CoCalc -- mips

GitHub Repository: Tetragramm/opencv
Path: blob/master/3rdparty/libwebp/src/dsp/mips_macro.h
¹⁶³⁴⁸ views
1
// Copyright 2014 Google Inc. All Rights Reserved.
2
//
3
// Use of this source code is governed by a BSD-style license
4
// that can be found in the COPYING file in the root of the source
5
// tree. An additional intellectual property rights grant can be found
6
// in the file PATENTS. All contributing project authors may
7
// be found in the AUTHORS file in the root of the source tree.
8
// -----------------------------------------------------------------------------
9
//
10
// MIPS common macros
11

12
#ifndef WEBP_DSP_MIPS_MACRO_H_
13
#define WEBP_DSP_MIPS_MACRO_H_
14

15
#if defined(__GNUC__) && defined(__ANDROID__) && LOCAL_GCC_VERSION == 0x409
16
#define WORK_AROUND_GCC
17
#endif
18

19
#define STR(s) #s
20
#define XSTR(s) STR(s)
21

22
// O0[31..16 | 15..0] = I0[31..16 | 15..0] + I1[31..16 | 15..0]
23
// O1[31..16 | 15..0] = I0[31..16 | 15..0] - I1[31..16 | 15..0]
24
// O - output
25
// I - input (macro doesn't change it)
26
#define ADD_SUB_HALVES(O0, O1,                                                 \
27
                       I0, I1)                                                 \
28
  "addq.ph          %[" #O0 "],   %[" #I0 "],  %[" #I1 "]           \n\t"      \
29
  "subq.ph          %[" #O1 "],   %[" #I0 "],  %[" #I1 "]           \n\t"
30

31
// O - output
32
// I - input (macro doesn't change it)
33
// I[0/1] - offset in bytes
34
#define LOAD_IN_X2(O0, O1,                                                     \
35
                   I0, I1)                                                     \
36
  "lh               %[" #O0 "],   " #I0 "(%[in])                  \n\t"        \
37
  "lh               %[" #O1 "],   " #I1 "(%[in])                  \n\t"
38

39
// I0 - location
40
// I1..I9 - offsets in bytes
41
#define LOAD_WITH_OFFSET_X4(O0, O1, O2, O3,                                    \
42
                            I0, I1, I2, I3, I4, I5, I6, I7, I8, I9)            \
43
  "ulw    %[" #O0 "],    " #I1 "+" XSTR(I9) "*" #I5 "(%[" #I0 "])       \n\t"  \
44
  "ulw    %[" #O1 "],    " #I2 "+" XSTR(I9) "*" #I6 "(%[" #I0 "])       \n\t"  \
45
  "ulw    %[" #O2 "],    " #I3 "+" XSTR(I9) "*" #I7 "(%[" #I0 "])       \n\t"  \
46
  "ulw    %[" #O3 "],    " #I4 "+" XSTR(I9) "*" #I8 "(%[" #I0 "])       \n\t"
47

