loopfilter_msa.h 17.1 KB
Newer Older
1
/*
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3
 *
4 5 6 7 8 9
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 11
 */

12 13
#ifndef VPX_DSP_LOOPFILTER_MSA_H_
#define VPX_DSP_LOOPFILTER_MSA_H_
14

Yaowu Xu's avatar
Yaowu Xu committed
15
#include "aom_dsp/mips/macros_msa.h"
16

clang-format's avatar
clang-format committed
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66
#define VPX_LPF_FILTER4_8W(p1_in, p0_in, q0_in, q1_in, mask_in, hev_in, \
                           p1_out, p0_out, q0_out, q1_out)              \
  {                                                                     \
    v16i8 p1_m, p0_m, q0_m, q1_m, q0_sub_p0, filt_sign;                 \
    v16i8 filt, filt1, filt2, cnst4b, cnst3b;                           \
    v8i16 q0_sub_p0_r, filt_r, cnst3h;                                  \
                                                                        \
    p1_m = (v16i8)__msa_xori_b(p1_in, 0x80);                            \
    p0_m = (v16i8)__msa_xori_b(p0_in, 0x80);                            \
    q0_m = (v16i8)__msa_xori_b(q0_in, 0x80);                            \
    q1_m = (v16i8)__msa_xori_b(q1_in, 0x80);                            \
                                                                        \
    filt = __msa_subs_s_b(p1_m, q1_m);                                  \
    filt = filt & (v16i8)hev_in;                                        \
    q0_sub_p0 = q0_m - p0_m;                                            \
    filt_sign = __msa_clti_s_b(filt, 0);                                \
                                                                        \
    cnst3h = __msa_ldi_h(3);                                            \
    q0_sub_p0_r = (v8i16)__msa_ilvr_b(q0_sub_p0, q0_sub_p0);            \
    q0_sub_p0_r = __msa_dotp_s_h((v16i8)q0_sub_p0_r, (v16i8)cnst3h);    \
    filt_r = (v8i16)__msa_ilvr_b(filt_sign, filt);                      \
    filt_r += q0_sub_p0_r;                                              \
    filt_r = __msa_sat_s_h(filt_r, 7);                                  \
                                                                        \
    /* combine left and right part */                                   \
    filt = __msa_pckev_b((v16i8)filt_r, (v16i8)filt_r);                 \
                                                                        \
    filt = filt & (v16i8)mask_in;                                       \
    cnst4b = __msa_ldi_b(4);                                            \
    filt1 = __msa_adds_s_b(filt, cnst4b);                               \
    filt1 >>= 3;                                                        \
                                                                        \
    cnst3b = __msa_ldi_b(3);                                            \
    filt2 = __msa_adds_s_b(filt, cnst3b);                               \
    filt2 >>= 3;                                                        \
                                                                        \
    q0_m = __msa_subs_s_b(q0_m, filt1);                                 \
    q0_out = __msa_xori_b((v16u8)q0_m, 0x80);                           \
    p0_m = __msa_adds_s_b(p0_m, filt2);                                 \
    p0_out = __msa_xori_b((v16u8)p0_m, 0x80);                           \
                                                                        \
    filt = __msa_srari_b(filt1, 1);                                     \
    hev_in = __msa_xori_b((v16u8)hev_in, 0xff);                         \
    filt = filt & (v16i8)hev_in;                                        \
                                                                        \
    q1_m = __msa_subs_s_b(q1_m, filt);                                  \
    q1_out = __msa_xori_b((v16u8)q1_m, 0x80);                           \
    p1_m = __msa_adds_s_b(p1_m, filt);                                  \
    p1_out = __msa_xori_b((v16u8)p1_m, 0x80);                           \
  }
67

