[Mesa-dev] [PATCH 3/3] util: Change util/set to use quadratic probing

Jason Ekstrand jason at jlekstrand.net
Mon Apr 13 15:05:38 PDT 2015


On Sat, Apr 11, 2015 at 4:25 PM, Thomas Helland
<thomashelland90 at gmail.com> wrote:
> The same rationale applies here as for the hash table.
> Power of two size should give better performance,
> and using the algorithm hash = sh + i/2 + i*i/2
> should result in only distinct hash values when hitting collisions.
>
> Difference at 95.0% confidence
>      -7.9505 +/- 2.44011
>    -5.04357% +/- 1.54794%
>
> V3: Feedback from Eric Anholt
>    - Don't change load factor and starting size.
>
> V2: Feedback from Connor Abbott
>    - Don't set initial hash address before potential rehash
>    - Remove hash_sizes table
>    - Correct the quadratic hashing algorithm
>    - Use correct comment style
>
>    Feedback from Jason Ekstrand
>    - Use unreachable() to detect if we fail to insert
>
> Signed-off-by: Thomas Helland <thomashelland90 at gmail.com>
> ---
>  src/util/set.c | 118 ++++++++++++++++++++-------------------------------------
>  src/util/set.h |   3 +-
>  2 files changed, 43 insertions(+), 78 deletions(-)
>
> diff --git a/src/util/set.c b/src/util/set.c
> index f01f869..7ff9520 100644
> --- a/src/util/set.c
> +++ b/src/util/set.c
> @@ -32,6 +32,17 @@
>   *    Keith Packard <keithp at keithp.com>
>   */
>
> +/**
> + * Implements an open-addressing, quadratic probing hash-set.
> + *
> + * We choose set sizes that's a power of two.
> + * This is computationally less expensive than primes.
> + * As a bonus the size and free space can be calculated instead of looked up.
> + * FNV-1a has good avalanche properties, so collision is not an issue.
> + * These sets are sized to have an extra 10% free to avoid
> + * exponential performance degradation as the set fills.
> + */
> +
>  #include <stdlib.h>
>  #include <assert.h>
>
> @@ -39,51 +50,9 @@
>  #include "ralloc.h"
>  #include "set.h"
>
> -/*
> - * From Knuth -- a good choice for hash/rehash values is p, p-2 where
> - * p and p-2 are both prime.  These tables are sized to have an extra 10%
> - * free to avoid exponential performance degradation as the hash table fills
> - */
> -
>  uint32_t deleted_key_value;
>  const void *deleted_key = &deleted_key_value;
>
> -static const struct {
> -   uint32_t max_entries, size, rehash;
> -} hash_sizes[] = {
> -   { 2,            5,            3            },
> -   { 4,            7,            5            },
> -   { 8,            13,           11           },
> -   { 16,           19,           17           },
> -   { 32,           43,           41           },
> -   { 64,           73,           71           },
> -   { 128,          151,          149          },
> -   { 256,          283,          281          },
> -   { 512,          571,          569          },
> -   { 1024,         1153,         1151         },
> -   { 2048,         2269,         2267         },
> -   { 4096,         4519,         4517         },
> -   { 8192,         9013,         9011         },
> -   { 16384,        18043,        18041        },
> -   { 32768,        36109,        36107        },
> -   { 65536,        72091,        72089        },
> -   { 131072,       144409,       144407       },
> -   { 262144,       288361,       288359       },
> -   { 524288,       576883,       576881       },
> -   { 1048576,      1153459,      1153457      },
> -   { 2097152,      2307163,      2307161      },
> -   { 4194304,      4613893,      4613891      },
> -   { 8388608,      9227641,      9227639      },
> -   { 16777216,     18455029,     18455027     },
> -   { 33554432,     36911011,     36911009     },
> -   { 67108864,     73819861,     73819859     },
> -   { 134217728,    147639589,    147639587    },
> -   { 268435456,    295279081,    295279079    },
> -   { 536870912,    590559793,    590559791    },
> -   { 1073741824,   1181116273,   1181116271   },
> -   { 2147483648ul, 2362232233ul, 2362232231ul }
> -};
> -
>  static int
>  entry_is_free(struct set_entry *entry)
>  {
> @@ -114,10 +83,9 @@ _mesa_set_create(void *mem_ctx,
>     if (ht == NULL)
>        return NULL;
>
> -   ht->size_index = 0;
> -   ht->size = hash_sizes[ht->size_index].size;
> -   ht->rehash = hash_sizes[ht->size_index].rehash;
> -   ht->max_entries = hash_sizes[ht->size_index].max_entries;
> +   ht->size_iteration = 2;
> +   ht->size = 1 << ht->size_iteration;
> +   ht->max_entries = ht->size * 0.9;
>     ht->key_hash_function = key_hash_function;
>     ht->key_equals_function = key_equals_function;
>     ht->table = rzalloc_array(ht, struct set_entry, ht->size);
> @@ -163,12 +131,11 @@ _mesa_set_destroy(struct set *ht, void (*delete_function)(struct set_entry *entr
>  static struct set_entry *
>  set_search(const struct set *ht, uint32_t hash, const void *key)
>  {
> -   uint32_t hash_address;
> +   uint32_t start_hash_address = hash & (ht->size - 1);
> +   uint32_t hash_address = start_hash_address;
> +   uint32_t quad_hash = 1;
>
> -   hash_address = hash % ht->size;
>     do {
> -      uint32_t double_hash;
> -
>        struct set_entry *entry = ht->table + hash_address;
>
>        if (entry_is_free(entry)) {
> @@ -179,10 +146,10 @@ set_search(const struct set *ht, uint32_t hash, const void *key)
>           }
>        }
>
> -      double_hash = 1 + hash % ht->rehash;
> -
> -      hash_address = (hash_address + double_hash) % ht->size;
> -   } while (hash_address != hash % ht->size);
> +      hash_address = (start_hash_address +
> +                (quad_hash + (quad_hash * quad_hash)) / 2) & (ht->size - 1);
> +      quad_hash++;
> +   } while (hash_address != start_hash_address);
>
>     return NULL;
>  }
> @@ -207,35 +174,30 @@ static struct set_entry *
>  set_add(struct set *ht, uint32_t hash, const void *key);
>
>  static void
> -set_rehash(struct set *ht, unsigned new_size_index)
> +set_rehash(struct set *ht, uint32_t new_size_iteration)
>  {
>     struct set old_ht;
>     struct set_entry *table, *entry;
>
> -   if (new_size_index >= ARRAY_SIZE(hash_sizes))
> +   if (new_size_iteration >= 31)
>        return;
>
>     table = rzalloc_array(ht, struct set_entry,
> -                         hash_sizes[new_size_index].size);
> +                         1 << new_size_iteration);
>     if (table == NULL)
>        return;
>
>     old_ht = *ht;
>
>     ht->table = table;
> -   ht->size_index = new_size_index;
> -   ht->size = hash_sizes[ht->size_index].size;
> -   ht->rehash = hash_sizes[ht->size_index].rehash;
> -   ht->max_entries = hash_sizes[ht->size_index].max_entries;
> +   ht->size_iteration = new_size_iteration;
> +   ht->size = 1 << new_size_iteration;
> +   ht->max_entries = ht->size * 0.7;
>     ht->entries = 0;
>     ht->deleted_entries = 0;
>
> -   for (entry = old_ht.table;
> -        entry != old_ht.table + old_ht.size;
> -        entry++) {
> -      if (entry_is_present(entry)) {
> -         set_add(ht, entry->hash, entry->key);
> -      }
> +   set_foreach(&old_ht, entry) {
> +      set_add(ht, entry->hash, entry->key);
>     }

