wingo pushed a commit to branch wip-whippet
in repository guile.

commit d425620d3700671243ce41b0795f3494561fbaa9
Author: Andy Wingo <wi...@igalia.com>
AuthorDate: Tue Apr 12 21:41:26 2022 +0200

    Add address map and set
---
 address-hash.h     |  45 ++++++++++++
 address-map.h      | 210 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 address-set.h      | 193 ++++++++++++++++++++++++++++++++++++++++++++++++
 test-address-map.c | 109 +++++++++++++++++++++++++++
 test-address-set.c |  98 +++++++++++++++++++++++++
 5 files changed, 655 insertions(+)

diff --git a/address-hash.h b/address-hash.h
new file mode 100644
index 000000000..49c33be97
--- /dev/null
+++ b/address-hash.h
@@ -0,0 +1,45 @@
+#ifndef ADDRESS_HASH_H
+#define ADDRESS_HASH_H
+
+#include <stdint.h>
+
+static uintptr_t hash_address(uintptr_t x) {
+  if (sizeof (x) < 8) {
+    // Chris Wellon's lowbias32, from https://nullprogram.com/blog/2018/07/31/.
+    x ^= x >> 16;
+    x *= 0x7feb352dU;
+    x ^= x >> 15;
+    x *= 0x846ca68bU;
+    x ^= x >> 16;
+    return x;
+  } else {
+    // Sebastiano Vigna's splitmix64 integer mixer, from
+    // https://prng.di.unimi.it/splitmix64.c.
+    x ^= x >> 30;
+    x *= 0xbf58476d1ce4e5b9U;
+    x ^= x >> 27;
+    x *= 0x94d049bb133111ebU;
+    x ^= x >> 31;
+    return x;
+  }
+}
+// Inverse of hash_address from https://nullprogram.com/blog/2018/07/31/.
+static uintptr_t unhash_address(uintptr_t x) {
+  if (sizeof (x) < 8) {
+    x ^= x >> 16;
+    x *= 0x43021123U;
+    x ^= x >> 15 ^ x >> 30;
+    x *= 0x1d69e2a5U;
+    x ^= x >> 16;
+    return x;
+  } else {
+    x ^= x >> 31 ^ x >> 62;
+    x *= 0x319642b2d24d8ec3U;
+    x ^= x >> 27 ^ x >> 54;
+    x *= 0x96de1b173f119089U;
+    x ^= x >> 30 ^ x >> 60;
+    return x;
+  }
+}
+
+#endif // ADDRESS_HASH_H
diff --git a/address-map.h b/address-map.h
new file mode 100644
index 000000000..4b6b0c47f
--- /dev/null
+++ b/address-map.h
@@ -0,0 +1,210 @@
+#ifndef ADDRESS_MAP_H
+#define ADDRESS_MAP_H
+
+#include <malloc.h>
+#include <stdint.h>
+#include <string.h>
+
+#include "address-hash.h"
+
+struct hash_map_entry {
+  uintptr_t k;
+  uintptr_t v;
+};
+
+struct hash_map {
+  struct hash_map_entry *data;
+  size_t size;         // total number of slots
+  size_t n_items;      // number of items in set
+  uint8_t *bits;        // bitvector indicating set slots
+};
+
+static void hash_map_clear(struct hash_map *map) {
+  memset(map->bits, 0, map->size / 8);
+  map->n_items = 0;
+}
+  
+// Size must be a power of 2.
