1
0
mirror of https://github.com/Zygo/bees.git synced 2025-05-18 13:55:44 +02:00
Zygo Blaxell 4943a07cce crucible: cache: linked-list LRU implementation
We need a better cache expiration algorithm than "make a copy of
the entire thing, sort it while holding a lock, and delete half
the items in a single burst."

Replace the Lamport clock with a double-linked list.  Each insert
or lookup operation moves the affected item to the head of the list.
Each erase operation deletes one single item at the tail of the list.

Also sort out some iterator invalidation nonsense by doing erases before
inserts instead of "insert, erase, find the inserted item again because
we invalidated the found iterator during the erase."

The new implementation adds a second word-sized member to each Value
as well as a copy of the Key.  Hopefully the enlarged size is not
a deal-breaker.

Signed-off-by: Zygo Blaxell <bees@furryterror.org>
2018-01-17 22:58:44 -05:00

298 lines
7.1 KiB
C++

#ifndef CRUCIBLE_CACHE_H
#define CRUCIBLE_CACHE_H
#include "crucible/lockset.h"
#include <algorithm>
#include <functional>
#include <map>
#include <mutex>
#include <tuple>
#include <vector>
namespace crucible {
using namespace std;
template <class Return, class... Arguments>
class LRUCache {
public:
using Key = tuple<Arguments...>;
using Func = function<Return(Arguments...)>;
private:
struct Value {
Value *fp = nullptr;
Value *bp = nullptr;
Key key;
Return ret;
Value(Key k, Return r) : key(k), ret(r) { }
// Crash early!
~Value() { fp = bp = nullptr; };
};
Func m_fn;
map<Key, Value> m_map;
LockSet<Key> m_lockset;
size_t m_max_size;
mutex m_mutex;
Value *m_last = nullptr;
void check_overflow();
void move_to_front(Value *vp);
void erase_one(Value *vp);
public:
LRUCache(Func f = Func(), size_t max_size = 100);
void func(Func f);
void max_size(size_t new_max_size);
Return operator()(Arguments... args);
Return refresh(Arguments... args);
void expire(Arguments... args);
void prune(function<bool(const Return &)> predicate);
void insert(const Return &r, Arguments... args);
void clear();
};
template <class Return, class... Arguments>
LRUCache<Return, Arguments...>::LRUCache(Func f, size_t max_size) :
m_fn(f),
m_max_size(max_size)
{
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::erase_one(Value *vp)
{
THROW_CHECK0(invalid_argument, vp);
Value *vp_bp = vp->bp;
THROW_CHECK0(runtime_error, vp_bp);
Value *vp_fp = vp->fp;
THROW_CHECK0(runtime_error, vp_fp);
vp_fp->bp = vp_bp;
vp_bp->fp = vp_fp;
// If we delete the head of the list then advance the head by one
if (vp == m_last) {
// If the head of the list is also the tail of the list then clear m_last
if (vp_fp == m_last) {
m_last = nullptr;
} else {
m_last = vp_fp;
}
}
m_map.erase(vp->key);
if (!m_last) {
THROW_CHECK0(runtime_error, m_map.empty());
} else {
THROW_CHECK0(runtime_error, !m_map.empty());
}
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::check_overflow()
{
while (m_map.size() >= m_max_size) {
THROW_CHECK0(runtime_error, m_last);
THROW_CHECK0(runtime_error, m_last->bp);
erase_one(m_last->bp);
}
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::move_to_front(Value *vp)
{
if (!m_last) {
// Create new LRU list
m_last = vp->fp = vp->bp = vp;
} else if (m_last != vp) {
Value *vp_fp = vp->fp;
Value *vp_bp = vp->bp;
if (vp_fp && vp_bp) {
// There are at least two and we are removing one that isn't m_last
// Connect adjacent nodes to each other (has no effect if vp is new), removing vp from list
vp_fp->bp = vp_bp;
vp_bp->fp = vp_fp;
} else {
// New insertion, both must be null
THROW_CHECK0(runtime_error, !vp_fp);
THROW_CHECK0(runtime_error, !vp_bp);
}
// Splice new node into list
Value *last_bp = m_last->bp;
THROW_CHECK0(runtime_error, last_bp);