This hunk, while I think it's correct, is unrelated.  Please drop it.
I think I'm ok with the cleanup, but it should be a separate patch if
you want to do that.
--Jason

>
>     ralloc_free(old_ht.table);
> @@ -250,19 +212,21 @@ set_rehash(struct set *ht, unsigned new_size_index)
>  static struct set_entry *
>  set_add(struct set *ht, uint32_t hash, const void *key)
>  {
> -   uint32_t hash_address;
> +   uint32_t start_hash_address, hash_address;
> +   uint32_t quad_hash = 1;
>     struct set_entry *available_entry = NULL;
>
>     if (ht->entries >= ht->max_entries) {
> -      set_rehash(ht, ht->size_index + 1);
> +      set_rehash(ht, ht->size_iteration + 1);
>     } else if (ht->deleted_entries + ht->entries >= ht->max_entries) {
> -      set_rehash(ht, ht->size_index);
> +      set_rehash(ht, ht->size_iteration);
>     }
>
> -   hash_address = hash % ht->size;
> +   start_hash_address = hash & (ht->size - 1);
> +   hash_address = start_hash_address;
> +
>     do {
>        struct set_entry *entry = ht->table + hash_address;
> -      uint32_t double_hash;
>
>        if (!entry_is_present(entry)) {
>           /* Stash the first available entry we find */
> @@ -288,10 +252,11 @@ set_add(struct set *ht, uint32_t hash, const void *key)
>           return entry;
>        }
>
> -      double_hash = 1 + hash % ht->rehash;
> +      hash_address = (start_hash_address +
> +                (quad_hash + (quad_hash * quad_hash)) / 2) & (ht->size - 1);
> +      quad_hash++;
> +   } while (hash_address != start_hash_address);
>
> -      hash_address = (hash_address + double_hash) % ht->size;
> -   } while (hash_address != hash % ht->size);
>
>     if (available_entry) {
>        if (entry_is_deleted(available_entry))
> @@ -305,6 +270,7 @@ set_add(struct set *ht, uint32_t hash, const void *key)
>     /* We could hit here if a required resize failed. An unchecked-malloc
>      * application could ignore this result.
>      */
> +   unreachable("Failed to insert entry in hash set");
>     return NULL;
>  }
>
> @@ -368,7 +334,7 @@ _mesa_set_random_entry(struct set *ht,
>                         int (*predicate)(struct set_entry *entry))
>  {
>     struct set_entry *entry;
> -   uint32_t i = rand() % ht->size;
> +   uint32_t i = rand() & (ht->size - 1);
>
>     if (ht->entries == 0)
>        return NULL;
> diff --git a/src/util/set.h b/src/util/set.h
> index 9acd2c2..fa17d3a 100644
> --- a/src/util/set.h
> +++ b/src/util/set.h
> @@ -46,9 +46,8 @@ struct set {
>     uint32_t (*key_hash_function)(const void *key);
>     bool (*key_equals_function)(const void *a, const void *b);
>     uint32_t size;
> -   uint32_t rehash;
>     uint32_t max_entries;
> -   uint32_t size_index;
> +   uint32_t size_iteration;
>     uint32_t entries;
>     uint32_t deleted_entries;
>  };
> --
> 2.3.4
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev


More information about the mesa-dev mailing list