48
// O - output
49
// IO - input/output
50
// I - input (macro doesn't change it)
51
#define MUL_SHIFT_SUM(O0, O1, O2, O3, O4, O5, O6, O7,                          \
52
                      IO0, IO1, IO2, IO3,                                      \
53
                      I0, I1, I2, I3, I4, I5, I6, I7)                          \
54
  "mul              %[" #O0 "],   %[" #I0 "],   %[kC2]        \n\t"            \
55
  "mul              %[" #O1 "],   %[" #I0 "],   %[kC1]        \n\t"            \
56
  "mul              %[" #O2 "],   %[" #I1 "],   %[kC2]        \n\t"            \
57
  "mul              %[" #O3 "],   %[" #I1 "],   %[kC1]        \n\t"            \
58
  "mul              %[" #O4 "],   %[" #I2 "],   %[kC2]        \n\t"            \
59
  "mul              %[" #O5 "],   %[" #I2 "],   %[kC1]        \n\t"            \
60
  "mul              %[" #O6 "],   %[" #I3 "],   %[kC2]        \n\t"            \
61
  "mul              %[" #O7 "],   %[" #I3 "],   %[kC1]        \n\t"            \
62
  "sra              %[" #O0 "],   %[" #O0 "],   16            \n\t"            \
63
  "sra              %[" #O1 "],   %[" #O1 "],   16            \n\t"            \
64
  "sra              %[" #O2 "],   %[" #O2 "],   16            \n\t"            \
65
  "sra              %[" #O3 "],   %[" #O3 "],   16            \n\t"            \
66
  "sra              %[" #O4 "],   %[" #O4 "],   16            \n\t"            \
67
  "sra              %[" #O5 "],   %[" #O5 "],   16            \n\t"            \
68
  "sra              %[" #O6 "],   %[" #O6 "],   16            \n\t"            \
69
  "sra              %[" #O7 "],   %[" #O7 "],   16            \n\t"            \
70
  "addu             %[" #IO0 "],  %[" #IO0 "],  %[" #I4 "]    \n\t"            \
71
  "addu             %[" #IO1 "],  %[" #IO1 "],  %[" #I5 "]    \n\t"            \
72
  "subu             %[" #IO2 "],  %[" #IO2 "],  %[" #I6 "]    \n\t"            \
73
  "subu             %[" #IO3 "],  %[" #IO3 "],  %[" #I7 "]    \n\t"
74

75
// O - output
76
// I - input (macro doesn't change it)
77
#define INSERT_HALF_X2(O0, O1,                                                 \
78
                       I0, I1)                                                 \
79
  "ins              %[" #O0 "],   %[" #I0 "], 16,    16           \n\t"        \
80
  "ins              %[" #O1 "],   %[" #I1 "], 16,    16           \n\t"
81

82
// O - output
83
// I - input (macro doesn't change it)
84
#define SRA_16(O0, O1, O2, O3,                                                 \
85
               I0, I1, I2, I3)                                                 \
86
  "sra              %[" #O0 "],  %[" #I0 "],  16                  \n\t"        \
87
  "sra              %[" #O1 "],  %[" #I1 "],  16                  \n\t"        \
88
  "sra              %[" #O2 "],  %[" #I2 "],  16                  \n\t"        \
89
  "sra              %[" #O3 "],  %[" #I3 "],  16                  \n\t"
90

91
// temp0[31..16 | 15..0] = temp8[31..16 | 15..0] + temp12[31..16 | 15..0]
92
// temp1[31..16 | 15..0] = temp8[31..16 | 15..0] - temp12[31..16 | 15..0]
93
// temp0[31..16 | 15..0] = temp0[31..16 >> 3 | 15..0 >> 3]
94
// temp1[31..16 | 15..0] = temp1[31..16 >> 3 | 15..0 >> 3]
95
// O - output
96
// I - input (macro doesn't change it)
97
#define SHIFT_R_SUM_X2(O0, O1, O2, O3, O4, O5, O6, O7,                         \
98
                       I0, I1, I2, I3, I4, I5, I6, I7)                         \
99
  "addq.ph          %[" #O0 "],   %[" #I0 "],   %[" #I4 "]    \n\t"            \
100
  "subq.ph          %[" #O1 "],   %[" #I0 "],   %[" #I4 "]    \n\t"            \
101
  "addq.ph          %[" #O2 "],   %[" #I1 "],   %[" #I5 "]    \n\t"            \
102
  "subq.ph          %[" #O3 "],   %[" #I1 "],   %[" #I5 "]    \n\t"            \
103
  "addq.ph          %[" #O4 "],   %[" #I2 "],   %[" #I6 "]    \n\t"            \
104
  "subq.ph          %[" #O5 "],   %[" #I2 "],   %[" #I6 "]    \n\t"            \
105
  "addq.ph          %[" #O6 "],   %[" #I3 "],   %[" #I7 "]    \n\t"            \
106
  "subq.ph          %[" #O7 "],   %[" #I3 "],   %[" #I7 "]    \n\t"            \
107
  "shra.ph          %[" #O0 "],   %[" #O0 "],   3             \n\t"            \
108
  "shra.ph          %[" #O1 "],   %[" #O1 "],   3             \n\t"            \
109
  "shra.ph          %[" #O2 "],   %[" #O2 "],   3             \n\t"            \
110
  "shra.ph          %[" #O3 "],   %[" #O3 "],   3             \n\t"            \
111
  "shra.ph          %[" #O4 "],   %[" #O4 "],   3             \n\t"            \
112
  "shra.ph          %[" #O5 "],   %[" #O5 "],   3             \n\t"            \
113
  "shra.ph          %[" #O6 "],   %[" #O6 "],   3             \n\t"            \
114
  "shra.ph          %[" #O7 "],   %[" #O7 "],   3             \n\t"
115

