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
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
|
/*
* Copyright © 2015-2016 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.
*/
#include "brw_compiler.h"
#include "brw_shader.h"
#include "brw_eu.h"
#include "dev/gen_debug.h"
#include "compiler/nir/nir.h"
#include "main/errors.h"
#include "util/debug.h"
#define COMMON_OPTIONS \
.lower_sub = true, \
.lower_fdiv = true, \
.lower_scmp = true, \
.lower_flrp16 = true, \
.lower_fmod = true, \
.lower_bitfield_extract = true, \
.lower_bitfield_insert = true, \
.lower_uadd_carry = true, \
.lower_usub_borrow = true, \
.lower_fdiv = true, \
.lower_flrp64 = true, \
.lower_isign = true, \
.lower_ldexp = true, \
.lower_device_index_to_zero = true, \
.vectorize_io = true, \
.use_interpolated_input_intrinsics = true, \
.vertex_id_zero_based = true, \
.lower_base_vertex = true
#define COMMON_SCALAR_OPTIONS \
.lower_to_scalar = true, \
.lower_pack_half_2x16 = true, \
.lower_pack_snorm_2x16 = true, \
.lower_pack_snorm_4x8 = true, \
.lower_pack_unorm_2x16 = true, \
.lower_pack_unorm_4x8 = true, \
.lower_unpack_half_2x16 = true, \
.lower_unpack_snorm_2x16 = true, \
.lower_unpack_snorm_4x8 = true, \
.lower_unpack_unorm_2x16 = true, \
.lower_unpack_unorm_4x8 = true, \
.lower_usub_sat64 = true, \
.lower_hadd64 = true, \
.max_unroll_iterations = 32
static const struct nir_shader_compiler_options scalar_nir_options = {
COMMON_OPTIONS,
COMMON_SCALAR_OPTIONS,
};
static const struct nir_shader_compiler_options vector_nir_options = {
COMMON_OPTIONS,
/* In the vec4 backend, our dpN instruction replicates its result to all the
* components of a vec4. We would like NIR to give us replicated fdot
* instructions because it can optimize better for us.
*/
.fdot_replicates = true,
.lower_pack_snorm_2x16 = true,
.lower_pack_unorm_2x16 = true,
.lower_unpack_snorm_2x16 = true,
.lower_unpack_unorm_2x16 = true,
.lower_extract_byte = true,
.lower_extract_word = true,
.intel_vec4 = true,
.max_unroll_iterations = 32,
};
struct brw_compiler *
brw_compiler_create(void *mem_ctx, const struct gen_device_info *devinfo)
{
struct brw_compiler *compiler = rzalloc(mem_ctx, struct brw_compiler);
compiler->devinfo = devinfo;
brw_fs_alloc_reg_sets(compiler);
brw_vec4_alloc_reg_set(compiler);
brw_init_compaction_tables(devinfo);
compiler->precise_trig = env_var_as_boolean("INTEL_PRECISE_TRIG", false);
compiler->use_tcs_8_patch =
devinfo->gen >= 12 ||
(devinfo->gen >= 9 && (INTEL_DEBUG & DEBUG_TCS_EIGHT_PATCH));
if (devinfo->gen >= 10) {
/* We don't support vec4 mode on Cannonlake. */
for (int i = MESA_SHADER_VERTEX; i < MESA_SHADER_STAGES; i++)
compiler->scalar_stage[i] = true;
} else {
compiler->scalar_stage[MESA_SHADER_VERTEX] =
devinfo->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_VS", true);
compiler->scalar_stage[MESA_SHADER_TESS_CTRL] =
devinfo->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_TCS", true);
compiler->scalar_stage[MESA_SHADER_TESS_EVAL] =
devinfo->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_TES", true);
compiler->scalar_stage[MESA_SHADER_GEOMETRY] =
devinfo->gen >= 8 && env_var_as_boolean("INTEL_SCALAR_GS", true);
compiler->scalar_stage[MESA_SHADER_FRAGMENT] = true;
compiler->scalar_stage[MESA_SHADER_COMPUTE] = true;
}
nir_lower_int64_options int64_options =
nir_lower_imul64 |
nir_lower_isign64 |
nir_lower_divmod64 |
nir_lower_imul_high64;
nir_lower_doubles_options fp64_options =
nir_lower_drcp |
nir_lower_dsqrt |
nir_lower_drsq |
nir_lower_dtrunc |
nir_lower_dfloor |
nir_lower_dceil |
nir_lower_dfract |
nir_lower_dround_even |
nir_lower_dmod |
nir_lower_dsub |
nir_lower_ddiv;
if (!devinfo->has_64bit_float || (INTEL_DEBUG & DEBUG_SOFT64)) {
int64_options |= nir_lower_mov64 |
nir_lower_icmp64 |
nir_lower_iadd64 |
nir_lower_iabs64 |
nir_lower_ineg64 |
nir_lower_logic64 |
nir_lower_minmax64 |
nir_lower_shift64 |
nir_lower_extract64;
fp64_options |= nir_lower_fp64_full_software;
}
/* The Bspec's section tittled "Instruction_multiply[DevBDW+]" claims that
* destination type can be Quadword and source type Doubleword for Gen8 and
* Gen9. So, lower 64 bit multiply instruction on rest of the platforms.
