diff --git a/vendored/libfuse/include/thread_pool.hpp b/vendored/libfuse/include/thread_pool.hpp index e1125fbb..eb7e8235 100644 --- a/vendored/libfuse/include/thread_pool.hpp +++ b/vendored/libfuse/include/thread_pool.hpp @@ -202,7 +202,7 @@ ThreadPool::_now_usecs() { struct timespec ts; clock_gettime(CLOCK_MONOTONIC,&ts); - return (std::uint64_t)ts.tv_sec * 1000000ULL + (std::uint64_t)ts.tv_nsec / 1000ULL; + return static_cast(ts.tv_sec) * 1000000ULL + static_cast(ts.tv_nsec) / 1000ULL; } @@ -608,7 +608,7 @@ ThreadPool::start_routine_autoscale(void *arg_) // would let all workers self-exit, draining the pool // to zero threads and stalling non-autoscale enqueues. std::size_t const min_thr = std::max(btp->_scaling_config.min_threads, - (std::size_t)1); + static_cast(1)); if(btp->_try_remove_self_if_above(pthread_self(),min_thr)) { syslog(LOG_DEBUG, @@ -697,28 +697,36 @@ ThreadPool::monitor_routine(void *arg_) // next real cycle doesn't accumulate the skipped interval's // completions into a single inflated throughput sample. std::uint64_t elapsed = _now_usecs() - t0; - if(elapsed < (std::uint64_t)interval / 2) + if(elapsed < static_cast(interval) / 2) { prev_completed = btp->_tasks_completed.load(std::memory_order_relaxed); continue; } std::uint64_t completed = btp->_tasks_completed.load(std::memory_order_relaxed); - std::uint64_t raw_throughput = completed - prev_completed; + std::uint64_t delta = completed - prev_completed; prev_completed = completed; + // Normalize to a rate (tasks per sample interval) so the + // hill-climber isn't distorted by variable elapsed times. + // The spurious-wakeup filter above accepts any cycle ≥ 50% + // of the target interval, so raw counts can vary by up to + // 2x from timing alone. Normalizing removes that noise. + double throughput_rate = static_cast(delta) * + (static_cast(interval) / static_cast(elapsed)); + // Warm up: seed the EMA with the first two samples before acting. if(warmup_samples < 2) { ema_throughput = (warmup_samples == 0) - ? (double)raw_throughput - : ema_alpha * (double)raw_throughput + (1.0 - ema_alpha) * ema_throughput; + ? throughput_rate + : ema_alpha * throughput_rate + (1.0 - ema_alpha) * ema_throughput; ++warmup_samples; continue; } double prev_ema = ema_throughput; - ema_throughput = ema_alpha * (double)raw_throughput + (1.0 - ema_alpha) * ema_throughput; + ema_throughput = ema_alpha * throughput_rate + (1.0 - ema_alpha) * ema_throughput; // Only adjust if there's meaningful work happening. // Reset direction to +1 (bias toward growth) so a stale @@ -790,7 +798,7 @@ ThreadPool::monitor_routine(void *arg_) std::uint64_t now = _now_usecs(); std::uint64_t last = btp->_last_scale_time_usecs.load(std::memory_order_relaxed); - if((now - last) < (std::uint64_t)btp->_scaling_config.cooldown_usecs) + if((now - last) < static_cast(btp->_scaling_config.cooldown_usecs)) continue; // Re-read thread count under _scaling_mutex so the bound @@ -962,7 +970,7 @@ ThreadPool::_maybe_grow_on_pressure() { std::uint64_t now = _now_usecs(); std::uint64_t last = _last_scale_time_usecs.load(std::memory_order_relaxed); - if((now - last) < (std::uint64_t)_scaling_config.cooldown_usecs) + if((now - last) < static_cast(_scaling_config.cooldown_usecs)) return; } if(thread_count() >= _scaling_config.max_threads) @@ -974,7 +982,7 @@ ThreadPool::_maybe_grow_on_pressure() std::uint64_t now = _now_usecs(); std::uint64_t last = _last_scale_time_usecs.load(std::memory_order_relaxed); - if((now - last) < (std::uint64_t)_scaling_config.cooldown_usecs) + if((now - last) < static_cast(_scaling_config.cooldown_usecs)) return; // Read count under _scaling_mutex for the log message below. @@ -1251,7 +1259,7 @@ ThreadPool::enqueue_work_autoscale(ThreadPool::PToken &ptok_, // INVARIANT: BoundedQueue::try_enqueue checks the semaphore // (tryWait) *before* forwarding the item into the inner queue. // On failure the item is never moved-from, so f remains valid. - // The assert below guards against future BoundedQueue changes + // The check below guards against future BoundedQueue changes // that could violate this ordering. Func f(std::forward(func_));