+static void hash_map_init(struct hash_map *map, size_t size) {
+  map->size = size;
+  map->data = malloc(sizeof(struct hash_map_entry) * size);
+  map->bits = malloc(size / 8);
+  hash_map_clear(map);
+}
+static void hash_map_destroy(struct hash_map *map) {
+  free(map->data);
+  free(map->bits);
+}
+
+static size_t hash_map_slot_index(struct hash_map *map, size_t idx) {
+  return idx & (map->size - 1);
+}
+static struct hash_map_entry* hash_map_slot_entry(struct hash_map *map,
+                                                  size_t idx) {
+  return &map->data[hash_map_slot_index(map, idx)];
+}
+static int hash_map_slot_is_empty(struct hash_map *map, size_t idx) {
+  idx = hash_map_slot_index(map, idx);
+  return (map->bits[idx / 8] & (1 << (idx % 8))) == 0;
+}
+static void hash_map_slot_acquire(struct hash_map *map, size_t idx) {
+  idx = hash_map_slot_index(map, idx);
+  map->bits[idx / 8] |= (1 << (idx % 8));
+  map->n_items++;
+}
+static void hash_map_slot_release(struct hash_map *map, size_t idx) {
+  idx = hash_map_slot_index(map, idx);
+  map->bits[idx / 8] &= ~(1 << (idx % 8));
+  map->n_items--;
+}
+static size_t hash_map_slot_distance(struct hash_map *map, size_t idx) {
+  return hash_map_slot_index(map, idx - hash_map_slot_entry(map, idx)->k);
+}
+static int hash_map_should_shrink(struct hash_map *map) {
+  return map->size > 8 && map->n_items <= (map->size >> 3);
+}
+static int hash_map_should_grow(struct hash_map *map) {
+  return map->n_items >= map->size - (map->size >> 3);
+}
+
+static void hash_map_do_insert(struct hash_map *map, uintptr_t k, uintptr_t v) 
{
+  size_t displacement = 0;
+  while (!hash_map_slot_is_empty(map, k + displacement)
+         && displacement < hash_map_slot_distance(map, k + displacement))
+    displacement++;
+  while (!hash_map_slot_is_empty(map, k + displacement)
+         && displacement == hash_map_slot_distance(map, k + displacement)) {
+    if (hash_map_slot_entry(map, k + displacement)->k == k) {
+      hash_map_slot_entry(map, k + displacement)->v = v;
+      return;
+    }
+    displacement++;
+  }
+  size_t idx = k + displacement;
+  size_t slots_to_move = 0;
+  while (!hash_map_slot_is_empty(map, idx + slots_to_move))
+    slots_to_move++;
+  hash_map_slot_acquire(map, idx + slots_to_move);
+  while (slots_to_move--)
+    *hash_map_slot_entry(map, idx + slots_to_move + 1) =
+      *hash_map_slot_entry(map, idx + slots_to_move);
+  *hash_map_slot_entry(map, idx) = (struct hash_map_entry){ k, v };
+}
+
+static void hash_map_populate(struct hash_map *dst, struct hash_map *src) {
+  for (size_t i = 0; i < src->size; i++)
+    if (!hash_map_slot_is_empty(src, i))
+      hash_map_do_insert(dst, hash_map_slot_entry(src, i)->k,
+                         hash_map_slot_entry(src, i)->v);
+}
+static void hash_map_grow(struct hash_map *map) {
+  struct hash_map fresh;
+  hash_map_init(&fresh, map->size << 1);
+  hash_map_populate(&fresh, map);
+  hash_map_destroy(map);
+  memcpy(map, &fresh, sizeof(fresh));
+}
+static void hash_map_shrink(struct hash_map *map) {
+  struct hash_map fresh;
+  hash_map_init(&fresh, map->size >> 1);
+  hash_map_populate(&fresh, map);
+  hash_map_destroy(map);
+  memcpy(map, &fresh, sizeof(fresh));
+}
+
+static void hash_map_insert(struct hash_map *map, uintptr_t k, uintptr_t v) {
+  if (hash_map_should_grow(map))
+    hash_map_grow(map);
+  hash_map_do_insert(map, k, v);
+}
+static void hash_map_remove(struct hash_map *map, uintptr_t k) {
+  size_t slot = k;
+  while (!hash_map_slot_is_empty(map, slot) && hash_map_slot_entry(map, 
slot)->k != k)
+    slot++;
+  if (hash_map_slot_is_empty(map, slot))
+    __builtin_trap();
+  while (!hash_map_slot_is_empty(map, slot + 1)
+         && hash_map_slot_distance(map, slot + 1)) {
+    *hash_map_slot_entry(map, slot) = *hash_map_slot_entry(map, slot + 1);
+    slot++;
+  }
