sse4_1/;
 406   }
 407   add_proto qw/void av1_fwd_txfm2d_16x16/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 408   if (aom_config("CONFIG_DAALA_TX16") ne "yes") {
 409     specialize qw/av1_fwd_txfm2d_16x16 sse4_1/;
 410   }
 411   add_proto qw/void av1_fwd_txfm2d_32x32/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 412   if (aom_config("CONFIG_DAALA_TX32") ne "yes") {
 413     specialize qw/av1_fwd_txfm2d_32x32 sse4_1/;
 414   }
 415
 416   if (aom_config("CONFIG_TX64X64") eq "yes") {
 417     add_proto qw/void av1_fwd_txfm2d_64x64/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 418     add_proto qw/void av1_fwd_txfm2d_32x64/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 419     add_proto qw/void av1_fwd_txfm2d_64x32/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 420     add_proto qw/void av1_fwd_txfm2d_16x64/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 421     add_proto qw/void av1_fwd_txfm2d_64x16/, "const int16_t *input, int32_t *output, int stride, TX_TYPE tx_type, int bd";
 422   }
 423   #
 424   # Motion search
 425   #
 426   add_proto qw/int av1_full_search_sad/, "const struct macroblock *x, const struct mv *ref_mv, int sad_per_bit, int distance, const struct aom_variance_vtable *fn_ptr, const struct mv *center_mv, struct mv *best_mv";
 427   specialize qw/av1_full_search_sad sse3 sse4_1/;
 428   $av1_full_search_sad_sse3=av1_full_search_sadx3;
 429   $av1_full_search_sad_sse4_1=av1_full_search_sadx8;
 430
 431   add_proto qw/int av1_diamond_search_sad/, "struct macroblock *x, const struct search_site_config *cfg,  struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, const struct aom_variance_vtable *fn_ptr, const struct mv *center_mv";
 432
 433   add_proto qw/int av1_full_range_search/, "const struct macroblock *x, const struct search_site_config *cfg, struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, const struct aom_variance_vtable *fn_ptr, const struct mv *center_mv";
 434
 435   add_proto qw/void av1_temporal_filter_apply/, "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count";
 436   specialize qw/av1_temporal_filter_apply sse2 msa/;
 437
 438   if (aom_config("CONFIG_AOM_QM") eq "yes") {
 439     add_proto qw/void av1_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr, int log_scale";
 440   } else {
 441     add_proto qw/void av1_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, int log_scale";
 442   }
 443
 444   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 445
 446     # ENCODEMB INVOKE
 447     if (aom_config("CONFIG_NEW_QUANT") eq "yes") {
 448       add_proto qw/void highbd_quantize_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 449
 450       add_proto qw/void highbd_quantize_fp_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 451
 452       add_proto qw/void highbd_quantize_32x32_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 453
 454       add_proto qw/void highbd_quantize_32x32_fp_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 455
 456       if (aom_config("CONFIG_TX64X64") eq "yes") {
 457         add_proto qw/void highbd_quantize_64x64_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 458
 459         add_proto qw/void highbd_quantize_64x64_fp_nuq/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *quant_ptr, const int16_t *dequant_ptr, const cuml_bins_type_nuq *cuml_bins_ptr, const dequant_val_type_nuq *dequant_val, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, uint16_t *eob_ptr, const int16_t *scan";
 460       }
 461     }
 462