clang-format's avatar
clang-format committed
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124
#define VPX_LPF_FILTER4_4W(p1_in, p0_in, q0_in, q1_in, mask_in, hev_in, \
                           p1_out, p0_out, q0_out, q1_out)              \
  {                                                                     \
    v16i8 p1_m, p0_m, q0_m, q1_m, q0_sub_p0, filt_sign;                 \
    v16i8 filt, filt1, filt2, cnst4b, cnst3b;                           \
    v8i16 q0_sub_p0_r, q0_sub_p0_l, filt_l, filt_r, cnst3h;             \
                                                                        \
    p1_m = (v16i8)__msa_xori_b(p1_in, 0x80);                            \
    p0_m = (v16i8)__msa_xori_b(p0_in, 0x80);                            \
    q0_m = (v16i8)__msa_xori_b(q0_in, 0x80);                            \
    q1_m = (v16i8)__msa_xori_b(q1_in, 0x80);                            \
                                                                        \
    filt = __msa_subs_s_b(p1_m, q1_m);                                  \
                                                                        \
    filt = filt & (v16i8)hev_in;                                        \
                                                                        \
    q0_sub_p0 = q0_m - p0_m;                                            \
    filt_sign = __msa_clti_s_b(filt, 0);                                \
                                                                        \
    cnst3h = __msa_ldi_h(3);                                            \
    q0_sub_p0_r = (v8i16)__msa_ilvr_b(q0_sub_p0, q0_sub_p0);            \
    q0_sub_p0_r = __msa_dotp_s_h((v16i8)q0_sub_p0_r, (v16i8)cnst3h);    \
    filt_r = (v8i16)__msa_ilvr_b(filt_sign, filt);                      \
    filt_r += q0_sub_p0_r;                                              \
    filt_r = __msa_sat_s_h(filt_r, 7);                                  \
                                                                        \
    q0_sub_p0_l = (v8i16)__msa_ilvl_b(q0_sub_p0, q0_sub_p0);            \
    q0_sub_p0_l = __msa_dotp_s_h((v16i8)q0_sub_p0_l, (v16i8)cnst3h);    \
    filt_l = (v8i16)__msa_ilvl_b(filt_sign, filt);                      \
    filt_l += q0_sub_p0_l;                                              \
    filt_l = __msa_sat_s_h(filt_l, 7);                                  \
                                                                        \
    filt = __msa_pckev_b((v16i8)filt_l, (v16i8)filt_r);                 \
    filt = filt & (v16i8)mask_in;                                       \
                                                                        \
    cnst4b = __msa_ldi_b(4);                                            \
    filt1 = __msa_adds_s_b(filt, cnst4b);                               \
    filt1 >>= 3;                                                        \
                                                                        \
    cnst3b = __msa_ldi_b(3);                                            \
    filt2 = __msa_adds_s_b(filt, cnst3b);                               \
    filt2 >>= 3;                                                        \
                                                                        \
    q0_m = __msa_subs_s_b(q0_m, filt1);                                 \
    q0_out = __msa_xori_b((v16u8)q0_m, 0x80);                           \
    p0_m = __msa_adds_s_b(p0_m, filt2);                                 \
    p0_out = __msa_xori_b((v16u8)p0_m, 0x80);                           \
                                                                        \
    filt = __msa_srari_b(filt1, 1);                                     \
    hev_in = __msa_xori_b((v16u8)hev_in, 0xff);                         \
    filt = filt & (v16i8)hev_in;                                        \
                                                                        \
    q1_m = __msa_subs_s_b(q1_m, filt);                                  \
    q1_out = __msa_xori_b((v16u8)q1_m, 0x80);                           \
    p1_m = __msa_adds_s_b(p1_m, filt);                                  \
    p1_out = __msa_xori_b((v16u8)p1_m, 0x80);                           \
  }
125

clang-format's avatar
clang-format committed
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
#define VPX_FLAT4(p3_in, p2_in, p0_in, q0_in, q2_in, q3_in, flat_out) \
  {                                                                   \
    v16u8 tmp, p2_a_sub_p0, q2_a_sub_q0, p3_a_sub_p0, q3_a_sub_q0;    \
    v16u8 zero_in = { 0 };                                            \
                                                                      \
    tmp = __msa_ori_b(zero_in, 1);                                    \
    p2_a_sub_p0 = __msa_asub_u_b(p2_in, p0_in);                       \
    q2_a_sub_q0 = __msa_asub_u_b(q2_in, q0_in);                       \
    p3_a_sub_p0 = __msa_asub_u_b(p3_in, p0_in);                       \
    q3_a_sub_q0 = __msa_asub_u_b(q3_in, q0_in);                       \
                                                                      \
    p2_a_sub_p0 = __msa_max_u_b(p2_a_sub_p0, q2_a_sub_q0);            \
    flat_out = __msa_max_u_b(p2_a_sub_p0, flat_out);                  \
    p3_a_sub_p0 = __msa_max_u_b(p3_a_sub_p0, q3_a_sub_q0);            \
    flat_out = __msa_max_u_b(p3_a_sub_p0, flat_out);                  \
                                                                      \
    flat_out = (tmp < (v16u8)flat_out);                               \
    flat_out = __msa_xori_b(flat_out, 0xff);                          \
    flat_out = flat_out & (mask);                                     \
  }
146

