2 * Copyright © 2016 Intel Corporation
3 * Copyright © 2019 Valve Corporation
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
28 * \file nir_opt_move.c
30 * This pass can move various operations just before their first use inside the
31 * same basic block. Usually this is to reduce register usage. It's probably
32 * not a good idea to use this in an optimization loop.
34 * Moving comparisons is useful because many GPUs generate condition codes
35 * for comparisons, and use predication for conditional selects and control
36 * flow. In a sequence such as:
38 * vec1 32 ssa_1 = flt a b
39 * <some other operations>
40 * vec1 32 ssa_2 = bcsel ssa_1 c d
42 * the backend would likely do the comparison, producing condition codes,
43 * then save those to a boolean value. The intervening operations might
44 * trash the condition codes. Then, in order to do the bcsel, it would
45 * need to re-populate the condition code register based on the boolean.
47 * By moving the comparison just before the bcsel, the condition codes could
48 * be used directly. This eliminates the need to reload them from the boolean
49 * (generally eliminating an instruction). It may also eliminate the need to
50 * create a boolean value altogether (unless it's used elsewhere), which could
51 * lower register pressure.
55 move_source(nir_src
*src
, nir_block
*block
, nir_instr
*before
, nir_move_options options
)
60 nir_instr
*src_instr
= src
->ssa
->parent_instr
;
62 if (src_instr
->block
== block
&& nir_can_move_instr(src_instr
, options
)) {
63 exec_node_remove(&src_instr
->node
);
66 exec_node_insert_node_before(&before
->node
, &src_instr
->node
);
68 exec_list_push_tail(&block
->instr_list
, &src_instr
->node
);
75 struct source_cb_data
{
77 nir_move_options options
;
81 move_source_cb(nir_src
*src
, void *data_ptr
)
83 struct source_cb_data data
= *(struct source_cb_data
*)data_ptr
;
85 nir_instr
*instr
= src
->parent_instr
;
86 if (move_source(src
, instr
->block
, instr
, data
.options
))
87 *data
.progress
= true;
89 return true; /* nir_foreach_src should keep going */
93 move(nir_block
*block
, nir_move_options options
)
95 bool progress
= false;
97 /* We use a simple approach: walk instructions backwards.
99 * If the instruction's source is a comparison from the same block,
100 * simply move it here. This may break SSA if it's used earlier in
101 * the block as well. However, as we walk backwards, we'll find the
102 * earlier use and move it again, further up. It eventually ends up
103 * dominating all uses again, restoring SSA form.
105 * Before walking instructions, we consider the if-condition at the
106 * end of the block, if one exists. It's effectively a use at the
107 * bottom of the block.
109 nir_if
*iff
= nir_block_get_following_if(block
);
111 progress
|= move_source(&iff
->condition
, block
, NULL
, options
);
114 nir_foreach_instr_reverse(instr
, block
) {
115 /* The sources of phi instructions happen after the predecessor block
116 * but before this block. (Yes, that's between blocks). This means
117 * that we don't need to move them in order for them to be correct.
118 * We could move them to encourage comparisons that are used in a phi to
119 * the end of the block, doing so correctly would make the pass
120 * substantially more complicated and wouldn't gain us anything since
121 * the phi can't use a flag value anyway.
124 if (instr
->type
== nir_instr_type_phi
) {
125 /* We're going backwards so everything else is a phi too */
127 } else if (instr
->type
== nir_instr_type_alu
) {
128 /* Walk ALU instruction sources backwards so that bcsel's boolean
129 * condition is processed last for when comparisons are being moved.
131 nir_alu_instr
*alu
= nir_instr_as_alu(instr
);
132 for (int i
= nir_op_infos
[alu
->op
].num_inputs
- 1; i
>= 0; i
--) {
133 progress
|= move_source(&alu
->src
[i
].src
, block
, instr
, options
);
136 struct source_cb_data data
;
137 data
.progress
= &progress
;
138 data
.options
= options
;
139 nir_foreach_src(instr
, move_source_cb
, &data
);
147 nir_opt_move(nir_shader
*shader
, nir_move_options options
)
149 bool progress
= false;
151 nir_foreach_function(func
, shader
) {
155 nir_foreach_block(block
, func
->impl
) {
156 if (move(block
, options
)) {
157 nir_metadata_preserve(func
->impl
, nir_metadata_block_index
|
158 nir_metadata_dominance
|
159 nir_metadata_live_ssa_defs
);