 463     add_proto qw/int64_t av1_highbd_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd";
 464     if (aom_config("CONFIG_DAALA_TX") ne "yes") {
 465       specialize qw/av1_highbd_block_error sse2/;
 466     }
 467
 468     add_proto qw/void av1_highbd_temporal_filter_apply/, "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count";
 469
 470   }
 471
 472   add_proto qw/void av1_highbd_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, int log_scale";
 473   specialize qw/av1_highbd_quantize_fp sse4_1 avx2/;
 474
 475   add_proto qw/void av1_highbd_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
 476
 477   # End av1_high encoder functions
 478
 479   add_proto qw/uint64_t av1_wedge_sse_from_residuals/, "const int16_t *r1, const int16_t *d, const uint8_t *m, int N";
 480   specialize qw/av1_wedge_sse_from_residuals sse2/;
 481   add_proto qw/int av1_wedge_sign_from_residuals/, "const int16_t *ds, const uint8_t *m, int N, int64_t limit";
 482   specialize qw/av1_wedge_sign_from_residuals sse2/;
 483   add_proto qw/void av1_wedge_compute_delta_squares/, "int16_t *d, const int16_t *a, const int16_t *b, int N";
 484   specialize qw/av1_wedge_compute_delta_squares sse2/;
 485
 486 }
 487 # end encoder functions
 488
 489 # Deringing Functions
 490
 491 add_proto qw/int cdef_find_dir/, "const uint16_t *img, int stride, int32_t *var, int coeff_shift";
 492 if (aom_config("CONFIG_CDEF_SINGLEPASS") ne "yes") {
 493   add_proto qw/void aom_clpf_block_hbd/, "uint16_t *dst, const uint16_t *src, int dstride, int sstride, int sizex, int sizey, unsigned int strength, unsigned int bd";
 494   add_proto qw/void aom_clpf_hblock_hbd/, "uint16_t *dst, const uint16_t *src, int dstride, int sstride, int sizex, int sizey, unsigned int strength, unsigned int bd";
 495   add_proto qw/void aom_clpf_block/, "uint8_t *dst, const uint16_t *src, int dstride, int sstride, int sizex, int sizey, unsigned int strength, unsigned int bd";
 496   add_proto qw/void aom_clpf_hblock/, "uint8_t *dst, const uint16_t *src, int dstride, int sstride, int sizex, int sizey, unsigned int strength, unsigned int bd";
 497   add_proto qw/void cdef_direction_4x4/, "uint16_t *y, int ystride, const uint16_t *in, int threshold, int dir, int damping";
 498   add_proto qw/void cdef_direction_8x8/, "uint16_t *y, int ystride, const uint16_t *in, int threshold, int dir, int damping";
 499   add_proto qw/void copy_8x8_16bit_to_8bit/, "uint8_t *dst, int dstride, const uint16_t *src, int sstride";
 500   add_proto qw/void copy_4x4_16bit_to_8bit/, "uint8_t *dst, int dstride, const uint16_t *src, int sstride";
 501   add_proto qw/void copy_8x8_16bit_to_16bit/, "uint16_t *dst, int dstride, const uint16_t *src, int sstride";
 502   add_proto qw/void copy_4x4_16bit_to_16bit/, "uint16_t *dst, int dstride, const uint16_t *src, int sstride";
 503 } else {
 504   add_proto qw/void cdef_filter_block/, "uint8_t *dst8, uint16_t *dst16, int dstride, const uint16_t *in, int pri_strength, int sec_strength, int dir, int pri_damping, int sec_damping, int bsize, int max, int coeff_shift";
 505 }
 506
 507 add_proto qw/void copy_rect8_8bit_to_16bit/, "uint16_t *dst, int dstride, const uint8_t *src, int sstride, int v, int h";
 508 add_proto qw/void copy_rect8_16bit_to_16bit/, "uint16_t *dst, int dstride, const uint16_t *src, int sstride, int v, int h";
 509
 510 # VS compiling for 32 bit targets does not support vector types in
 511 # structs as arguments, which makes the v256 type of the intrinsics
 512 # hard to support, so optimizations for this target are disabled.
