vp9_rtcd_defs.pl 8.98 KB
Newer Older
Johann's avatar
Johann committed
1 2 3 4 5 6 7 8 9 10
##
##  Copyright (c) 2017 The WebM project authors. All Rights Reserved.
##
##  Use of this source code is governed by a BSD-style license
##  that can be found in the LICENSE file in the root of the source
##  tree. An additional intellectual property rights grant can be found
##  in the file PATENTS.  All contributing project authors may
##  be found in the AUTHORS file in the root of the source tree.
##

11 12 13 14 15 16 17
sub vp9_common_forward_decls() {
print <<EOF
/*
 * VP9
 */

#include "vpx/vpx_integer.h"
18
#include "vp9/common/vp9_common.h"
19
#include "vp9/common/vp9_enums.h"
20
#include "vp9/common/vp9_filter.h"
21 22 23 24 25 26

struct macroblockd;

/* Encoder forward decls */
struct macroblock;
struct vp9_variance_vtable;
27
struct search_site_config;
28 29 30 31 32 33 34
struct mv;
union int_mv;
struct yv12_buffer_config;
EOF
}
forward_decls qw/vp9_common_forward_decls/;

Johann's avatar
Johann committed
35 36
# functions that are 64 bit only.
$mmx_x86_64 = $sse2_x86_64 = $ssse3_x86_64 = $avx_x86_64 = $avx2_x86_64 = '';
37 38 39 40 41 42
if ($opts{arch} eq "x86_64") {
  $mmx_x86_64 = 'mmx';
  $sse2_x86_64 = 'sse2';
  $ssse3_x86_64 = 'ssse3';
  $avx_x86_64 = 'avx';
  $avx2_x86_64 = 'avx2';
43
  $avx512_x86_64 = 'avx512';
44 45 46 47 48 49
}

#
# post proc
#
if (vpx_config("CONFIG_VP9_POSTPROC") eq "yes") {
50
add_proto qw/void vp9_filter_by_weight16x16/, "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int src_weight";
51
specialize qw/vp9_filter_by_weight16x16 sse2 msa/;
52 53

add_proto qw/void vp9_filter_by_weight8x8/, "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int src_weight";
54
specialize qw/vp9_filter_by_weight8x8 sse2 msa/;
55 56 57 58 59
}

#
# dct
#
60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77
# Force C versions if CONFIG_EMULATE_HARDWARE is 1
add_proto qw/void vp9_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int stride, int tx_type";

add_proto qw/void vp9_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int stride, int tx_type";

add_proto qw/void vp9_iht16x16_256_add/, "const tran_low_t *input, uint8_t *output, int pitch, int tx_type";

if (vpx_config("CONFIG_EMULATE_HARDWARE") ne "yes") {
  # Note that there are more specializations appended when
  # CONFIG_VP9_HIGHBITDEPTH is off.
  specialize qw/vp9_iht4x4_16_add sse2/;
  specialize qw/vp9_iht8x8_64_add sse2/;
  specialize qw/vp9_iht16x16_256_add sse2/;
  if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") ne "yes") {
    # Note that these specializations are appended to the above ones.
    specialize qw/vp9_iht4x4_16_add neon dspr2 msa/;
    specialize qw/vp9_iht8x8_64_add neon dspr2 msa/;
    specialize qw/vp9_iht16x16_256_add dspr2 msa/;
78
  }
79
}
80

81 82
# High bitdepth functions
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
83 84 85 86 87 88 89 90 91 92 93
  #
  # post proc
  #
  if (vpx_config("CONFIG_VP9_POSTPROC") eq "yes") {
    add_proto qw/void vp9_highbd_mbpost_proc_down/, "uint16_t *dst, int pitch, int rows, int cols, int flimit";

    add_proto qw/void vp9_highbd_mbpost_proc_across_ip/, "uint16_t *src, int pitch, int rows, int cols, int flimit";

    add_proto qw/void vp9_highbd_post_proc_down_and_across/, "const uint16_t *src_ptr, uint16_t *dst_ptr, int src_pixels_per_line, int dst_pixels_per_line, int rows, int cols, int flimit";
  }

94 95 96
  #
  # dct
  #
97 98
  # Note as optimized versions of these functions are added we need to add a check to ensure
  # that when CONFIG_EMULATE_HARDWARE is on, it defaults to the C versions only.
99
  add_proto qw/void vp9_highbd_iht4x4_16_add/, "const tran_low_t *input, uint16_t *dest, int stride, int tx_type, int bd";
100

101
  add_proto qw/void vp9_highbd_iht8x8_64_add/, "const tran_low_t *input, uint16_t *dest, int stride, int tx_type, int bd";
102

