freedreno/ir3/ir3_cache.c

/*
 * Copyright (C) 2015 Rob Clark <robclark@freedesktop.org>
 *
 * 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.
 *
 * Authors:
 *    Rob Clark <robclark@freedesktop.org>
 */

#include "util/hash_table.h"
#include "util/ralloc.h"
#define XXH_INLINE_ALL
#include "util/xxhash.h"

#include "ir3_cache.h"
#include "ir3_gallium.h"

static uint32_t
key_hash(const void *_key)
{
   const struct ir3_cache_key *key = _key;
   return XXH32(key, sizeof(*key), 0);
}

static bool
key_equals(const void *_a, const void *_b)
{
   const struct ir3_cache_key *a = _a;
   const struct ir3_cache_key *b = _b;
   // TODO we could optimize the key shader-variant key comparison by not
   // ignoring has_per_samp.. not really sure if that helps..
   return memcmp(a, b, sizeof(struct ir3_cache_key)) == 0;
}

struct ir3_cache {
   /* cache mapping gallium/etc shader state-objs + shader-key to backend
    * specific state-object
    */
   struct hash_table *ht;

   const struct ir3_cache_funcs *funcs;
   void *data;
};

struct ir3_cache *
ir3_cache_create(const struct ir3_cache_funcs *funcs, void *data)
{
   struct ir3_cache *cache = rzalloc(NULL, struct ir3_cache);

   cache->ht = _mesa_hash_table_create(cache, key_hash, key_equals);
   cache->funcs = funcs;
   cache->data = data;

   return cache;
}

void
ir3_cache_destroy(struct ir3_cache *cache)
{
   if (!cache)
      return;

   /* _mesa_hash_table_destroy is so *almost* useful.. */
   hash_table_foreach (cache->ht, entry) {
      cache->funcs->destroy_state(cache->data, entry->data);
   }

   ralloc_free(cache);
}

struct ir3_program_state *
ir3_cache_lookup(struct ir3_cache *cache, const struct ir3_cache_key *key,
                 struct pipe_debug_callback *debug)
{
   uint32_t hash = key_hash(key);
   struct hash_entry *entry =
      _mesa_hash_table_search_pre_hashed(cache->ht, hash, key);

   if (entry) {
      return entry->data;
   }

   if (key->hs)
      debug_assert(key->ds);

   struct ir3_shader *shaders[MESA_SHADER_STAGES] = {
      [MESA_SHADER_VERTEX] = ir3_get_shader(key->vs),
      [MESA_SHADER_TESS_CTRL] = ir3_get_shader(key->hs),
      [MESA_SHADER_TESS_EVAL] = ir3_get_shader(key->ds),
      [MESA_SHADER_GEOMETRY] = ir3_get_shader(key->gs),
      [MESA_SHADER_FRAGMENT] = ir3_get_shader(key->fs),
   };

   struct ir3_shader_variant *variants[MESA_SHADER_STAGES];
   struct ir3_shader_key shader_key = key->key;

   for (gl_shader_stage stage = MESA_SHADER_VERTEX; stage < MESA_SHADER_STAGES;
        stage++) {
      if (shaders[stage]) {
         variants[stage] =
            ir3_shader_variant(shaders[stage], shader_key, false, debug);
         if (!variants[stage])
            return NULL;
      } else {
         variants[stage] = NULL;
      }
   }

   struct ir3_compiler *compiler = shaders[MESA_SHADER_VERTEX]->compiler;
   uint32_t safe_constlens = ir3_trim_constlen(variants, compiler);
   shader_key.safe_constlen = true;

   for (gl_shader_stage stage = MESA_SHADER_VERTEX; stage < MESA_SHADER_STAGES;
        stage++) {
      if (safe_constlens & (1 << stage)) {
         variants[stage] =
            ir3_shader_variant(shaders[stage], shader_key, false, debug);
         if (!variants[stage])
            return NULL;
      }
   }

   struct ir3_shader_variant *bs;

   if (ir3_has_binning_vs(&key->key)) {
      /* starting with a6xx, the same const state is used for binning and draw
       * passes, so the binning pass VS variant needs to match the main VS
       */
      shader_key.safe_constlen = (compiler->gen >= 6) &&
            !!(safe_constlens & (1 << MESA_SHADER_VERTEX));
      bs =
         ir3_shader_variant(shaders[MESA_SHADER_VERTEX], shader_key, true, debug);
      if (!bs)
         return NULL;
   } else {
      bs = variants[MESA_SHADER_VERTEX];
   }

   struct ir3_program_state *state = cache->funcs->create_state(
      cache->data, bs, variants[MESA_SHADER_VERTEX],
      variants[MESA_SHADER_TESS_CTRL], variants[MESA_SHADER_TESS_EVAL],
      variants[MESA_SHADER_GEOMETRY], variants[MESA_SHADER_FRAGMENT],
      key);
   state->key = *key;

   /* NOTE: uses copy of key in state obj, because pointer passed by caller
    * is probably on the stack
    */
   _mesa_hash_table_insert_pre_hashed(cache->ht, hash, &state->key, state);

   return state;
}

/* call when an API level state object is destroyed, to invalidate
 * cache entries which reference that state object.
 */
void
ir3_cache_invalidate(struct ir3_cache *cache, void *stobj)
{
   if (!cache)
      return;

   hash_table_foreach (cache->ht, entry) {
      const struct ir3_cache_key *key = entry->key;
      if ((key->fs == stobj) || (key->vs == stobj) || (key->ds == stobj) ||
          (key->hs == stobj) || (key->gs == stobj)) {
         cache->funcs->destroy_state(cache->data, entry->data);
         _mesa_hash_table_remove(cache->ht, entry);
         return;
      }
   }
}