// New elemnt points to both ends of list
vp->fp = m_last;
vp->bp = last_bp;
// Insert vp as fp from the end of the list
last_bp->fp = vp;
// Insert vp as bp from the second from the start of the list
m_last->bp = vp;
// Update start of list
m_last = vp;
}
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::max_size(size_t new_max_size)
{
unique_lock<mutex> lock(m_mutex);
m_max_size = new_max_size;
// FIXME: this really reduces the cache size to new_max_size - 1
// because every other time we call this method, it is immediately
// followed by insert.
check_overflow();
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::func(Func func)
{
unique_lock<mutex> lock(m_mutex);
m_fn = func;
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::clear()
{
unique_lock<mutex> lock(m_mutex);
m_map.clear();
m_last = nullptr;
}
template <class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::prune(function<bool(const Return &)> pred)
{
unique_lock<mutex> lock(m_mutex);
for (auto it = m_map.begin(); it != m_map.end(); ) {
auto next_it = ++it;
if (pred(it.second.ret)) {
erase_one(&it.second);
}
it = next_it;
}
}
template<class Return, class... Arguments>
Return
LRUCache<Return, Arguments...>::operator()(Arguments... args)
{
Key k(args...);
bool inserted = false;
// Do we have it cached?
unique_lock<mutex> lock(m_mutex);
auto found = m_map.find(k);
if (found == m_map.end()) {
// No, release cache lock and acquire key lock
lock.unlock();
auto key_lock = m_lockset.make_lock(k);
// Did item appear in cache while we were waiting for key?
lock.lock();
found = m_map.find(k);
if (found == m_map.end()) {
// No, we hold key and cache locks, but item not in cache.
// Release cache lock and call function
lock.unlock();
// Create new value
Value v(k, m_fn(args...));
// Reacquire cache lock
lock.lock();
// Make room
check_overflow();
// Reacquire cache lock and insert return value
tie(found, inserted) = m_map.insert(make_pair(k, v));
// We hold a lock on this key so we are the ones to insert it
THROW_CHECK0(runtime_error, inserted);
// Release key lock, keep the cache lock
key_lock.unlock();
}
}
// Item should be in cache now
THROW_CHECK0(runtime_error, found != m_map.end());
// (Re)insert at head of LRU
move_to_front(&(found->second));
// Make copy before releasing lock
auto rv = found->second.ret;
return rv;
}
template<class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::expire(Arguments... args)
{
Key k(args...);
unique_lock<mutex> lock(m_mutex);
auto found = m_map.find(k);
if (found != m_map.end()) {
erase_one(&found->second);
}
}
template<class Return, class... Arguments>
Return
LRUCache<Return, Arguments...>::refresh(Arguments... args)
{
expire(args...);
return operator()(args...);
}
template<class Return, class... Arguments>
void
LRUCache<Return, Arguments...>::insert(const Return &r, Arguments... args)
{
Key k(args...);
bool inserted = false;
// Do we have it cached?
unique_lock<mutex> lock(m_mutex);
auto found = m_map.find(k);
if (found == m_map.end()) {
// No, release cache lock and acquire key lock
lock.unlock();
auto key_lock = m_lockset.make_lock(k);
// Did item appear in cache while we were waiting for key?
lock.lock();
found = m_map.find(k);
if (found == m_map.end()) {
// Make room
check_overflow();
// No, we hold key and cache locks, but item not in cache.
// Insert the provided return value (no need to unlock here)
Value v(k, r);
tie(found, inserted) = m_map.insert(make_pair(k, v));
// We hold a lock on this key so we are the ones to insert it
THROW_CHECK0(runtime_error, inserted);
}
}
// Item should be in cache now
THROW_CHECK0(runtime_error, found != m_map.end());
// (Re)insert at head of LRU
move_to_front(&(found->second));
}
}
#endif // CRUCIBLE_CACHE_H