116
// precrq.ph.w temp0, temp8, temp2
117
//   temp0 = temp8[31..16] | temp2[31..16]
118
// ins temp2, temp8, 16, 16
119
//   temp2 = temp8[31..16] | temp2[15..0]
120
// O - output
121
// IO - input/output
122
// I - input (macro doesn't change it)
123
#define PACK_2_HALVES_TO_WORD(O0, O1, O2, O3,                                  \
124
                              IO0, IO1, IO2, IO3,                              \
125
                              I0, I1, I2, I3)                                  \
126
  "precrq.ph.w      %[" #O0 "],    %[" #I0 "],  %[" #IO0 "]       \n\t"        \
127
  "precrq.ph.w      %[" #O1 "],    %[" #I1 "],  %[" #IO1 "]       \n\t"        \
128
  "ins              %[" #IO0 "],   %[" #I0 "],  16,    16         \n\t"        \
129
  "ins              %[" #IO1 "],   %[" #I1 "],  16,    16         \n\t"        \
130
  "precrq.ph.w      %[" #O2 "],    %[" #I2 "],  %[" #IO2 "]       \n\t"        \
131
  "precrq.ph.w      %[" #O3 "],    %[" #I3 "],  %[" #IO3 "]       \n\t"        \
132
  "ins              %[" #IO2 "],   %[" #I2 "],  16,    16         \n\t"        \
133
  "ins              %[" #IO3 "],   %[" #I3 "],  16,    16         \n\t"
134

135
// preceu.ph.qbr temp0, temp8
136
//   temp0 = 0 | 0 | temp8[23..16] | temp8[7..0]
137
// preceu.ph.qbl temp1, temp8
138
//   temp1 = temp8[23..16] | temp8[7..0] | 0 | 0
139
// O - output
140
// I - input (macro doesn't change it)
141
#define CONVERT_2_BYTES_TO_HALF(O0, O1, O2, O3, O4, O5, O6, O7,                \
142
                                I0, I1, I2, I3)                                \
143
  "preceu.ph.qbr    %[" #O0 "],   %[" #I0 "]                      \n\t"        \
144
  "preceu.ph.qbl    %[" #O1 "],   %[" #I0 "]                      \n\t"        \
145
  "preceu.ph.qbr    %[" #O2 "],   %[" #I1 "]                      \n\t"        \
146
  "preceu.ph.qbl    %[" #O3 "],   %[" #I1 "]                      \n\t"        \
147
  "preceu.ph.qbr    %[" #O4 "],   %[" #I2 "]                      \n\t"        \
148
  "preceu.ph.qbl    %[" #O5 "],   %[" #I2 "]                      \n\t"        \
149
  "preceu.ph.qbr    %[" #O6 "],   %[" #I3 "]                      \n\t"        \
150
  "preceu.ph.qbl    %[" #O7 "],   %[" #I3 "]                      \n\t"
151

