greensky00/skiplist

Generic lock-free Skiplist container pure C implementation, STL-style set, map

skiplist's Issues

Compile Error in MSVS

skiplist\src\skiplist.cc(42): error C2010: “.”

Visual Studio 2017 c++14

42 #define __SLD_P(args...)

SkipList Benchamrk

Test environment：Intel(R) Xeon(R) Platinum 8352Y CPU @ 2.20GHz 12 core 24G memory
Single writer and multiple readers
Randomly insert and read 100K integers
Results：
W1R1 Writer ops/s：417,792 Reader ops/s： 652,800
W1R2 Writer ops/s：209,350 Reader ops/s：513,066
W1R4 Writer ops/s：161,677 Reader ops/s：809129
Why is it so different from the Benchmark results?

Is this able to have a higher performance than the implementation in Redis?

From the information in your benchmark result, it has a higher performance than STL set with mutex. But I'd like to know if there has a higher performance than the skiplist in Redis? (BTW, skiplist in Redis is implemented without thread-safe guarantee .)

need clear() interface

for both sl_set and sl_map, we need clear() interface.

map erase error

hello, when I use sl_map :: erase(key) , program stop in skiplist_wait_for_free
how to resolve?

Is this really lock free? It seems to have spin locks.

For instance:

static inline void _sl_read_lock_an(skiplist_node* node) {
    for(;;) {
        // Wait for active writer to release the lock
        uint32_t accessing_next = 0;
        ATM_LOAD(node->accessing_next, accessing_next);
        while (accessing_next & 0xfff00000) {
            YIELD();
            ATM_LOAD(node->accessing_next, accessing_next);
        }

        ATM_FETCH_ADD(node->accessing_next, 0x1);
        ATM_LOAD(node->accessing_next, accessing_next);
        if ((accessing_next & 0xfff00000) == 0) {
            return;
        }

        ATM_FETCH_SUB(node->accessing_next, 0x1);
    }
}

This seems to be the cause of the problem here in #1 (deadlock). Do you just mean std::mutex free?