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
|
/*
* Copyright (c) 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.
*/
/**
* \file lower_buffer_access.cpp
*
* Helper for IR lowering pass to replace dereferences of buffer object based
* shader variables with intrinsic function calls.
*
* This helper is used by lowering passes for UBOs, SSBOs and compute shader
* shared variables.
*/
#include "lower_buffer_access.h"
#include "ir_builder.h"
#include "main/macros.h"
#include "util/list.h"
#include "glsl_parser_extras.h"
using namespace ir_builder;
namespace lower_buffer_access {
static inline int
writemask_for_size(unsigned n)
{
return ((1 << n) - 1);
}
/**
* Takes a deref and recursively calls itself to break the deref down to the
* point that the reads or writes generated are contiguous scalars or vectors.
*/
void
lower_buffer_access::emit_access(void *mem_ctx,
bool is_write,
ir_dereference *deref,
ir_variable *base_offset,
unsigned int deref_offset,
bool row_major,
int matrix_columns,
unsigned int packing,
unsigned int write_mask)
{
if (deref->type->is_record()) {
unsigned int field_offset = 0;
for (unsigned i = 0; i < deref->type->length; i++) {
const struct glsl_struct_field *field =
&deref->type->fields.structure[i];
ir_dereference *field_deref =
new(mem_ctx) ir_dereference_record(deref->clone(mem_ctx, NULL),
field->name);
field_offset =
glsl_align(field_offset,
field->type->std140_base_alignment(row_major));
emit_access(mem_ctx, is_write, field_deref, base_offset,
deref_offset + field_offset,
row_major, 1, packing,
writemask_for_size(field_deref->type->vector_elements));
field_offset += field->type->std140_size(row_major);
}
return;
}
if (deref->type->is_array()) {
unsigned array_stride = packing == GLSL_INTERFACE_PACKING_STD430 ?
deref->type->fields.array->std430_array_stride(row_major) :
glsl_align(deref->type->fields.array->std140_size(row_major), 16);
for (unsigned i = 0; i < deref->type->length; i++) {
ir_constant *element = new(mem_ctx) ir_constant(i);
ir_dereference *element_deref =
new(mem_ctx) ir_dereference_array(deref->clone(mem_ctx, NULL),
element);
emit_access(mem_ctx, is_write, element_deref, base_offset,
deref_offset + i * array_stride,
row_major, 1, packing,
writemask_for_size(element_deref->type->vector_elements));
}
return;
}
if (deref->type->is_matrix()) {
for (unsigned i = 0; i < deref->type->matrix_columns; i++) {
ir_constant *col = new(mem_ctx) ir_constant(i);
ir_dereference *col_deref =
new(mem_ctx) ir_dereference_array(deref->clone(mem_ctx, NULL), col);
if (row_major) {
/* For a row-major matrix, the next column starts at the next
* element.
*/
int size_mul = deref->type->is_double() ? 8 : 4;
emit_access(mem_ctx, is_write, col_deref, base_offset,
deref_offset + i * size_mul,
row_major, deref->type->matrix_columns, packing,
writemask_for_size(col_deref->type->vector_elements));
} else {
int size_mul;
/* std430 doesn't round up vec2 size to a vec4 size */
if (packing == GLSL_INTERFACE_PACKING_STD430 &&
deref->type->vector_elements == 2 &&
!deref->type->is_double()) {
size_mul = 8;
} else {
/* std140 always rounds the stride of arrays (and matrices) to a
* vec4, so matrices are always 16 between columns/rows. With
* doubles, they will be 32 apart when there are more than 2 rows.
*
* For both std140 and std430, if the member is a
* three-'component vector with components consuming N basic
* machine units, the base alignment is 4N. For vec4, base
* alignment is 4N.
*/
size_mul = (deref->type->is_double() &&
deref->type->vector_elements > 2) ? 32 : 16;
}
emit_access(mem_ctx, is_write, col_deref, base_offset,
deref_offset + i * size_mul,
row_major, deref->type->matrix_columns, packing,
writemask_for_size(col_deref->type->vector_elements));
}
}
return;
}
assert(deref->type->is_scalar() || deref->type->is_vector());
if (!row_major) {
ir_rvalue *offset =
add(base_offset, new(mem_ctx) ir_constant(deref_offset));
unsigned mask =
is_write ? write_mask : (1 << deref->type->vector_elements) - 1;
insert_buffer_access(mem_ctx, deref, deref->type, offset, mask, -1);
} else {
unsigned N = deref->type->is_double() ? 8 : 4;
/* We're dereffing a column out of a row-major matrix, so we
* gather the vector from each stored row.
*/
assert(deref->type->base_type == GLSL_TYPE_FLOAT ||
deref->type->base_type == GLSL_TYPE_DOUBLE);
/* Matrices, row_major or not, are stored as if they were
* arrays of vectors of the appropriate size in std140.
* Arrays have their strides rounded up to a vec4, so the
* matrix stride is always 16. However a double matrix may either be 16
* or 32 depending on the number of columns.
*/
assert(matrix_columns <= 4);
unsigned matrix_stride = 0;
/* Matrix stride for std430 mat2xY matrices are not rounded up to
* vec4 size. From OpenGL 4.3 spec, section 7.6.2.2 "Standard Uniform
* Block Layout":
*
* "2. If the member is a two- or four-component vector with components
* consuming N basic machine units, the base alignment is 2N or 4N,
* respectively." [...]
* "4. If the member is an array of scalars or vectors, the base alignment
* and array stride are set to match the base alignment of a single array
* element, according to rules (1), (2), and (3), and rounded up to the
* base alignment of a vec4." [...]
* "7. If the member is a row-major matrix with C columns and R rows, the
* matrix is stored identically to an array of R row vectors with C
* components each, according to rule (4)." [...]
* "When using the std430 storage layout, shader storage blocks will be
* laid out in buffer storage identically to uniform and shader storage
* blocks using the std140 layout, except that the base alignment and
* stride of arrays of scalars and vectors in rule 4 and of structures in
* rule 9 are not rounded up a multiple of the base alignment of a vec4."
*/
if (packing == GLSL_INTERFACE_PACKING_STD430 && matrix_columns == 2)
matrix_stride = 2 * N;
else
matrix_stride = glsl_align(matrix_columns * N, 16);
const glsl_type *deref_type = deref->type->base_type == GLSL_TYPE_FLOAT ?
glsl_type::float_type : glsl_type::double_type;
for (unsigned i = 0; i < deref->type->vector_elements; i++) {
ir_rvalue *chan_offset =
add(base_offset,
new(mem_ctx) ir_constant(deref_offset + i * matrix_stride));
if (!is_write || ((1U << i) & write_mask))
insert_buffer_access(mem_ctx, deref, deref_type, chan_offset,
(1U << i), i);
}
}
}
} /* namespace lower_buffer_access */
|
