sharded_cache.h

Go to the documentation of this file.

/*
 * This program source code file is part of KiCad, a free EDA CAD application.
 *
 * Copyright The KiCad Developers, see AUTHORS.txt for contributors.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <https://www.gnu.org/licenses/>.
 */
 
#ifndef SHARDED_CACHE_H
#define SHARDED_CACHE_H
 
#include <array>
#include <shared_mutex>
#include <unordered_map>
 
#include <thread_pool.h>

template <typename KEY, typename VALUE, std::size_t SHARDS = 256>
class SHARDED_CACHE
{
public:
    bool Get( const KEY& aKey, VALUE& aValue ) const
    {
        const SHARD&                        shard = shardFor( aKey );
        std::shared_lock<std::shared_mutex> lock( shard.mutex );
 
        auto it = shard.map.find( aKey );
 
        if( it == shard.map.end() )
            return false;
 
        aValue = it->second;
        return true;
    }
 
    void Set( const KEY& aKey, const VALUE& aValue )
    {
        SHARD&                              shard = shardFor( aKey );
        std::unique_lock<std::shared_mutex> lock( shard.mutex );
 
        shard.map[aKey] = aValue;
    }
 
    void Clear()
    {
        // Clearing a node-based map frees every entry individually.  A dense board with
        // heavy custom rules can accumulate hundreds of millions of predicate-cache entries,
        // and that serial deallocation otherwise stalls DRC re-initialization for tens of
        // seconds.  The shards are independent, so fan their clears across the worker pool
        // once the cache is large enough to outweigh the scheduling overhead.
        if( Size() < PARALLEL_CLEAR_THRESHOLD )
        {
            for( SHARD& shard : m_shards )
            {
                std::unique_lock<std::shared_mutex> lock( shard.mutex );
                shard.map.clear();
            }
 
            return;
        }
 
        thread_pool& tp = GetKiCadThreadPool();
 
        tp.submit_loop( std::size_t( 0 ), SHARDS,
                [this]( std::size_t aShard )
                {
                    std::unique_lock<std::shared_mutex> lock( m_shards[aShard].mutex );
                    m_shards[aShard].map.clear();
                },
                SHARDS ).wait();
    }
 
    bool Empty() const
    {
        for( const SHARD& shard : m_shards )
        {
            std::shared_lock<std::shared_mutex> lock( shard.mutex );
 
            if( !shard.map.empty() )
                return false;
        }
 
        return true;
    }
 
private:
    static constexpr std::size_t PARALLEL_CLEAR_THRESHOLD = 100000;

    std::size_t Size() const
    {
        std::size_t total = 0;
 
        for( const SHARD& shard : m_shards )
        {
            std::shared_lock<std::shared_mutex> lock( shard.mutex );
            total += shard.map.size();
        }
 
        return total;
    }
 
    struct SHARD
    {
        mutable std::shared_mutex      mutex;
        std::unordered_map<KEY, VALUE> map;
    };
 
    SHARD& shardFor( const KEY& aKey )
    {
        return m_shards[std::hash<KEY>{}( aKey ) % SHARDS];
    }
 
    const SHARD& shardFor( const KEY& aKey ) const
    {
        return m_shards[std::hash<KEY>{}( aKey ) % SHARDS];
    }
 
    std::array<SHARD, SHARDS> m_shards;
};
 
#endif // SHARDED_CACHE_H