generic-library/cxx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

This commit establishes a new bucket_count policy in the unordered containers: The policy now allows a power-of-2 number of buckets to be requested (and that request honored) by the client. And if the number of buckets is set to a power of 2, then the constraint of the hash to the number of buckets uses & instead of %. If the client does not specify a number of buckets, then the policy remains unchanged: a prime number of buckets is selected. The growth policy is that the number of buckets is roughly doubled when needed. While growing, either the prime, or the power-of-2 strategy will be preserved. There is a small run time cost for putting in this switch. For very cheap hash functions, e.g. identity for int, the cost can be as high as 18%. However with more typical use cases, e.g. strings, the cost is in the noise level. I've measured cases with very cheap hash functions (int) that using a power-of-2 number of buckets can make look up about twice as fast. However I've also noted that a power-of-2 number of buckets is more susceptible to accidental catastrophic collisions. Though I've also noted that accidental catastrophic collisions are also possible when using a prime number of buckets (but seems far less likely). In short, this patch adds an extra tuning knob for those clients trying to get the last bit of performance squeezed out of their hash containers. Casual users of the hash containers will not notice the introduction of this tuning knob. Those clients who swear by power-of-2 hash containers can now opt-in to that strategy. Clients who prefer a prime number of buckets can continue as they have.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@159836 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Howard Hinnant
2012-07-06 17:31:14 +00:00
parent 820e00755a
commit 7a44515588
5 changed files with 68 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

