Spinlock with exponential backoff, test before lock, randomized backoff

spinlock.h

/*
==============================================================================
Copyright 2019 Rory Jaffe
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
    http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================
*/  
#pragma once
#include <atomic>
#include <chrono>
#include <cstddef>
#include <cstdint>
#include <limits>
#include <tuple>
#include <type_traits>

#ifndef __ARM_ARCH
extern "C" {
   extern void _mm_pause();
}
#define MIDI2LR_spin_pause _mm_pause()
#else
#define MIDI2LR_spin_pause __asm__ __volatile__("yield")
#endif

   /* from http://prng.di.unimi.it/splitmix64.c, made state atomically updated and added methods to
    * satisfy concept std::uniform_random_bit_generator */

   class PRNG {
    public:
      using result_type = uint64_t;
      static result_type NextRandom() noexcept
      {
         result_type z {state_.fetch_add(0x9e3779b97f4a7c15ULL, std::memory_order_relaxed)
                        + 0x9e3779b97f4a7c15ULL};
         z = (z ^ z >> 30) * 0xbf58476d1ce4e5b9ULL;
         z = (z ^ z >> 27) * 0x94d049bb133111ebULL;
         return z ^ z >> 31;
         static_assert(std::is_unsigned_v<decltype(z)>, "Avoid sign extension");
      }
      [[nodiscard]] constexpr result_type min() const noexcept
      {
         return std::numeric_limits<result_type>::min();
      }
      [[nodiscard]] constexpr result_type max() const noexcept
      {
         return std::numeric_limits<result_type>::max();
      }
      result_type operator()() noexcept { return NextRandom(); }

    private:
      alignas(64) static std::atomic<result_type> state_;
      // NOLINTNEXTLINE(cppcoreguidelines-avoid-c-arrays, modernize-avoid-c-arrays)
      [[maybe_unused]] alignas(1) static std::byte padding_[64 - sizeof(state_)];
   };
   /* have to separate declaration from definition due to MSVC bug when std:c++latest
    * https://developercommunity.visualstudio.com/content/problem/1079261/alignas-not-accepted-when-applied-to-inline-static.html
    */
   alignas(64) inline std::atomic<PRNG::result_type> PRNG::state_ {[] {
      static_assert(sizeof(std::random_device::result_type) * 2 == sizeof(result_type)
                    && sizeof(std::random_device::result_type) == sizeof(uint32_t));
      auto rd {std::random_device {}};
      return static_cast<result_type>(rd()) << 32 | static_cast<result_type>(rd());
   }()};

   class SpinLock {
    public:
      SpinLock() noexcept = default;
      ~SpinLock() = default;
      SpinLock(const SpinLock& other) = delete;
      SpinLock(SpinLock&& other) = delete;
      SpinLock& operator=(const SpinLock& other) = delete;
      SpinLock& operator=(SpinLock&& other) = delete;
      void lock() noexcept
      { /* if uncontested, don't bother with exponential back-off */
         if (!flag_.load(std::memory_order_relaxed)
             && !flag_.exchange(true, std::memory_order_acquire)) {
            return;
         }
         using LoopT = std::uint64_t;
         const auto [bo1, bo2, bo3] {[]() noexcept -> std::tuple<LoopT, LoopT, LoopT> {
            static_assert(std::is_unsigned_v<
                              PRNG::result_type> && sizeof(PRNG::result_type) >= sizeof(uint16_t));
            const auto rn {PRNG::NextRandom()};
            return {1 + (rn & 0b11), 12 + (rn >> 2 & 0b111), 56 + (rn >> 5 & 0b1111)};
         }()};
         do {
            /* avoid cache invalidation if lock appears to be unavailable */
            for (LoopT k {0}; flag_.load(std::memory_order_relaxed); ++k) {
               using namespace std::chrono_literals;
               if (k < bo1) { continue; }
               if (k < bo2) { MIDI2LR_spin_pause; }
               else if (k < bo3) {
                  std::this_thread::yield();
               }
               else {
#pragma warning(suppress : 26447)
                  std::this_thread::sleep_for(1ms); /* never throws, analyzer false pos */
               }
            }
         } while (flag_.load(std::memory_order_relaxed)
                  || flag_.exchange(true, std::memory_order_acquire));
      }
      bool try_lock() noexcept
      {
         /* avoid cache invalidation if lock appears to be unavailable */
         if (flag_.load(std::memory_order_relaxed)) { return false; }
         return !flag_.exchange(true, std::memory_order_acquire); /* try to acquire lock */
      }
      void unlock() noexcept { flag_.store(false, std::memory_order_release); }

    private:
      alignas(64) std::atomic<bool> flag_ {false};
      // NOLINTNEXTLINE(cppcoreguidelines-avoid-c-arrays, modernize-avoid-c-arrays)
      [[maybe_unused]] alignas(1) std::byte padding_[64 - sizeof(flag_)];
   };