+  hash_map_slot_release(map, slot);
+  if (hash_map_should_shrink(map))
+    hash_map_shrink(map);
+}
+static int hash_map_contains(struct hash_map *map, uintptr_t k) {
+  for (size_t slot = k; !hash_map_slot_is_empty(map, slot); slot++) {
+    if (hash_map_slot_entry(map, slot)->k == k)
+      return 1;
+    if (hash_map_slot_distance(map, slot) < (slot - k))
+      return 0;
+  }
+  return 0;
+}
+static uintptr_t hash_map_lookup(struct hash_map *map, uintptr_t k, uintptr_t 
default_) {
+  for (size_t slot = k; !hash_map_slot_is_empty(map, slot); slot++) {
+    if (hash_map_slot_entry(map, slot)->k == k)
+      return hash_map_slot_entry(map, slot)->v;
+    if (hash_map_slot_distance(map, slot) < (slot - k))
+      break;
+  }
+  return default_;
+}
+static inline void hash_map_for_each (struct hash_map *map,
+                                      void (*f)(uintptr_t, uintptr_t, void*),
+                                      void *data) 
__attribute__((always_inline));
+static inline void hash_map_for_each(struct hash_map *map,
+                                     void (*f)(uintptr_t, uintptr_t, void*),
+                                     void *data) {
+  for (size_t i = 0; i < map->size; i++)
+    if (!hash_map_slot_is_empty(map, i))
+      f(hash_map_slot_entry(map, i)->k, hash_map_slot_entry(map, i)->v, data);
+}
+  
+struct address_map {
+  struct hash_map hash_map;
+};
+
+static void address_map_init(struct address_map *map) {
+  hash_map_init(&map->hash_map, 8);
+}
+static void address_map_destroy(struct address_map *map) {
+  hash_map_destroy(&map->hash_map);
+}
+static void address_map_clear(struct address_map *map) {
+  hash_map_clear(&map->hash_map);
+}
+  
+static void address_map_add(struct address_map *map, uintptr_t addr, uintptr_t 
v) {
+  hash_map_insert(&map->hash_map, hash_address(addr), v);
+}
+static void address_map_remove(struct address_map *map, uintptr_t addr) {
+  hash_map_remove(&map->hash_map, hash_address(addr));
+}
+static int address_map_contains(struct address_map *map, uintptr_t addr) {
+  return hash_map_contains(&map->hash_map, hash_address(addr));
+}
+static uintptr_t address_map_lookup(struct address_map *map, uintptr_t addr,
+                                 uintptr_t default_) {
+  return hash_map_lookup(&map->hash_map, hash_address(addr), default_);
+}
+
+struct address_map_for_each_data {
+  void (*f)(uintptr_t, uintptr_t, void *);
+  void *data;
+};
+static void address_map_do_for_each(uintptr_t k, uintptr_t v, void *data) {
+  struct address_map_for_each_data *for_each_data = data;
+  for_each_data->f(unhash_address(k), v, for_each_data->data);
+}
+static inline void address_map_for_each (struct address_map *map,
+                                         void (*f)(uintptr_t, uintptr_t, 
void*),
+                                         void *data) 
__attribute__((always_inline));
+static inline void address_map_for_each (struct address_map *map,
+                                         void (*f)(uintptr_t, uintptr_t, 
void*),
+                                         void *data) {
+  struct address_map_for_each_data for_each_data = { f, data };
+  hash_map_for_each(&map->hash_map, address_map_do_for_each, &for_each_data);
+}
+
+#endif // ADDRESS_MAP_H
diff --git a/address-set.h b/address-set.h
new file mode 100644
index 000000000..5bbcb6168
--- /dev/null
+++ b/address-set.h
@@ -0,0 +1,193 @@
+#ifndef ADDRESS_SET_H
+#define ADDRESS_SET_H
+
+#include <malloc.h>
+#include <stdint.h>
+#include <string.h>
+
+#include "address-hash.h"
+
+struct hash_set {
+  uintptr_t *data;
+  size_t size;         // total number of slots
+  size_t n_items;      // number of items in set
+  uint8_t *bits;        // bitvector indicating set slots
+};
+
+static void hash_set_clear(struct hash_set *set) {
+  memset(set->bits, 0, set->size / 8);
+  set->n_items = 0;
+}
+  
+// Size must be a power of 2.