103
include/__hash_table
View File

@@ -58,6 +58,27 @@ struct __hash_node
value_type __value_;
};
inline _LIBCPP_INLINE_VISIBILITY
bool
__is_power2(size_t __bc)
{
return __bc > 2 && !(__bc & (__bc - 1));
}
inline _LIBCPP_INLINE_VISIBILITY
size_t
__constrain_hash(size_t __h, size_t __bc)
{
return !(__bc & (__bc - 1)) ? __h & (__bc - 1) : __h % __bc;
}
inline _LIBCPP_INLINE_VISIBILITY
size_t
__next_pow2(size_t __n)
{
return size_t(1) << (std::numeric_limits<size_t>::digits - __clz(__n-1));
}
template <class _Tp, class _Hash, class _Equal, class _Alloc> class __hash_table;
template <class _ConstNodePtr> class __hash_const_iterator;
template <class _HashIterator> class __hash_map_iterator;
@@ -240,7 +261,7 @@ public:
__hash_local_iterator& operator++()
{
__node_ = __node_->__next_;
if (__node_ != nullptr && __node_->__hash_ % __bucket_count_ != __bucket_)
if (__node_ != nullptr && __constrain_hash(__node_->__hash_, __bucket_count_) != __bucket_)
__node_ = nullptr;
return *this;
}
@@ -330,7 +351,7 @@ public:
__hash_const_local_iterator& operator++()
{
__node_ = __node_->__next_;
if (__node_ != nullptr && __node_->__hash_ % __bucket_count_ != __bucket_)
if (__node_ != nullptr && __constrain_hash(__node_->__hash_, __bucket_count_) != __bucket_)
__node_ = nullptr;
return *this;
}
@@ -637,7 +658,7 @@ public:
template <class _Key>
_LIBCPP_INLINE_VISIBILITY
size_type bucket(const _Key& __k) const
{return hash_function()(__k) % bucket_count();}
{return __constrain_hash(hash_function()(__k), bucket_count());}
template <class _Key>
iterator find(const _Key& __x);
@@ -871,7 +892,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(__hash_table&& __u)
{
if (size() > 0)
{
__bucket_list_[__p1_.first().__next_->__hash_ % bucket_count()] =
__bucket_list_[__constrain_hash(__p1_.first().__next_->__hash_, bucket_count())] =
static_cast<__node_pointer>(_VSTD::addressof(__p1_.first()));
__u.__p1_.first().__next_ = nullptr;
__u.size() = 0;
@@ -895,7 +916,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(__hash_table&& __u,
{
__p1_.first().__next_ = __u.__p1_.first().__next_;
__u.__p1_.first().__next_ = nullptr;
__bucket_list_[__p1_.first().__next_->__hash_ % bucket_count()] =
__bucket_list_[__constrain_hash(__p1_.first().__next_->__hash_, bucket_count())] =
static_cast<__node_pointer>(_VSTD::addressof(__p1_.first()));
size() = __u.size();
__u.size() = 0;
@@ -992,7 +1013,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__move_assign(
__p1_.first().__next_ = __u.__p1_.first().__next_;
if (size() > 0)
{
__bucket_list_[__p1_.first().__next_->__hash_ % bucket_count()] =
__bucket_list_[__constrain_hash(__p1_.first().__next_->__hash_, bucket_count())] =
static_cast<__node_pointer>(_VSTD::addressof(__p1_.first()));
__u.__p1_.first().__next_ = nullptr;
__u.size() = 0;
@@ -1190,12 +1211,12 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_unique(__node_pointer __
size_t __chash;
if (__bc != 0)
{
__chash = __nd->__hash_ % __bc;
__chash = __constrain_hash(__nd->__hash_, __bc);
__ndptr = __bucket_list_[__chash];
if (__ndptr != nullptr)
{
for (__ndptr = __ndptr->__next_; __ndptr != nullptr &&
__ndptr->__hash_ % __bc == __chash;
__constrain_hash(__ndptr->__hash_, __bc) == __chash;
__ndptr = __ndptr->__next_)
{
if (key_eq()(__ndptr->__value_, __nd->__value_))
@@ -1206,10 +1227,10 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_unique(__node_pointer __
{
if (size()+1 > __bc * max_load_factor() || __bc == 0)
{
rehash(_VSTD::max<size_type>(2 * __bc + 1,
rehash(_VSTD::max<size_type>(2 * __bc + !__is_power2(__bc),
size_type(ceil(float(size() + 1) / max_load_factor()))));
__bc = bucket_count();
__chash = __nd->__hash_ % __bc;
__chash = __constrain_hash(__nd->__hash_, __bc);
}
// insert_after __bucket_list_[__chash], or __first_node if bucket is null
__node_pointer __pn = __bucket_list_[__chash];
@@ -1221,7 +1242,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_unique(__node_pointer __
// fix up __bucket_list_
__bucket_list_[__chash] = __pn;
if (__nd->__next_ != nullptr)
__bucket_list_[__nd->__next_->__hash_ % __bc] = __nd;
__bucket_list_[__constrain_hash(__nd->__next_->__hash_, __bc)] = __nd;
}
else
{
@@ -1245,11 +1266,11 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi(__node_pointer __c
size_type __bc = bucket_count();
