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
|
/*
* Copyright © 2015 Broadcom
*
* 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 "util/macros.h"
#include "nir.h"
#include "nir_builder.h"
/** @file nir_lower_load_const_to_scalar.c
*
* Replaces vector nir_load_const instructions with a series of loads and a
* vec[234] to reconstruct the original vector (on the assumption that
* nir_lower_alu_to_scalar() will then be used to split it up).
*
* This gives NIR a chance to CSE more operations on a scalar shader, when the
* same value was used in different vector contant loads.
*/
static bool
lower_load_const_instr_scalar(nir_load_const_instr *lower)
{
if (lower->def.num_components == 1)
return false;
nir_builder b;
nir_builder_init(&b, nir_cf_node_get_function(&lower->instr.block->cf_node));
b.cursor = nir_before_instr(&lower->instr);
/* Emit the individual loads. */
nir_ssa_def *loads[NIR_MAX_VEC_COMPONENTS];
for (unsigned i = 0; i < lower->def.num_components; i++) {
nir_load_const_instr *load_comp =
nir_load_const_instr_create(b.shader, 1, lower->def.bit_size);
switch (lower->def.bit_size) {
case 64:
load_comp->value.u64[0] = lower->value.u64[i];
break;
case 32:
load_comp->value.u32[0] = lower->value.u32[i];
break;
case 16:
load_comp->value.u16[0] = lower->value.u16[i];
break;
case 8:
load_comp->value.u8[0] = lower->value.u8[i];
break;
case 1:
load_comp->value.b[0] = lower->value.b[i];
break;
default:
assert(!"invalid bit size");
}
nir_builder_instr_insert(&b, &load_comp->instr);
loads[i] = &load_comp->def;
}
/* Batch things back together into a vector. */
nir_ssa_def *vec = nir_vec(&b, loads, lower->def.num_components);
/* Replace the old load with a reference to our reconstructed vector. */
nir_ssa_def_rewrite_uses(&lower->def, nir_src_for_ssa(vec));
nir_instr_remove(&lower->instr);
return true;
}
static bool
nir_lower_load_const_to_scalar_impl(nir_function_impl *impl)
{
bool progress = false;
nir_foreach_block(block, impl) {
nir_foreach_instr_safe(instr, block) {
if (instr->type == nir_instr_type_load_const)
progress |=
lower_load_const_instr_scalar(nir_instr_as_load_const(instr));
}
}
if (progress) {
nir_metadata_preserve(impl, nir_metadata_block_index |
nir_metadata_dominance);
} else {
#ifndef NDEBUG
impl->valid_metadata &= ~nir_metadata_not_properly_reset;
#endif
}
return progress;
}
bool
nir_lower_load_const_to_scalar(nir_shader *shader)
{
bool progress = false;
nir_foreach_function(function, shader) {
if (function->impl)
progress |= nir_lower_load_const_to_scalar_impl(function->impl);
}
return progress;
}
|