152
// temp0[31..16 | 15..0] = temp0[31..16 | 15..0] + temp8[31..16 | 15..0]
153
// temp0[31..16 | 15..0] = temp0[31..16 <<(s) 7 | 15..0 <<(s) 7]
154
// temp1..temp7 same as temp0
155
// precrqu_s.qb.ph temp0, temp1, temp0:
156
//   temp0 = temp1[31..24] | temp1[15..8] | temp0[31..24] | temp0[15..8]
157
// store temp0 to dst
158
// IO - input/output
159
// I - input (macro doesn't change it)
160
#define STORE_SAT_SUM_X2(IO0, IO1, IO2, IO3, IO4, IO5, IO6, IO7,               \
161
                         I0, I1, I2, I3, I4, I5, I6, I7,                       \
162
                         I8, I9, I10, I11, I12, I13)                           \
163
  "addq.ph          %[" #IO0 "],  %[" #IO0 "],  %[" #I0 "]          \n\t"      \
164
  "addq.ph          %[" #IO1 "],  %[" #IO1 "],  %[" #I1 "]          \n\t"      \
165
  "addq.ph          %[" #IO2 "],  %[" #IO2 "],  %[" #I2 "]          \n\t"      \
166
  "addq.ph          %[" #IO3 "],  %[" #IO3 "],  %[" #I3 "]          \n\t"      \
167
  "addq.ph          %[" #IO4 "],  %[" #IO4 "],  %[" #I4 "]          \n\t"      \
168
  "addq.ph          %[" #IO5 "],  %[" #IO5 "],  %[" #I5 "]          \n\t"      \
169
  "addq.ph          %[" #IO6 "],  %[" #IO6 "],  %[" #I6 "]          \n\t"      \
170
  "addq.ph          %[" #IO7 "],  %[" #IO7 "],  %[" #I7 "]          \n\t"      \
171
  "shll_s.ph        %[" #IO0 "],  %[" #IO0 "],  7                   \n\t"      \
172
  "shll_s.ph        %[" #IO1 "],  %[" #IO1 "],  7                   \n\t"      \
173
  "shll_s.ph        %[" #IO2 "],  %[" #IO2 "],  7                   \n\t"      \
174
  "shll_s.ph        %[" #IO3 "],  %[" #IO3 "],  7                   \n\t"      \
175
  "shll_s.ph        %[" #IO4 "],  %[" #IO4 "],  7                   \n\t"      \
176
  "shll_s.ph        %[" #IO5 "],  %[" #IO5 "],  7                   \n\t"      \
177
  "shll_s.ph        %[" #IO6 "],  %[" #IO6 "],  7                   \n\t"      \
178
  "shll_s.ph        %[" #IO7 "],  %[" #IO7 "],  7                   \n\t"      \
179
  "precrqu_s.qb.ph  %[" #IO0 "],  %[" #IO1 "],  %[" #IO0 "]         \n\t"      \
180
  "precrqu_s.qb.ph  %[" #IO2 "],  %[" #IO3 "],  %[" #IO2 "]         \n\t"      \
181
  "precrqu_s.qb.ph  %[" #IO4 "],  %[" #IO5 "],  %[" #IO4 "]         \n\t"      \
182
  "precrqu_s.qb.ph  %[" #IO6 "],  %[" #IO7 "],  %[" #IO6 "]         \n\t"      \
183
  "usw              %[" #IO0 "],  " XSTR(I13) "*" #I9 "(%[" #I8 "])   \n\t"    \
184
  "usw              %[" #IO2 "],  " XSTR(I13) "*" #I10 "(%[" #I8 "])  \n\t"    \
185
  "usw              %[" #IO4 "],  " XSTR(I13) "*" #I11 "(%[" #I8 "])  \n\t"    \
186
  "usw              %[" #IO6 "],  " XSTR(I13) "*" #I12 "(%[" #I8 "])  \n\t"
187

188
#define OUTPUT_EARLY_CLOBBER_REGS_10()                                         \
189
  : [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), [temp3]"=&r"(temp3),             \
190
    [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), [temp6]"=&r"(temp6),             \
191
    [temp7]"=&r"(temp7), [temp8]"=&r"(temp8), [temp9]"=&r"(temp9),             \
192
    [temp10]"=&r"(temp10)
193

194
#define OUTPUT_EARLY_CLOBBER_REGS_18()                                         \
195
  OUTPUT_EARLY_CLOBBER_REGS_10(),                                              \
196
  [temp11]"=&r"(temp11), [temp12]"=&r"(temp12), [temp13]"=&r"(temp13),         \
197
  [temp14]"=&r"(temp14), [temp15]"=&r"(temp15), [temp16]"=&r"(temp16),         \
198
  [temp17]"=&r"(temp17), [temp18]"=&r"(temp18)
199

200
#endif  // WEBP_DSP_MIPS_MACRO_H_
201

202
Product

Resources

Company