if (size()+1 > __bc * max_load_factor() || __bc == 0)
{
rehash(_VSTD::max<size_type>(2 * __bc + 1,
rehash(_VSTD::max<size_type>(2 * __bc + !__is_power2(__bc),
size_type(ceil(float(size() + 1) / max_load_factor()))));
__bc = bucket_count();
}
size_t __chash = __cp->__hash_ % __bc;
size_t __chash = __constrain_hash(__cp->__hash_, __bc);
__node_pointer __pn = __bucket_list_[__chash];
if (__pn == nullptr)
{
@@ -1259,12 +1280,12 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi(__node_pointer __c
// fix up __bucket_list_
__bucket_list_[__chash] = __pn;
if (__cp->__next_ != nullptr)
__bucket_list_[__cp->__next_->__hash_ % __bc] = __cp;
__bucket_list_[__constrain_hash(__cp->__next_->__hash_, __bc)] = __cp;
}
else
{
for (bool __found = false; __pn->__next_ != nullptr &&
__pn->__next_->__hash_ % __bc == __chash;
__constrain_hash(__pn->__next_->__hash_, __bc) == __chash;
__pn = __pn->__next_)
{
// __found key_eq() action
@@ -1285,7 +1306,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi(__node_pointer __c
__pn->__next_ = __cp;
if (__cp->__next_ != nullptr)
{
size_t __nhash = __cp->__next_->__hash_ % __bc;
size_t __nhash = __constrain_hash(__cp->__next_->__hash_, __bc);
if (__nhash != __chash)
__bucket_list_[__nhash] = __cp;
}
@@ -1306,11 +1327,11 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi(
size_type __bc = bucket_count();
if (size()+1 > __bc * max_load_factor() || __bc == 0)
{
rehash(_VSTD::max<size_type>(2 * __bc + 1,
rehash(_VSTD::max<size_type>(2 * __bc + !__is_power2(__bc),
size_type(ceil(float(size() + 1) / max_load_factor()))));
__bc = bucket_count();
}
size_t __chash = __cp->__hash_ % __bc;
size_t __chash = __constrain_hash(__cp->__hash_, __bc);
__node_pointer __pp = __bucket_list_[__chash];
while (__pp->__next_ != __np)
__pp = __pp->__next_;
@@ -1333,12 +1354,12 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__insert_unique(const value_type& __x)
size_t __chash;
if (__bc != 0)
{
__chash = __hash % __bc;
__chash = __constrain_hash(__hash, __bc);
__nd = __bucket_list_[__chash];
if (__nd != nullptr)
{
for (__nd = __nd->__next_; __nd != nullptr &&
__nd->__hash_ % __bc == __chash;
__constrain_hash(__nd->__hash_, __bc) == __chash;
__nd = __nd->__next_)
{
if (key_eq()(__nd->__value_, __x))
@@ -1350,10 +1371,10 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__insert_unique(const value_type& __x)
__node_holder __h = __construct_node(__x, __hash);
if (size()+1 > __bc * max_load_factor() || __bc == 0)
{
rehash(_VSTD::max<size_type>(2 * __bc + 1,
rehash(_VSTD::max<size_type>(2 * __bc + !__is_power2(__bc),
size_type(ceil(float(size() + 1) / max_load_factor()))));
__bc = bucket_count();
__chash = __hash % __bc;
__chash = __constrain_hash(__hash, __bc);
}
// insert_after __bucket_list_[__chash], or __first_node if bucket is null
__node_pointer __pn = __bucket_list_[__chash];
@@ -1365,7 +1386,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__insert_unique(const value_type& __x)
// fix up __bucket_list_
__bucket_list_[__chash] = __pn;
if (__h->__next_ != nullptr)
__bucket_list_[__h->__next_->__hash_ % __bc] = __h.get();
__bucket_list_[__constrain_hash(__h->__next_->__hash_, __bc)] = __h.get();
}
else
{
@@ -1489,16 +1510,20 @@ template <class _Tp, class _Hash, class _Equal, class _Alloc>
void
__hash_table<_Tp, _Hash, _Equal, _Alloc>::rehash(size_type __n)
{
__n = __next_prime(_VSTD::max<size_type>(__n, size() > 0));
if (__n == 1)
__n = 2;
else if (__n & (__n - 1))
__n = __next_prime(__n);
size_type __bc = bucket_count();
if (__n > __bc)
__rehash(__n);
else
else if (__n < __bc)
{
__n = _VSTD::max<size_type>
(
__n,
__next_prime(size_t(ceil(float(size()) / max_load_factor())))
__is_power2(__bc) ? __next_pow2(size_t(ceil(float(size()) / max_load_factor()))) :
__next_prime(size_t(ceil(float(size()) / max_load_factor())))
);
if (__n < __bc)
__rehash(__n);
@@ -1521,13 +1546,13 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::__rehash(size_type __nbc)
__node_pointer __cp = __pp->__next_;
if (__cp != nullptr)
{
size_type __chash = __cp->__hash_ % __nbc;
size_type __chash = __constrain_hash(__cp->__hash_, __nbc);
__bucket_list_[__chash] = __pp;
size_type __phash = __chash;
for (__pp = __cp, __cp = __cp->__next_; __cp != nullptr;
__cp = __pp->__next_)
{