+static void hash_set_init(struct hash_set *set, size_t size) {
+  set->size = size;
+  set->data = malloc(sizeof(uintptr_t) * size);
+  set->bits = malloc(size / 8);
+  hash_set_clear(set);
+}
+static void hash_set_destroy(struct hash_set *set) {
+  free(set->data);
+  free(set->bits);
+}
+
+static size_t hash_set_slot_index(struct hash_set *set, size_t idx) {
+  return idx & (set->size - 1);
+}
+static int hash_set_slot_is_empty(struct hash_set *set, size_t idx) {
+  idx = hash_set_slot_index(set, idx);
+  return (set->bits[idx / 8] & (1 << (idx % 8))) == 0;
+}
+static uintptr_t hash_set_slot_ref(struct hash_set *set, size_t idx) {
+  return set->data[hash_set_slot_index(set, idx)];
+}
+static void hash_set_slot_set(struct hash_set *set, size_t idx, uintptr_t v) {
+  set->data[hash_set_slot_index(set, idx)] = v;
+}
+static void hash_set_slot_acquire(struct hash_set *set, size_t idx) {
+  idx = hash_set_slot_index(set, idx);
+  set->bits[idx / 8] |= (1 << (idx % 8));
+  set->n_items++;
+}
+static void hash_set_slot_release(struct hash_set *set, size_t idx) {
+  idx = hash_set_slot_index(set, idx);
+  set->bits[idx / 8] &= ~(1 << (idx % 8));
+  set->n_items--;
+}
+static size_t hash_set_slot_distance(struct hash_set *set, size_t idx) {
+  return hash_set_slot_index(set, idx - hash_set_slot_ref(set, idx));
+}
+static int hash_set_should_shrink(struct hash_set *set) {
+  return set->size > 8 && set->n_items <= (set->size >> 3);
+}
+static int hash_set_should_grow(struct hash_set *set) {
+  return set->n_items >= set->size - (set->size >> 3);
+}
+
+static void hash_set_do_insert(struct hash_set *set, uintptr_t v) {
+  size_t displacement = 0;
+  while (!hash_set_slot_is_empty(set, v + displacement)
+         && displacement < hash_set_slot_distance(set, v + displacement))
+    displacement++;
+  while (!hash_set_slot_is_empty(set, v + displacement)
+         && displacement == hash_set_slot_distance(set, v + displacement)) {
+    if (hash_set_slot_ref(set, v + displacement) == v)
+      return;
+    displacement++;
+  }
+  size_t idx = v + displacement;
+  size_t slots_to_move = 0;
+  while (!hash_set_slot_is_empty(set, idx + slots_to_move))
+    slots_to_move++;
+  hash_set_slot_acquire(set, idx + slots_to_move);
+  while (slots_to_move--)
+    hash_set_slot_set(set, idx + slots_to_move + 1,
+                      hash_set_slot_ref(set, idx + slots_to_move));
+  hash_set_slot_set(set, idx, v);
+}
+
+static void hash_set_populate(struct hash_set *dst, struct hash_set *src) {
+  for (size_t i = 0; i < src->size; i++)
+    if (!hash_set_slot_is_empty(src, i))
+      hash_set_do_insert(dst, hash_set_slot_ref(src, i));
+}
+static void hash_set_grow(struct hash_set *set) {
+  struct hash_set fresh;
+  hash_set_init(&fresh, set->size << 1);
+  hash_set_populate(&fresh, set);
+  hash_set_destroy(set);
+  memcpy(set, &fresh, sizeof(fresh));
+}
+static void hash_set_shrink(struct hash_set *set) {
+  struct hash_set fresh;
+  hash_set_init(&fresh, set->size >> 1);
+  hash_set_populate(&fresh, set);
+  hash_set_destroy(set);
+  memcpy(set, &fresh, sizeof(fresh));
+}
+
+static void hash_set_insert(struct hash_set *set, uintptr_t v) {
+  if (hash_set_should_grow(set))
+    hash_set_grow(set);
+  hash_set_do_insert(set, v);
+}
+
+static void hash_set_remove(struct hash_set *set, uintptr_t v) {