*/
if (devinfo->gen < 8 || devinfo->gen > 9)
int64_options |= nir_lower_imul_2x32_64;
/* We want the GLSL compiler to emit code that uses condition codes */
for (int i = 0; i < MESA_SHADER_STAGES; i++) {
compiler->glsl_compiler_options[i].MaxUnrollIterations = 0;
compiler->glsl_compiler_options[i].MaxIfDepth =
devinfo->gen < 6 ? 16 : UINT_MAX;
compiler->glsl_compiler_options[i].EmitNoIndirectInput = true;
compiler->glsl_compiler_options[i].EmitNoIndirectUniform = false;
bool is_scalar = compiler->scalar_stage[i];
compiler->glsl_compiler_options[i].EmitNoIndirectOutput = is_scalar;
compiler->glsl_compiler_options[i].EmitNoIndirectTemp = is_scalar;
compiler->glsl_compiler_options[i].OptimizeForAOS = !is_scalar;
struct nir_shader_compiler_options *nir_options =
rzalloc(compiler, struct nir_shader_compiler_options);
if (is_scalar) {
*nir_options = scalar_nir_options;
} else {
*nir_options = vector_nir_options;
}
/* Prior to Gen6, there are no three source operations, and Gen11 loses
* LRP.
*/
nir_options->lower_ffma = devinfo->gen < 6;
nir_options->lower_flrp32 = devinfo->gen < 6 || devinfo->gen >= 11;
nir_options->lower_fpow = devinfo->gen >= 12;
nir_options->lower_rotate = devinfo->gen < 11;
nir_options->lower_bitfield_reverse = devinfo->gen < 7;
nir_options->lower_int64_options = int64_options;
nir_options->lower_doubles_options = fp64_options;
nir_options->unify_interfaces = i < MESA_SHADER_FRAGMENT;
compiler->glsl_compiler_options[i].NirOptions = nir_options;
compiler->glsl_compiler_options[i].ClampBlockIndicesToArrayBounds = true;
}
compiler->glsl_compiler_options[MESA_SHADER_TESS_CTRL].EmitNoIndirectInput = false;
compiler->glsl_compiler_options[MESA_SHADER_TESS_EVAL].EmitNoIndirectInput = false;
compiler->glsl_compiler_options[MESA_SHADER_TESS_CTRL].EmitNoIndirectOutput = false;
if (compiler->scalar_stage[MESA_SHADER_GEOMETRY])
compiler->glsl_compiler_options[MESA_SHADER_GEOMETRY].EmitNoIndirectInput = false;
return compiler;
}
static void
insert_u64_bit(uint64_t *val, bool add)
{
*val = (*val << 1) | !!add;
}
uint64_t
brw_get_compiler_config_value(const struct brw_compiler *compiler)
{
uint64_t config = 0;
insert_u64_bit(&config, compiler->precise_trig);
if (compiler->devinfo->gen >= 8 && compiler->devinfo->gen < 10) {
insert_u64_bit(&config, compiler->scalar_stage[MESA_SHADER_VERTEX]);
insert_u64_bit(&config, compiler->scalar_stage[MESA_SHADER_TESS_CTRL]);
insert_u64_bit(&config, compiler->scalar_stage[MESA_SHADER_TESS_EVAL]);
insert_u64_bit(&config, compiler->scalar_stage[MESA_SHADER_GEOMETRY]);
}
uint64_t debug_bits = INTEL_DEBUG;
uint64_t mask = DEBUG_DISK_CACHE_MASK;
while (mask != 0) {
const uint64_t bit = 1ULL << (ffsll(mask) - 1);
insert_u64_bit(&config, (debug_bits & bit) != 0);
mask &= ~bit;
}
return config;
}
unsigned
brw_prog_data_size(gl_shader_stage stage)
{
STATIC_ASSERT(MESA_SHADER_VERTEX == 0);
STATIC_ASSERT(MESA_SHADER_TESS_CTRL == 1);
STATIC_ASSERT(MESA_SHADER_TESS_EVAL == 2);
STATIC_ASSERT(MESA_SHADER_GEOMETRY == 3);
STATIC_ASSERT(MESA_SHADER_FRAGMENT == 4);
STATIC_ASSERT(MESA_SHADER_COMPUTE == 5);
static const size_t stage_sizes[] = {
sizeof(struct brw_vs_prog_data),
sizeof(struct brw_tcs_prog_data),
sizeof(struct brw_tes_prog_data),
sizeof(struct brw_gs_prog_data),
sizeof(struct brw_wm_prog_data),
sizeof(struct brw_cs_prog_data),
};
assert((int)stage >= 0 && stage < ARRAY_SIZE(stage_sizes));
return stage_sizes[stage];
}
unsigned
brw_prog_key_size(gl_shader_stage stage)
{
static const size_t stage_sizes[] = {
sizeof(struct brw_vs_prog_key),
sizeof(struct brw_tcs_prog_key),
sizeof(struct brw_tes_prog_key),
sizeof(struct brw_gs_prog_key),
sizeof(struct brw_wm_prog_key),
sizeof(struct brw_cs_prog_key),
};
assert((int)stage >= 0 && stage < ARRAY_SIZE(stage_sizes));
return stage_sizes[stage];
}
|