__chash = __cp->__hash_ % __nbc;
__chash = __constrain_hash(__cp->__hash_, __nbc);
if (__chash == __phash)
__pp = __cp;
else
@@ -1565,12 +1590,12 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::find(const _Key& __k)
size_type __bc = bucket_count();
if (__bc != 0)
{
size_t __chash = __hash % __bc;
size_t __chash = __constrain_hash(__hash, __bc);
__node_pointer __nd = __bucket_list_[__chash];
if (__nd != nullptr)
{
for (__nd = __nd->__next_; __nd != nullptr &&
__nd->__hash_ % __bc == __chash;
__constrain_hash(__nd->__hash_, __bc) == __chash;
__nd = __nd->__next_)
{
if (key_eq()(__nd->__value_, __k))
@@ -1590,12 +1615,12 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::find(const _Key& __k) const
size_type __bc = bucket_count();
if (__bc != 0)
{
size_t __chash = __hash % __bc;
size_t __chash = __constrain_hash(__hash, __bc);
__node_const_pointer __nd = __bucket_list_[__chash];
if (__nd != nullptr)
{
for (__nd = __nd->__next_; __nd != nullptr &&
__nd->__hash_ % __bc == __chash;
__constrain_hash(__nd->__hash_, __bc) == __chash;
__nd = __nd->__next_)
{
if (key_eq()(__nd->__value_, __k))
@@ -1734,7 +1759,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::remove(const_iterator __p) _NOEXCEPT
// current node
__node_pointer __cn = const_cast<__node_pointer>(__p.__node_);
size_type __bc = bucket_count();
size_t __chash = __cn->__hash_ % __bc;
size_t __chash = __constrain_hash(__cn->__hash_, __bc);
// find previous node
__node_pointer __pn = __bucket_list_[__chash];
for (; __pn->__next_ != __cn; __pn = __pn->__next_)
@@ -1742,15 +1767,15 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::remove(const_iterator __p) _NOEXCEPT
// Fix up __bucket_list_
// if __pn is not in same bucket (before begin is not in same bucket) &&
// if __cn->__next_ is not in same bucket (nullptr is not in same bucket)
if (__pn == _VSTD::addressof(__p1_.first()) || __pn->__hash_ % __bc != __chash)
if (__pn == _VSTD::addressof(__p1_.first()) || __constrain_hash(__pn->__hash_, __bc) != __chash)
{
if (__cn->__next_ == nullptr || __cn->__next_->__hash_ % __bc != __chash)
if (__cn->__next_ == nullptr || __constrain_hash(__cn->__next_->__hash_, __bc) != __chash)
__bucket_list_[__chash] = nullptr;
}
// if __cn->__next_ is not in same bucket (nullptr is in same bucket)
if (__cn->__next_ != nullptr)
{
size_t __nhash = __cn->__next_->__hash_ % __bc;
size_t __nhash = __constrain_hash(__cn->__next_->__hash_, __bc);
if (__nhash != __chash)
__bucket_list_[__nhash] = __pn;
}
@@ -1881,10 +1906,10 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::swap(__hash_table& __u)
__p2_.swap(__u.__p2_);
__p3_.swap(__u.__p3_);
if (size() > 0)
__bucket_list_[__p1_.first().__next_->__hash_ % bucket_count()] =
__bucket_list_[__constrain_hash(__p1_.first().__next_->__hash_, bucket_count())] =
static_cast<__node_pointer>(_VSTD::addressof(__p1_.first()));
if (__u.size() > 0)
__u.__bucket_list_[__u.__p1_.first().__next_->__hash_ % __u.bucket_count()] =
__u.__bucket_list_[__constrain_hash(__u.__p1_.first().__next_->__hash_, __u.bucket_count())] =
static_cast<__node_pointer>(_VSTD::addressof(__u.__p1_.first()));
}
@@ -1898,7 +1923,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>::bucket_size(size_type __n) const
if (__np != nullptr)
{
for (__np = __np->__next_; __np != nullptr &&
__np->__hash_ % __bc == __n;
__constrain_hash(__np->__hash_, __bc) == __n;
__np = __np->__next_, ++__r)
;
}

2
test/containers/unord/unord.map/reserve.pass.cpp
View File

@@ -53,7 +53,7 @@ int main()
assert(c.bucket_count() >= 2);
test(c);
c.reserve(31);
assert(c.bucket_count() == 17);
assert(c.bucket_count() >= 16);
test(c);
}
}

2
test/containers/unord/unord.multimap/reserve.pass.cpp
View File

@@ -55,7 +55,7 @@ int main()
assert(c.bucket_count() == 3);
test(c);
c.reserve(31);
assert(c.bucket_count() == 17);
assert(c.bucket_count() >= 16);
test(c);
}
}

2
test/containers/unord/unord.multiset/reserve.pass.cpp
View File

@@ -52,7 +52,7 @@ int main()
assert(c.bucket_count() == 3);
test(c);
c.reserve(31);
assert(c.bucket_count() == 17);
assert(c.bucket_count() >= 16);
test(c);
}
}

2
test/containers/unord/unord.set/reserve.pass.cpp
View File

@@ -52,7 +52,7 @@ int main()
assert(c.bucket_count() >= 2);
test(c);
c.reserve(31);
assert(c.bucket_count() == 17);
assert(c.bucket_count() >= 16);
test(c);
}
}