+  size_t slot = v;
+  while (!hash_set_slot_is_empty(set, slot) && hash_set_slot_ref(set, slot) != 
v)
+    slot++;
+  if (hash_set_slot_is_empty(set, slot))
+    __builtin_trap();
+  while (!hash_set_slot_is_empty(set, slot + 1)
+         && hash_set_slot_distance(set, slot + 1)) {
+    hash_set_slot_set(set, slot, hash_set_slot_ref(set, slot + 1));
+    slot++;
+  }
+  hash_set_slot_release(set, slot);
+  if (hash_set_should_shrink(set))
+    hash_set_shrink(set);
+}
+static int hash_set_contains(struct hash_set *set, uintptr_t v) {
+  for (size_t slot = v; !hash_set_slot_is_empty(set, slot); slot++) {
+    if (hash_set_slot_ref(set, slot) == v)
+      return 1;
+    if (hash_set_slot_distance(set, slot) < (slot - v))
+      return 0;
+  }
+  return 0;
+}
+static inline void hash_set_for_each (struct hash_set *set,
+                                      void (*f)(uintptr_t, void*), void *data) 
__attribute__((always_inline));
+static inline void hash_set_for_each(struct hash_set *set,
+                                     void (*f)(uintptr_t, void*), void *data) {
+  for (size_t i = 0; i < set->size; i++)
+    if (!hash_set_slot_is_empty(set, i))
+      f(hash_set_slot_ref(set, i), data);
+}
+  
+struct address_set {
+  struct hash_set hash_set;
+};
+
+static void address_set_init(struct address_set *set) {
+  hash_set_init(&set->hash_set, 8);
+}
+static void address_set_destroy(struct address_set *set) {
+  hash_set_destroy(&set->hash_set);
+}
+static void address_set_clear(struct address_set *set) {
+  hash_set_clear(&set->hash_set);
+}
+  
+static void address_set_add(struct address_set *set, uintptr_t addr) {
+  hash_set_insert(&set->hash_set, hash_address(addr));
+}
+static void address_set_remove(struct address_set *set, uintptr_t addr) {
+  hash_set_remove(&set->hash_set, hash_address(addr));
+}
+static int address_set_contains(struct address_set *set, uintptr_t addr) {
+  return hash_set_contains(&set->hash_set, hash_address(addr));
+}
+static void address_set_union(struct address_set *set, struct address_set 
*other) {
+  while (set->hash_set.size < other->hash_set.size)
+    hash_set_grow(&set->hash_set);
+  hash_set_populate(&set->hash_set, &other->hash_set);
+}
+
+struct address_set_for_each_data {
+  void (*f)(uintptr_t, void *);
+  void *data;
+};
+static void address_set_do_for_each(uintptr_t v, void *data) {
+  struct address_set_for_each_data *for_each_data = data;
+  for_each_data->f(unhash_address(v), for_each_data->data);
+}
+static inline void address_set_for_each (struct address_set *set,
+                                         void (*f)(uintptr_t, void*), void 
*data) __attribute__((always_inline));
+static inline void address_set_for_each (struct address_set *set,
+                                         void (*f)(uintptr_t, void*), void 
*data) {
+  struct address_set_for_each_data for_each_data = { f, data };
+  hash_set_for_each(&set->hash_set, address_set_do_for_each, &for_each_data);
+}
+
+#endif // ADDRESS_SET_H
diff --git a/test-address-map.c b/test-address-map.c
new file mode 100644
index 000000000..abe11c4b0
--- /dev/null
+++ b/test-address-map.c
@@ -0,0 +1,109 @@
+#include <stdio.h>
+
+#include "address-map.h"
+
+#define COUNT (1000 * 1000)
+
+static void add_to_other(uintptr_t addr, uintptr_t val, void *data) {