 513 if ($opts{config} !~ /libs-x86-win32-vs.*/) {
 514   if (aom_config("CONFIG_CDEF_SINGLEPASS") eq "yes") {
 515     specialize qw/cdef_find_dir sse2 ssse3 sse4_1 avx2 neon/;
 516     specialize qw/cdef_filter_block sse2 ssse3 sse4_1 avx2 neon/;
 517     specialize qw/copy_rect8_8bit_to_16bit sse2 ssse3 sse4_1 avx2 neon/;
 518     specialize qw/copy_rect8_16bit_to_16bit sse2 ssse3 sse4_1 avx2 neon/;
 519   } else {
 520     specialize qw/cdef_find_dir sse2 ssse3 sse4_1 neon/;
 521     specialize qw/aom_clpf_block_hbd sse2 ssse3 sse4_1 neon/;
 522     specialize qw/aom_clpf_hblock_hbd sse2 ssse3 sse4_1 neon/;
 523     specialize qw/aom_clpf_block sse2 ssse3 sse4_1 neon/;
 524     specialize qw/aom_clpf_hblock sse2 ssse3 sse4_1 neon/;
 525     specialize qw/cdef_find_dir sse2 ssse3 sse4_1 neon/;
 526     specialize qw/cdef_direction_4x4 sse2 ssse3 sse4_1 neon/;
 527     specialize qw/cdef_direction_8x8 sse2 ssse3 sse4_1 neon/;
 528
 529     specialize qw/copy_8x8_16bit_to_8bit sse2 ssse3 sse4_1 neon/;
 530     specialize qw/copy_4x4_16bit_to_8bit sse2 ssse3 sse4_1 neon/;
 531     specialize qw/copy_8x8_16bit_to_16bit sse2 ssse3 sse4_1 neon/;
 532     specialize qw/copy_4x4_16bit_to_16bit sse2 ssse3 sse4_1 neon/;
 533     specialize qw/copy_rect8_8bit_to_16bit sse2 ssse3 sse4_1 neon/;
 534     specialize qw/copy_rect8_16bit_to_16bit sse2 ssse3 sse4_1 neon/;
 535   }
 536 }
 537
 538 # WARPED_MOTION / GLOBAL_MOTION functions
 539
 540 add_proto qw/void av1_warp_affine/, "const int32_t *mat, const uint8_t *ref, int width, int height, int stride, uint8_t *pred, int p_col, int p_row, int p_width, int p_height, int p_stride, int subsampling_x, int subsampling_y, ConvolveParams *conv_params, int16_t alpha, int16_t beta, int16_t gamma, int16_t delta";
 541
 542 if (aom_config("CONFIG_JNT_COMP") eq "yes") {
 543   if (aom_config("CONFIG_JNT_COMP") eq "yes") {
 544     specialize qw/av1_warp_affine sse4_1/;
 545   }
 546 } else {
 547   specialize qw/av1_warp_affine sse2 ssse3/;
 548 }
 549
 550 if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 551   add_proto qw/void av1_highbd_warp_affine/, "const int32_t *mat, const uint16_t *ref, int width, int height, int stride, uint16_t *pred, int p_col, int p_row, int p_width, int p_height, int p_stride, int subsampling_x, int subsampling_y, int bd, ConvolveParams *conv_params, int16_t alpha, int16_t beta, int16_t gamma, int16_t delta";
 552
 553 if (aom_config("CONFIG_JNT_COMP") eq "yes") {
 554   if (aom_config("CONFIG_JNT_COMP") eq "yes") {
 555     specialize qw/av1_highbd_warp_affine sse4_1/;
 556   }
 557 } else {
 558   specialize qw/av1_highbd_warp_affine ssse3/;
 559 }
 560 }
 561
 562 if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 563   add_proto qw/double compute_cross_correlation/, "unsigned char *im1, int stride1, int x1, int y1, unsigned char *im2, int stride2, int x2, int y2";
 564   specialize qw/compute_cross_correlation sse4_1/;
 565 }
 566
 567 # LOOP_RESTORATION functions
 568
 569 if (aom_config("CONFIG_LOOP_RESTORATION") eq "yes") {
 570   add_proto qw/void apply_selfguided_restoration/, "const uint8_t *dat, int width, int height, int stride, int eps, const int *xqd, uint8_t *dst, int dst_stride, int32_t *tmpbuf, int bit_depth, int highbd";
 571   specialize qw/apply_selfguided_restoration sse4_1/;
 572
 573   add_proto qw/void av1_selfguided_restoration/, "const uint8_t *dgd, int width, int height, int stride, int32_t *flt1, int32_t *flt2, int flt_stride, const sgr_params_type *params, int bit_depth, int highbd";
 574   specialize qw/av1_selfguided_restoration sse4_1/;
 575 }
 576
 577 # CONVOLVE_ROUND/COMPOUND_ROUND functions
 578
 579 if (aom_config("CONFIG_CONVOLVE_ROUND") eq "yes") {
