src/gallium/drivers/vc4/vc4_opt_cse.c

   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_cse.c
  26  *
  27  * Implements CSE for QIR without control flow.
  28  *
  29  * For each operation that writes a destination (and isn't just a MOV), put it
  30  * in the hash table of all instructions that do so.  When faced with another
  31  * one, look it up in the hash table by its opcode and operands.  If there's
  32  * an entry in the table, then just reuse the entry's destination as the
  33  * source of a MOV instead of reproducing the computation.  That MOV will then
  34  * get cleaned up by copy propagation.
  35  */
  36
  37 #include "vc4_qir.h"
  38
  39 #include "util/hash_table.h"
  40 #include "util/ralloc.h"
  41
  42 static bool debug;
  43
  44 struct inst_key {
  45         enum qop op;
  46         struct qreg src[4];
  47         /**
  48          * If the instruction depends on the flags, how many SFs have been
  49          * seen before this instruction.
  50          */
  51         uint32_t implicit_arg_update_count;
  52 };
  53
  54 static bool
  55 inst_key_equals(const void *a, const void *b)
  56 {
  57         const struct inst_key *key_a = a;
  58         const struct inst_key *key_b = b;
  59
  60         return memcmp(key_a, key_b, sizeof(*key_a)) == 0;
  61 }
  62
  63 static struct qinst *
  64 vc4_find_cse(struct vc4_compile *c, struct hash_table *ht,
  65              struct qinst *inst, uint32_t sf_count)
  66 {
  67         if (inst->dst.file != QFILE_TEMP ||
  68             inst->op == QOP_MOV ||
  69             qir_get_op_nsrc(inst->op) > 4) {
  70                 return NULL;
  71         }
  72
  73         struct inst_key key;
  74         memset(&key, 0, sizeof(key));
  75         key.op = inst->op;
  76         memcpy(key.src, inst->src,
  77                qir_get_op_nsrc(inst->op) * sizeof(key.src[0]));
  78         if (qir_depends_on_flags(inst))
  79                 key.implicit_arg_update_count = sf_count;
  80
  81         uint32_t hash = _mesa_hash_data(&key, sizeof(key));
  82         struct hash_entry *entry =
  83                 _mesa_hash_table_search_pre_hashed(ht, hash, &key);
  84
  85         if (entry) {
  86                 if (debug) {
  87                         fprintf(stderr, "CSE found match:\n");
  88
  89                         fprintf(stderr, "  Original inst: ");
  90                         qir_dump_inst(c, entry->data);
  91                         fprintf(stderr, "\n");
  92
  93                         fprintf(stderr, "  Our inst:      ");
  94                         qir_dump_inst(c, inst);
  95                         fprintf(stderr, "\n");
  96                 }
  97
  98                 return entry->data;
  99         }
 100
 101         struct inst_key *alloc_key = ralloc(ht, struct inst_key);
 102         if (!alloc_key)
 103                 return NULL;
 104         memcpy(alloc_key, &key, sizeof(*alloc_key));
 105         _mesa_hash_table_insert_pre_hashed(ht, hash, alloc_key, inst);
 106
 107         if (debug) {
 108                 fprintf(stderr, "Added to CSE HT: ");
 109                 qir_dump_inst(c, inst);
 110                 fprintf(stderr, "\n");
 111         }
 112
 113         return NULL;
 114 }
 115
 116 bool
 117 qir_opt_cse(struct vc4_compile *c)
 118 {
 119         bool progress = false;
 120         uint32_t sf_count = 0;
 121
 122         struct hash_table *ht = _mesa_hash_table_create(NULL, NULL,
 123                                                         inst_key_equals);
 124         if (!ht)
 125                 return false;
 126
 127         list_for_each_entry(struct qinst, inst, &c->instructions, link) {
 128                 if (qir_has_side_effects(c, inst) ||
 129                     qir_has_side_effect_reads(c, inst) ||
 130                     inst->op == QOP_TLB_COLOR_READ) {
 131                         continue;
 132                 }
 133
 134                 if (inst->sf) {
 135                         sf_count++;
 136                 } else {
 137                         struct qinst *cse = vc4_find_cse(c, ht, inst, sf_count);
 138                         if (cse) {
 139                                 inst->src[0] = cse->dst;
 140                                 for (int i = 1; i < qir_get_op_nsrc(inst->op);
 141                                      i++)
 142                                         inst->src[i] = c->undef;
 143                                 inst->op = QOP_MOV;
 144                                 progress = true;
 145
 146                                 if (debug) {
 147                                         fprintf(stderr, "  Turned into:   ");
 148                                         qir_dump_inst(c, inst);
 149                                         fprintf(stderr, "\n");
 150                                 }
 151                         }
 152                 }
 153         }
 154
 155         ralloc_free(ht);
 156
 157         return progress;
 158 }