Concurrency

Pad Away the False Sharing

Difficulty

Asked at Jump Trading, XTX Markets

Read the problem, hints and solution here. The editor needs a bigger screen: open this page on a laptop to write and run your code.

xxxx xxxx xxxxxx xxx xxx xxxxxxx xxx xxx xxxx xx xxxxxxxxx xxxxxxx: xx xxxxx, xx xxxxxx xxxxx, xxxxxxx xx xxxxxx xxxxx. xxxx xx xxxx xxxxxx xxxx x xxxxxx xxxxx, xxx xxxx xxxxx xx xxx xxx xxxxx.

xxx xxxxxxxx xxx xx xxx xxxxx, xx xxxx xxxxx x xxxxx xxxx, xxx x xxxx xxxx xxxxx xx xxxxx xxxx xxx xxxx xxxx xxxxxxxxxxx. xxxxx xxxxxxx xxxxxxxxxxxx xxxxx xxxxxxxxx xxxxxxxx xxxxxxxx xxxxxxx xxx xxxxxxx xxxx xxxxx xxxxxxxx xxxxxxxx xxxx xxx. xxx xxxxxxxx xxxxxx xxxxx xxx xxx xx xxxx xxxx xxxx xxxxxxxxx xxxxx.

x xxxxx<xxxxxxx> xxxxxxx xx xxxxxxxx: xxxxxxx, xxxx-xxxx, xxx xxxx. xxxxxxxxx xxxxxx<xxxxxxx>, xxxxx xx xxx xxxx xxxxx xxxx xxxx xxxxxxx xx xxx xxx xxxxx xxxx.

xxxx(xxxxx), xxxxx()
xxxxxx() — xxx xxxxxxxx xx xxxxx xxxxxxx xxxxxxx x xxx xxxxxxx x

xxxxx xx xx.

xxxxxxxx

xxxxxx<x> x;
x.xxxxxx();          // xx
xxxxxx(xxxxxx<x>);   // xxx
xxxxxx(xxxxx<x>);    // xx, xxx xxxxx xx xxx xxxx

xxxxxxxxxxx

xxx xxxxx xxxxxx xxx xxxxxx, xxx xxx xxxxxx. x xxxxxxxxx xx x xxxxxx xx xxx xxxxxx xxxxxxx, xxx xxx xxxxxx xx xxxx xxx xxx xxxxxxxx xx.
xxxx xxxxxxx xxxx xxxxx xx x xx-xxxx xxxxxxxx, xxxxxxxxxxx xxxxxxxx xxxx xx xx xxxxx xx xxxxx xxxxx, xxx xxx xxxxx xxxxxx xxxx xx xxxxxxx * xx xxxxx.
xxxxxxx x xxxxxx xxxx x xxxxxxxx xxxx xxxxx xxxx xxx xxxx xxxxx xxx xxx xxxxxxxxx xxxxx: xxx xxxxx xxxxxxx xxx xxxxx xxxxxxxx x xxxxxxxx xxx xxxxx xxxx xxxxxxxx xxxxxxxx xxx xxxxxx. xxx xxxxxxx xxxx xxx xxxx.
xxxxxxx xxx xxxxxx xxxxx xx xxx xxx xxxx xxxx. xxxxxxx xxxx xxxxx xxx xxxxxx xxx xxxxxx xx xxx xxxx xx xxxxxx xx xx.

This problem is only available to subscribers.

Language: C++Padded

#include <atomic>
#include <cstdint>
#include <cstddef>
#include <new>
#include <thread>
#include <vector>

// ...
// ...
// ...
// ...
// ...
// ...
template <std::size_t Threads>
struct Naked {
    std::atomic<long long> counters[Threads]{};

    void bump(std::size_t which) { counters[which].fetch_add(1, std::memory_order_relaxed); }
    long long total() const {
        long long sum = 0;
        for (std::size_t i = 0; i < Threads; ++i) sum += counters[i].load(std::memory_order_relaxed);
        return sum;
    }
};

// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
// ...
template <std::size_t Threads>
struct Padded {
    static constexpr std::size_t kLine = 64;

    void bump(std::size_t which) {
        // ...
        (void)which;
    }

    long long total() const {
        // ...
        return 0;
    }

    std::size_t stride() const {
        // ...
        return 0;
    }

    // ...
};

Unlock to runUpgrade

Sample tests

Unlock the problem to run it against the tests.