 580   add_proto qw/void av1_convolve_2d/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params";
 581   specialize qw/av1_convolve_2d sse2 avx2/;
 582   add_proto qw/void av1_convolve_rounding/, "const int32_t *src, int src_stride, uint8_t *dst, int dst_stride, int w, int h, int bits";
 583   specialize qw/av1_convolve_rounding avx2/;
 584
 585   if (aom_config("CONFIG_COMPOUND_ROUND") ne "yes") {
 586     add_proto qw/void av1_convolve_2d_copy/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params";
 587     specialize qw/av1_convolve_2d_copy sse2/;
 588     add_proto qw/void av1_convolve_x/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params";
 589     specialize qw/av1_convolve_x sse2/;
 590     add_proto qw/void av1_convolve_y/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params";
 591     specialize qw/av1_convolve_y sse2/;
 592   }
 593
 594   add_proto qw/void av1_convolve_2d_scale/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_qn, const int x_step_qn, const int subpel_y_q4, const int y_step_qn, ConvolveParams *conv_params";
 595   if (aom_config("CONFIG_COMPOUND_ROUND") ne "yes") {
 596     specialize qw/av1_convolve_2d_scale sse4_1/;
 597   }
 598
 599   if (aom_config("CONFIG_JNT_COMP") eq "yes") {
 600     add_proto qw/void av1_jnt_convolve_2d/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params";
 601     specialize qw/av1_jnt_convolve_2d sse4_1/;
 602
 603     if (aom_config("CONFIG_COMPOUND_ROUND") ne "yes") {
 604       add_proto qw/void av1_jnt_convolve_2d_copy/, "const uint8_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params";
 605       specialize qw/av1_jnt_convolve_2d_copy sse2/;
 606     }
 607   }
 608
 609   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 610     add_proto qw/void av1_highbd_convolve_2d/, "const uint16_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int subpel_y_q4, ConvolveParams *conv_params, int bd";
 611     specialize qw/av1_highbd_convolve_2d ssse3 avx2/;
 612     add_proto qw/void av1_highbd_convolve_rounding/, "const int32_t *src, int src_stride, uint8_t *dst, int dst_stride, int w, int h, int bits, int bd";
 613     specialize qw/av1_highbd_convolve_rounding avx2/;
 614
 615     add_proto qw/void av1_highbd_convolve_2d_scale/, "const uint16_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride, int w, int h, InterpFilterParams *filter_params_x, InterpFilterParams *filter_params_y, const int subpel_x_q4, const int x_step_qn, const int subpel_y_q4, const int y_step_qn, ConvolveParams *conv_params, int bd";
 616     if (aom_config("CONFIG_COMPOUND_ROUND") ne "yes") {
 617         specialize qw/av1_highbd_convolve_2d_scale sse4_1/;
 618     }
 619   }
 620 }
 621
 622 # INTRA_EDGE functions
 623 if (aom_config("CONFIG_INTRA_EDGE") eq "yes") {
 624   add_proto qw/void av1_filter_intra_edge/, "uint8_t *p, int sz, int strength";
 625   specialize qw/av1_filter_intra_edge sse4_1/;
 626   add_proto qw/void av1_upsample_intra_edge/, "uint8_t *p, int sz";
 627   specialize qw/av1_upsample_intra_edge sse4_1/;
 628   if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
 629     add_proto qw/void av1_filter_intra_edge_high/, "uint16_t *p, int sz, int strength";
 630     specialize qw/av1_filter_intra_edge_high sse4_1/;
 631     add_proto qw/void av1_upsample_intra_edge_high/, "uint16_t *p, int sz, int bd";
 632     specialize qw/av1_upsample_intra_edge_high sse4_1/;
 633   }
 634 }
 635
 636 # DAALA_TX functions
 637 if (aom_config("CONFIG_DAALA_TX") eq "yes") {
 638   add_proto qw/void daala_inv_txfm_add/, "const tran_low_t *input_coeffs, void *output_pixels, int output_stride, TxfmParam *txfm_param";
 639   specialize qw/daala_inv_txfm_add avx2/;
 640 }
 641
 642 1;