1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
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
67
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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
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
176
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
214
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
|
/* -*- c++ -*- */
/*
* Copyright © 2010-2015 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*/
#ifndef BRW_IR_FS_H
#define BRW_IR_FS_H
#include "brw_shader.h"
class fs_inst;
class fs_reg : public backend_reg {
public:
DECLARE_RALLOC_CXX_OPERATORS(fs_reg)
void init();
fs_reg();
explicit fs_reg(float f);
explicit fs_reg(int32_t i);
explicit fs_reg(uint32_t u);
explicit fs_reg(uint8_t vf[4]);
explicit fs_reg(uint8_t vf0, uint8_t vf1, uint8_t vf2, uint8_t vf3);
fs_reg(struct brw_reg fixed_hw_reg);
fs_reg(enum register_file file, int reg);
fs_reg(enum register_file file, int reg, enum brw_reg_type type);
fs_reg(enum register_file file, int reg, enum brw_reg_type type, uint8_t width);
bool equals(const fs_reg &r) const;
bool is_contiguous() const;
/** Smear a channel of the reg to all channels. */
fs_reg &set_smear(unsigned subreg);
/**
* Offset in bytes from the start of the register. Values up to a
* backend_reg::reg_offset unit are valid.
*/
int subreg_offset;
fs_reg *reladdr;
/**
* The register width. This indicates how many hardware values are
* represented by each virtual value. Valid values are 1, 8, or 16.
* For immediate values, this is 1. Most of the rest of the time, it
* will be equal to the dispatch width.
*/
uint8_t width;
/**
* Returns the effective register width when used as a source in the
* given instruction. Registers such as uniforms and immediates
* effectively take on the width of the instruction in which they are
* used.
*/
uint8_t effective_width;
/** Register region horizontal stride */
uint8_t stride;
};
static inline fs_reg
negate(fs_reg reg)
{
assert(reg.file != HW_REG && reg.file != IMM);
reg.negate = !reg.negate;
return reg;
}
static inline fs_reg
retype(fs_reg reg, enum brw_reg_type type)
{
reg.fixed_hw_reg.type = reg.type = type;
return reg;
}
static inline fs_reg
byte_offset(fs_reg reg, unsigned delta)
{
switch (reg.file) {
case BAD_FILE:
break;
case GRF:
case ATTR:
reg.reg_offset += delta / 32;
break;
case MRF:
reg.reg += delta / 32;
break;
default:
assert(delta == 0);
}
reg.subreg_offset += delta % 32;
return reg;
}
static inline fs_reg
horiz_offset(fs_reg reg, unsigned delta)
{
switch (reg.file) {
case BAD_FILE:
case UNIFORM:
case IMM:
/* These only have a single component that is implicitly splatted. A
* horizontal offset should be a harmless no-op.
*/
break;
case GRF:
case MRF:
case ATTR:
return byte_offset(reg, delta * reg.stride * type_sz(reg.type));
default:
assert(delta == 0);
}
return reg;
}
static inline fs_reg
offset(fs_reg reg, unsigned delta)
{
assert(reg.stride > 0);
switch (reg.file) {
case BAD_FILE:
break;
case GRF:
case MRF:
case ATTR:
return byte_offset(reg, delta * reg.width * reg.stride * type_sz(reg.type));
case UNIFORM:
reg.reg_offset += delta;
break;
default:
assert(delta == 0);
}
return reg;
}
static inline fs_reg
component(fs_reg reg, unsigned idx)
{
assert(reg.subreg_offset == 0);
assert(idx < reg.width);
reg.subreg_offset = idx * type_sz(reg.type);
reg.width = 1;
return reg;
}
/**
* Get either of the 8-component halves of a 16-component register.
*
* Note: this also works if \c reg represents a SIMD16 pair of registers.
*/
static inline fs_reg
half(fs_reg reg, unsigned idx)
{
assert(idx < 2);
if (reg.file == UNIFORM)
return reg;
assert(idx == 0 || (reg.file != HW_REG && reg.file != IMM));
assert(reg.width == 16);
reg.width = 8;
return horiz_offset(reg, 8 * idx);
}
static const fs_reg reg_undef;
class fs_inst : public backend_instruction {
fs_inst &operator=(const fs_inst &);
void init(enum opcode opcode, uint8_t exec_width, const fs_reg &dst,
const fs_reg *src, unsigned sources);
public:
DECLARE_RALLOC_CXX_OPERATORS(fs_inst)
fs_inst();
fs_inst(enum opcode opcode, uint8_t exec_size);
fs_inst(enum opcode opcode, const fs_reg &dst);
fs_inst(enum opcode opcode, uint8_t exec_size, const fs_reg &dst,
const fs_reg &src0);
fs_inst(enum opcode opcode, const fs_reg &dst, const fs_reg &src0);
fs_inst(enum opcode opcode, uint8_t exec_size, const fs_reg &dst,
const fs_reg &src0, const fs_reg &src1);
fs_inst(enum opcode opcode, const fs_reg &dst, const fs_reg &src0,
const fs_reg &src1);
fs_inst(enum opcode opcode, uint8_t exec_size, const fs_reg &dst,
const fs_reg &src0, const fs_reg &src1, const fs_reg &src2);
fs_inst(enum opcode opcode, const fs_reg &dst, const fs_reg &src0,
const fs_reg &src1, const fs_reg &src2);
fs_inst(enum opcode opcode, const fs_reg &dst, const fs_reg src[],
unsigned sources);
fs_inst(enum opcode opcode, uint8_t exec_size, const fs_reg &dst,
const fs_reg src[], unsigned sources);
fs_inst(const fs_inst &that);
~fs_inst();
void resize_sources(uint8_t num_sources);
bool equals(fs_inst *inst) const;
bool overwrites_reg(const fs_reg ®) const;
bool is_send_from_grf() const;
bool is_partial_write() const;
int regs_read(int arg) const;
bool can_do_source_mods(struct brw_context *brw);
bool has_side_effects() const;
bool reads_flag() const;
bool writes_flag() const;
fs_reg dst;
fs_reg *src;
uint8_t sources; /**< Number of fs_reg sources. */
/**
* Execution size of the instruction. This is used by the generator to
* generate the correct binary for the given fs_inst. Current valid
* values are 1, 8, 16.
*/
uint8_t exec_size;
bool eot:1;
bool force_uncompressed:1;
bool force_sechalf:1;
bool pi_noperspective:1; /**< Pixel interpolator noperspective flag */
};
#endif
|