103
  add_proto qw/void vp9_highbd_iht16x16_256_add/, "const tran_low_t *input, uint16_t *output, int pitch, int tx_type, int bd";
104
}
105 106 107 108 109 110 111 112

#
# Encoder functions below this point.
#
if (vpx_config("CONFIG_VP9_ENCODER") eq "yes") {

# ENCODEMB INVOKE

113 114 115 116 117
#
# Denoiser
#
if (vpx_config("CONFIG_VP9_TEMPORAL_DENOISING") eq "yes") {
  add_proto qw/int vp9_denoiser_filter/, "const uint8_t *sig, int sig_stride, const uint8_t *mc_avg, int mc_avg_stride, uint8_t *avg, int avg_stride, int increase_denoising, BLOCK_SIZE bs, int motion_magnitude";
118
  specialize qw/vp9_denoiser_filter neon sse2/;
119 120
}

121
add_proto qw/int64_t vp9_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz";
122

123
add_proto qw/int64_t vp9_block_error_fp/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, int block_size";
124

125 126 127 128 129
add_proto qw/void vp9_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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";
specialize qw/vp9_quantize_fp neon sse2/, "$ssse3_x86_64";

add_proto qw/void vp9_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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";
specialize qw/vp9_quantize_fp_32x32 neon/, "$ssse3_x86_64";
130

131 132 133 134
add_proto qw/void vp9_fdct8x8_quant/, "const int16_t *input, int stride, tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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";

if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
  specialize qw/vp9_block_error avx2 sse2/;
135

136
  specialize qw/vp9_block_error_fp avx2 sse2/;
137

138
  specialize qw/vp9_fdct8x8_quant neon ssse3/;
139 140 141

  add_proto qw/int64_t vp9_highbd_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd";
  specialize qw/vp9_highbd_block_error sse2/;
142
} else {
143
  specialize qw/vp9_block_error avx2 msa sse2/;
144

145
  specialize qw/vp9_block_error_fp neon avx2 sse2/;
146

147
  specialize qw/vp9_fdct8x8_quant sse2 ssse3 neon/;
148
}
149 150 151

# fdct functions

152
add_proto qw/void vp9_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
153

154
add_proto qw/void vp9_fht8x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
155

156
add_proto qw/void vp9_fht16x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
157

158
add_proto qw/void vp9_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
159

160 161 162 163 164 165 166 167 168 169 170 171
# Note that there are more specializations appended when CONFIG_VP9_HIGHBITDEPTH
# is off.
specialize qw/vp9_fht4x4 sse2/;
specialize qw/vp9_fht8x8 sse2/;
specialize qw/vp9_fht16x16 sse2/;
specialize qw/vp9_fwht4x4 sse2/;
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") ne "yes") {
  # Note that these specializations are appended to the above ones.
  specialize qw/vp9_fht4x4 msa/;
  specialize qw/vp9_fht8x8 msa/;
  specialize qw/vp9_fht16x16 msa/;
  specialize qw/vp9_fwht4x4 msa/;
172
}
173 174 175 176

#
# Motion search
#
177
add_proto qw/int vp9_diamond_search_sad/, "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 vp9_variance_vtable *fn_ptr, const struct mv *center_mv";
178
specialize qw/vp9_diamond_search_sad avx/;
179

180
if (vpx_config("CONFIG_REALTIME_ONLY") ne "yes") {
Johann's avatar
Johann committed
181
add_proto qw/void vp9_temporal_filter_apply/, "const uint8_t *frame1, unsigned int stride, const uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, uint32_t *accumulator, uint16_t *count";
182
specialize qw/vp9_temporal_filter_apply sse4_1/;
183
}
184

185 186
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {

187 188
  # ENCODEMB INVOKE

189
  add_proto qw/void vp9_highbd_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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";
190

191
  add_proto qw/void vp9_highbd_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *round_ptr, const int16_t *quant_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" ;
192 193

  # fdct functions
194
  add_proto qw/void vp9_highbd_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
195

196
  add_proto qw/void vp9_highbd_fht8x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
197

198
  add_proto qw/void vp9_highbd_fht16x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
199

200
  add_proto qw/void vp9_highbd_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
201

Johann's avatar
Johann committed
202
  add_proto qw/void vp9_highbd_temporal_filter_apply/, "const uint8_t *frame1, unsigned int stride, const uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, uint32_t *accumulator, uint16_t *count";
203

204
}
205
# End vp9_high encoder functions
206

207 208 209
#
# frame based scale
#
210
add_proto qw/void vp9_scale_and_extend_frame/, "const struct yv12_buffer_config *src, struct yv12_buffer_config *dst, INTERP_FILTER filter_type, int phase_scaler";
211
specialize qw/vp9_scale_and_extend_frame neon ssse3/;
212

213 214 215
}
# end encoder functions
1;