+  struct address_map *other = data;
+  if (addr >= COUNT)
+    fprintf(stdout, "unexpected address: %zu\n", addr);
+  if (address_map_contains(other, addr))
+    fprintf(stdout, "missing: %zu\n", addr);
+  address_map_add(other, addr, val);
+}
+
+int main(int argc, char *arv[]) {
+  struct address_map set;
+  address_map_init(&set);
+  for (size_t i = 0; i < COUNT; i++)
+    address_map_add(&set, i, -i);
+  fprintf(stdout, "after initial add, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  for (size_t i = 0; i < COUNT; i++) {
+    if (!address_map_contains(&set, i)) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+    if (address_map_lookup(&set, i, -1) != -i) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+  }
+  for (size_t i = COUNT; i < COUNT * 2; i++) {
+    if (address_map_contains(&set, i)) {
+      fprintf(stdout, "unexpectedly present: %zu\n", i);
+      return 1;
+    }
+  }
+  address_map_clear(&set);
+  fprintf(stdout, "after clear, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  for (size_t i = 0; i < COUNT; i++)
+    address_map_add(&set, i, 0);
+  // Now update.
+  fprintf(stdout, "after re-add, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  for (size_t i = 0; i < COUNT; i++)
+    address_map_add(&set, i, i + 1);
+  fprintf(stdout, "after idempotent re-add, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  for (size_t i = 0; i < COUNT; i++) {
+    if (!address_map_contains(&set, i)) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+    if (address_map_lookup(&set, i, -1) != i + 1) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+  }
+  for (size_t i = 0; i < COUNT; i++)
+    address_map_remove(&set, i);
+  fprintf(stdout, "after one-by-one removal, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  for (size_t i = COUNT; i < 2 * COUNT; i++) {
+    if (address_map_contains(&set, i)) {
+      fprintf(stdout, "unexpectedly present: %zu\n", i);
+      return 1;
+    }
+  }
+  for (size_t i = 0; i < COUNT; i++)
+    address_map_add(&set, i, i + 2);
+  struct address_map set2;
+  address_map_init(&set2);
+  address_map_for_each(&set, add_to_other, &set2);
+  fprintf(stdout, "after for-each set, %zu/%zu\n", set2.hash_map.n_items,
+          set2.hash_map.size);
+  for (size_t i = 0; i < COUNT; i++) {
+    if (address_map_lookup(&set2, i, -1) != i + 2) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+  }
+  address_map_destroy(&set2);
+
+  size_t burnin = 1000 * 1000 * 1000 / COUNT;
+  fprintf(stdout, "beginning clear then add %zu items, %zu times\n",
+          (size_t)COUNT, burnin);
+  for (size_t j = 0; j < burnin; j++) {
+    address_map_clear(&set);
+    for (size_t i = 0; i < COUNT; i++)
+      address_map_add(&set, i, i + 3);
+  }
+  fprintf(stdout, "after burnin, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  fprintf(stdout, "beginning lookup %zu items, %zu times\n",
+          (size_t)COUNT, burnin);
+  for (size_t j = 0; j < burnin; j++) {
+    for (size_t i = 0; i < COUNT; i++) {
+      if (address_map_lookup(&set, i, -1) != i + 3) {
+        fprintf(stdout, "missing: %zu\n", i);
+        return 1;
+      }
+    }
+  }
+  fprintf(stdout, "after burnin, %zu/%zu\n", set.hash_map.n_items,
+          set.hash_map.size);
+  address_map_destroy(&set);
+}
diff --git a/test-address-set.c b/test-address-set.c