clang-format's avatar
clang-format committed
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
#define VPX_FLAT5(p7_in, p6_in, p5_in, p4_in, p0_in, q0_in, q4_in, q5_in, \
                  q6_in, q7_in, flat_in, flat2_out)                       \
  {                                                                       \
    v16u8 tmp, zero_in = { 0 };                                           \
    v16u8 p4_a_sub_p0, q4_a_sub_q0, p5_a_sub_p0, q5_a_sub_q0;             \
    v16u8 p6_a_sub_p0, q6_a_sub_q0, p7_a_sub_p0, q7_a_sub_q0;             \
                                                                          \
    tmp = __msa_ori_b(zero_in, 1);                                        \
    p4_a_sub_p0 = __msa_asub_u_b(p4_in, p0_in);                           \
    q4_a_sub_q0 = __msa_asub_u_b(q4_in, q0_in);                           \
    p5_a_sub_p0 = __msa_asub_u_b(p5_in, p0_in);                           \
    q5_a_sub_q0 = __msa_asub_u_b(q5_in, q0_in);                           \
    p6_a_sub_p0 = __msa_asub_u_b(p6_in, p0_in);                           \
    q6_a_sub_q0 = __msa_asub_u_b(q6_in, q0_in);                           \
    p7_a_sub_p0 = __msa_asub_u_b(p7_in, p0_in);                           \
    q7_a_sub_q0 = __msa_asub_u_b(q7_in, q0_in);                           \
                                                                          \
    p4_a_sub_p0 = __msa_max_u_b(p4_a_sub_p0, q4_a_sub_q0);                \
    flat2_out = __msa_max_u_b(p5_a_sub_p0, q5_a_sub_q0);                  \
    flat2_out = __msa_max_u_b(p4_a_sub_p0, flat2_out);                    \
    p6_a_sub_p0 = __msa_max_u_b(p6_a_sub_p0, q6_a_sub_q0);                \
    flat2_out = __msa_max_u_b(p6_a_sub_p0, flat2_out);                    \
    p7_a_sub_p0 = __msa_max_u_b(p7_a_sub_p0, q7_a_sub_q0);                \
    flat2_out = __msa_max_u_b(p7_a_sub_p0, flat2_out);                    \
                                                                          \
    flat2_out = (tmp < (v16u8)flat2_out);                                 \
    flat2_out = __msa_xori_b(flat2_out, 0xff);                            \
    flat2_out = flat2_out & flat_in;                                      \
  }
176

