2
* Copyright (C) 2018 Alyssa Rosenzweig
3
* Copyright (C) 2019 Collabora, Ltd.
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 FROM,
21
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
26
#include "util/u_memory.h"
27
#include "midgard_ops.h"
29
/* SIMD-aware dead code elimination. Perform liveness analysis step-by-step,
30
* removing dead components. If an instruction ends up with a zero mask, the
31
* instruction in total is dead and should be removed. */
34
can_cull_mask(compiler_context *ctx, midgard_instruction *ins)
36
if (ins->dest >= ctx->temp_count)
39
if (ins->dest == ctx->blend_src1)
42
if (ins->type == TAG_LOAD_STORE_4)
43
if (load_store_opcode_props[ins->op].props & LDST_SPECIAL_MASK)
50
can_dce(midgard_instruction *ins)
55
if (ins->compact_branch)
58
if (ins->type == TAG_LOAD_STORE_4)
59
if (load_store_opcode_props[ins->op].props & LDST_SIDE_FX)
62
if (ins->type == TAG_TEXTURE_4)
63
if (ins->op == midgard_tex_op_barrier)
70
midgard_opt_dead_code_eliminate_block(compiler_context *ctx, midgard_block *block)
72
bool progress = false;
74
uint16_t *live = mem_dup(block->base.live_out, ctx->temp_count * sizeof(uint16_t));
76
mir_foreach_instr_in_block_rev(block, ins) {
77
if (can_cull_mask(ctx, ins)) {
78
unsigned type_size = nir_alu_type_get_type_size(ins->dest_type);
79
unsigned round_size = type_size;
80
unsigned oldmask = ins->mask;
82
/* Make sure we're packable */
83
if (type_size < 32 && ins->type == TAG_LOAD_STORE_4)
86
unsigned rounded = mir_round_bytemask_up(live[ins->dest], round_size);
87
unsigned cmask = mir_from_bytemask(rounded, type_size);
90
progress |= (ins->mask != oldmask);
93
mir_liveness_ins_update(live, ins, ctx->temp_count);
96
mir_foreach_instr_in_block_safe(block, ins) {
98
mir_remove_instruction(ins);
109
midgard_opt_dead_code_eliminate(compiler_context *ctx)
111
/* We track liveness. In fact, it's ok if we assume more things are
112
* live than they actually are, that just reduces the effectiveness of
113
* this iterations lightly. And DCE has the effect of strictly reducing
114
* liveness, so we can run DCE across all blocks while only computing
115
* liveness at the beginning. */
117
mir_invalidate_liveness(ctx);
118
mir_compute_liveness(ctx);
120
bool progress = false;
122
mir_foreach_block(ctx, block) {
123
progress |= midgard_opt_dead_code_eliminate_block(ctx, (midgard_block *) block);
129
/* Removes dead moves, that is, moves with a destination overwritten before
130
* being read. Normally handled implicitly as part of DCE, but this has to run
131
* after the out-of-SSA pass */
134
midgard_opt_dead_move_eliminate(compiler_context *ctx, midgard_block *block)
136
bool progress = false;
138
mir_foreach_instr_in_block_safe(block, ins) {
139
if (ins->type != TAG_ALU_4) continue;
140
if (ins->compact_branch) continue;
141
if (!OP_IS_MOVE(ins->op)) continue;
143
/* Check if it's overwritten in this block before being read */
144
bool overwritten = false;
146
mir_foreach_instr_in_block_from(block, q, mir_next_op(ins)) {
148
if (mir_has_arg(q, ins->dest))
151
/* Check if overwritten */
152
if (q->dest == ins->dest) {
153
/* Special case to vec4; component tracking is
156
overwritten = (q->mask == 0xF);
162
mir_remove_instruction(ins);