Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 5564 | serge | 1 | /* |
| 2 | * Copyright © 2014 Broadcom |
||
| 3 | * |
||
| 4 | * Permission is hereby granted, free of charge, to any person obtaining a |
||
| 5 | * copy of this software and associated documentation files (the "Software"), |
||
| 6 | * to deal in the Software without restriction, including without limitation |
||
| 7 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
||
| 8 | * and/or sell copies of the Software, and to permit persons to whom the |
||
| 9 | * Software is furnished to do so, subject to the following conditions: |
||
| 10 | * |
||
| 11 | * The above copyright notice and this permission notice (including the next |
||
| 12 | * paragraph) shall be included in all copies or substantial portions of the |
||
| 13 | * Software. |
||
| 14 | * |
||
| 15 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
||
| 16 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
||
| 17 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
||
| 18 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
||
| 19 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
||
| 20 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS |
||
| 21 | * IN THE SOFTWARE. |
||
| 22 | */ |
||
| 23 | |||
| 24 | /** |
||
| 25 | * @file vc4_opt_dead_code.c |
||
| 26 | * |
||
| 27 | * This is a simmple dead code eliminator for QIR with no control flow. |
||
| 28 | * |
||
| 29 | * It walks from the bottom of the instruction list, removing instructions |
||
| 30 | * with a destination that is never used, and marking the sources of non-dead |
||
| 31 | * instructions as used. |
||
| 32 | */ |
||
| 33 | |||
| 34 | #include "vc4_qir.h" |
||
| 35 | |||
| 36 | static bool debug; |
||
| 37 | |||
| 38 | static void |
||
| 39 | dce(struct vc4_compile *c, struct qinst *inst) |
||
| 40 | { |
||
| 41 | if (debug) { |
||
| 42 | fprintf(stderr, "Removing: "); |
||
| 43 | qir_dump_inst(c, inst); |
||
| 44 | fprintf(stderr, "\n"); |
||
| 45 | } |
||
| 46 | assert(!inst->sf); |
||
| 47 | qir_remove_instruction(c, inst); |
||
| 48 | } |
||
| 49 | |||
| 50 | static bool |
||
| 51 | has_nonremovable_reads(struct vc4_compile *c, struct qinst *inst) |
||
| 52 | { |
||
| 53 | for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) { |
||
| 54 | if (inst->src[i].file == QFILE_VPM) { |
||
| 55 | uint32_t attr = inst->src[i].index / 4; |
||
| 56 | uint32_t offset = (inst->src[i].index % 4) * 4; |
||
| 57 | |||
| 58 | if (c->vattr_sizes[attr] != offset + 4) |
||
| 59 | return true; |
||
| 60 | |||
| 61 | /* Can't get rid of the last VPM read, or the |
||
| 62 | * simulator (at least) throws an error. |
||
| 63 | */ |
||
| 64 | uint32_t total_size = 0; |
||
| 65 | for (uint32_t i = 0; i < ARRAY_SIZE(c->vattr_sizes); i++) |
||
| 66 | total_size += c->vattr_sizes[i]; |
||
| 67 | if (total_size == 4) |
||
| 68 | return true; |
||
| 69 | } |
||
| 70 | |||
| 71 | if (inst->src[i].file == QFILE_VARY && |
||
| 72 | c->input_semantics[inst->src[i].index].semantic == 0xff) { |
||
| 73 | return true; |
||
| 74 | } |
||
| 75 | } |
||
| 76 | |||
| 77 | return false; |
||
| 78 | } |
||
| 79 | |||
| 80 | bool |
||
| 81 | qir_opt_dead_code(struct vc4_compile *c) |
||
| 82 | { |
||
| 83 | bool progress = false; |
||
| 84 | bool *used = calloc(c->num_temps, sizeof(bool)); |
||
| 85 | bool sf_used = false; |
||
| 86 | /* Whether we're eliminating texture setup currently. */ |
||
| 87 | bool dce_tex = false; |
||
| 88 | |||
| 89 | struct simple_node *node, *t; |
||
| 90 | for (node = c->instructions.prev, t = node->prev; |
||
| 91 | &c->instructions != node; |
||
| 92 | node = t, t = t->prev) { |
||
| 93 | struct qinst *inst = (struct qinst *)node; |
||
| 94 | |||
| 95 | if (inst->dst.file == QFILE_TEMP && |
||
| 96 | !used[inst->dst.index] && |
||
| 97 | !inst->sf && |
||
| 98 | (!qir_has_side_effects(c, inst) || |
||
| 99 | inst->op == QOP_TEX_RESULT) && |
||
| 100 | !has_nonremovable_reads(c, inst)) { |
||
| 101 | if (inst->op == QOP_TEX_RESULT) { |
||
| 102 | dce_tex = true; |
||
| 103 | c->num_texture_samples--; |
||
| 104 | } |
||
| 105 | |||
| 106 | for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) { |
||
| 107 | if (inst->src[i].file != QFILE_VPM) |
||
| 108 | continue; |
||
| 109 | uint32_t attr = inst->src[i].index / 4; |
||
| 110 | uint32_t offset = (inst->src[i].index % 4) * 4; |
||
| 111 | |||
| 112 | if (c->vattr_sizes[attr] == offset + 4) { |
||
| 113 | c->num_inputs--; |
||
| 114 | c->vattr_sizes[attr] -= 4; |
||
| 115 | } |
||
| 116 | } |
||
| 117 | |||
| 118 | dce(c, inst); |
||
| 119 | progress = true; |
||
| 120 | continue; |
||
| 121 | } |
||
| 122 | |||
| 123 | if (qir_depends_on_flags(inst)) |
||
| 124 | sf_used = true; |
||
| 125 | if (inst->sf) { |
||
| 126 | if (!sf_used) { |
||
| 127 | if (debug) { |
||
| 128 | fprintf(stderr, "Removing SF on: "); |
||
| 129 | qir_dump_inst(c, inst); |
||
| 130 | fprintf(stderr, "\n"); |
||
| 131 | } |
||
| 132 | |||
| 133 | inst->sf = false; |
||
| 134 | progress = true; |
||
| 135 | } |
||
| 136 | sf_used = false; |
||
| 137 | } |
||
| 138 | |||
| 139 | if (inst->op == QOP_TEX_RESULT) |
||
| 140 | dce_tex = false; |
||
| 141 | |||
| 142 | if (dce_tex && (inst->op == QOP_TEX_S || |
||
| 143 | inst->op == QOP_TEX_T || |
||
| 144 | inst->op == QOP_TEX_R || |
||
| 145 | inst->op == QOP_TEX_B || |
||
| 146 | inst->op == QOP_TEX_DIRECT)) { |
||
| 147 | dce(c, inst); |
||
| 148 | progress = true; |
||
| 149 | continue; |
||
| 150 | } |
||
| 151 | |||
| 152 | for (int i = 0; i < qir_get_op_nsrc(inst->op); i++) { |
||
| 153 | if (inst->src[i].file == QFILE_TEMP) |
||
| 154 | used[inst->src[i].index] = true; |
||
| 155 | } |
||
| 156 | } |
||
| 157 | |||
| 158 | free(used); |
||
| 159 | |||
| 160 | return progress; |
||
| 161 | } |