clang-format's avatar
clang-format committed
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213
#define VPX_FILTER8(p3_in, p2_in, p1_in, p0_in, q0_in, q1_in, q2_in, q3_in, \
                    p2_filt8_out, p1_filt8_out, p0_filt8_out, q0_filt8_out, \
                    q1_filt8_out, q2_filt8_out)                             \
  {                                                                         \
    v8u16 tmp0, tmp1, tmp2;                                                 \
                                                                            \
    tmp2 = p2_in + p1_in + p0_in;                                           \
    tmp0 = p3_in << 1;                                                      \
                                                                            \
    tmp0 = tmp0 + tmp2 + q0_in;                                             \
    tmp1 = tmp0 + p3_in + p2_in;                                            \
    p2_filt8_out = (v8i16)__msa_srari_h((v8i16)tmp1, 3);                    \
                                                                            \
    tmp1 = tmp0 + p1_in + q1_in;                                            \
    p1_filt8_out = (v8i16)__msa_srari_h((v8i16)tmp1, 3);                    \
                                                                            \
    tmp1 = q2_in + q1_in + q0_in;                                           \
    tmp2 = tmp2 + tmp1;                                                     \
    tmp0 = tmp2 + (p0_in);                                                  \
    tmp0 = tmp0 + (p3_in);                                                  \
    p0_filt8_out = (v8i16)__msa_srari_h((v8i16)tmp0, 3);                    \
                                                                            \
    tmp0 = q2_in + q3_in;                                                   \
    tmp0 = p0_in + tmp1 + tmp0;                                             \
    tmp1 = q3_in + q3_in;                                                   \
    tmp1 = tmp1 + tmp0;                                                     \
    q2_filt8_out = (v8i16)__msa_srari_h((v8i16)tmp1, 3);                    \
                                                                            \
    tmp0 = tmp2 + q3_in;                                                    \
    tmp1 = tmp0 + q0_in;                                                    \
    q0_filt8_out = (v8i16)__msa_srari_h((v8i16)tmp1, 3);                    \
                                                                            \
    tmp1 = tmp0 - p2_in;                                                    \
    tmp0 = q1_in + q3_in;                                                   \
    tmp1 = tmp0 + tmp1;                                                     \
    q1_filt8_out = (v8i16)__msa_srari_h((v8i16)tmp1, 3);                    \
  }
214

clang-format's avatar
clang-format committed
215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
#define LPF_MASK_HEV(p3_in, p2_in, p1_in, p0_in, q0_in, q1_in, q2_in, q3_in, \
                     limit_in, b_limit_in, thresh_in, hev_out, mask_out,     \
                     flat_out)                                               \
  {                                                                          \
    v16u8 p3_asub_p2_m, p2_asub_p1_m, p1_asub_p0_m, q1_asub_q0_m;            \
    v16u8 p1_asub_q1_m, p0_asub_q0_m, q3_asub_q2_m, q2_asub_q1_m;            \
                                                                             \
    /* absolute subtraction of pixel values */                               \
    p3_asub_p2_m = __msa_asub_u_b(p3_in, p2_in);                             \
    p2_asub_p1_m = __msa_asub_u_b(p2_in, p1_in);                             \
    p1_asub_p0_m = __msa_asub_u_b(p1_in, p0_in);                             \
    q1_asub_q0_m = __msa_asub_u_b(q1_in, q0_in);                             \
    q2_asub_q1_m = __msa_asub_u_b(q2_in, q1_in);                             \
    q3_asub_q2_m = __msa_asub_u_b(q3_in, q2_in);                             \
    p0_asub_q0_m = __msa_asub_u_b(p0_in, q0_in);                             \
    p1_asub_q1_m = __msa_asub_u_b(p1_in, q1_in);                             \
                                                                             \
    /* calculation of hev */                                                 \
    flat_out = __msa_max_u_b(p1_asub_p0_m, q1_asub_q0_m);                    \
    hev_out = thresh_in < (v16u8)flat_out;                                   \
                                                                             \
    /* calculation of mask */                                                \
    p0_asub_q0_m = __msa_adds_u_b(p0_asub_q0_m, p0_asub_q0_m);               \
    p1_asub_q1_m >>= 1;                                                      \
    p0_asub_q0_m = __msa_adds_u_b(p0_asub_q0_m, p1_asub_q1_m);               \
                                                                             \
    mask_out = b_limit_in < p0_asub_q0_m;                                    \
    mask_out = __msa_max_u_b(flat_out, mask_out);                            \
    p3_asub_p2_m = __msa_max_u_b(p3_asub_p2_m, p2_asub_p1_m);                \
    mask_out = __msa_max_u_b(p3_asub_p2_m, mask_out);                        \
    q2_asub_q1_m = __msa_max_u_b(q2_asub_q1_m, q3_asub_q2_m);                \
    mask_out = __msa_max_u_b(q2_asub_q1_m, mask_out);                        \
                                                                             \
    mask_out = limit_in < (v16u8)mask_out;                                   \
    mask_out = __msa_xori_b(mask_out, 0xff);                                 \
  }
#endif /* VPX_DSP_LOOPFILTER_MSA_H_ */