new file mode 100644
index 000000000..ecd14b674
--- /dev/null
+++ b/test-address-set.c
@@ -0,0 +1,98 @@
+#include <stdio.h>
+
+#include "address-set.h"
+
+#define COUNT (1000 * 1000)
+
+static void remove_from_other(uintptr_t addr, void *data) {
+  struct address_set *other = data;
+  if (addr >= COUNT)
+    fprintf(stdout, "unexpected address: %zu\n", addr);
+  if (!address_set_contains(other, addr))
+    fprintf(stdout, "missing: %zu\n", addr);
+  address_set_remove(other, addr);
+}
+
+int main(int argc, char *arv[]) {
+  struct address_set set;
+  address_set_init(&set);
+  for (size_t i = 0; i < COUNT; i++)
+    address_set_add(&set, i);
+  fprintf(stdout, "after initial add, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  for (size_t i = 0; i < COUNT; i++) {
+    if (!address_set_contains(&set, i)) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+  }
+  for (size_t i = COUNT; i < COUNT * 2; i++) {
+    if (address_set_contains(&set, i)) {
+      fprintf(stdout, "unexpectedly present: %zu\n", i);
+      return 1;
+    }
+  }
+  address_set_clear(&set);
+  fprintf(stdout, "after clear, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  for (size_t i = 0; i < COUNT; i++)
+    address_set_add(&set, i);
+  // Do it twice.
+  fprintf(stdout, "after re-add, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  for (size_t i = 0; i < COUNT; i++)
+    address_set_add(&set, i);
+  fprintf(stdout, "after idempotent re-add, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  for (size_t i = 0; i < COUNT; i++) {
+    if (!address_set_contains(&set, i)) {
+      fprintf(stdout, "missing: %zu\n", i);
+      return 1;
+    }
+  }
+  for (size_t i = 0; i < COUNT; i++)
+    address_set_remove(&set, i);
+  fprintf(stdout, "after one-by-one removal, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  for (size_t i = COUNT; i < 2 * COUNT; i++) {
+    if (address_set_contains(&set, i)) {
+      fprintf(stdout, "unexpectedly present: %zu\n", i);
+      return 1;
+    }
+  }
+  for (size_t i = 0; i < COUNT; i++)
+    address_set_add(&set, i);
+  struct address_set set2;
+  address_set_init(&set2);
+  address_set_union(&set2, &set);
+  fprintf(stdout, "populated set2, %zu/%zu\n", set2.hash_set.n_items,
+          set2.hash_set.size);
+  address_set_for_each(&set, remove_from_other, &set2);
+  fprintf(stdout, "after for-each removal, %zu/%zu\n", set2.hash_set.n_items,
+          set2.hash_set.size);
+  address_set_destroy(&set2);
+
+  size_t burnin = 1000 * 1000 * 1000 / COUNT;
+  fprintf(stdout, "beginning clear then add %zu items, %zu times\n",
+          (size_t)COUNT, burnin);
+  for (size_t j = 0; j < burnin; j++) {
+    address_set_clear(&set);
+    for (size_t i = 0; i < COUNT; i++)
+      address_set_add(&set, i);
+  }
+  fprintf(stdout, "after burnin, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  fprintf(stdout, "beginning lookup %zu items, %zu times\n",
+          (size_t)COUNT, burnin);
+  for (size_t j = 0; j < burnin; j++) {
+    for (size_t i = 0; i < COUNT; i++) {
+      if (!address_set_contains(&set, i)) {
+        fprintf(stdout, "missing: %zu\n", i);
+        return 1;
+      }
+    }
+  }
+  fprintf(stdout, "after burnin, %zu/%zu\n", set.hash_set.n_items,
+          set.hash_set.size);
+  address_set_destroy(&set);
+}

Reply via email to