5 Commits

Author	SHA1	Message	Date
cucadmuh	a63ba3c9cb	fix(cover-backfill): kill idle CPU spin and offline-cache menu re-walks (#943) ... * fix(cover-backfill): snapshot-diff worklist and live-tunable parallelism Aggressive cover backfill pegged one tokio worker at ~100% on large, fully-synced libraries while the download queues stayed empty. - Take two snapshots once per pass — the DB catalog (single GROUP BY) and the on-disk cover bucket (one directory walk) — and download the set-difference. No per-row `stat` syscalls and no re-scan loop; the empty cache case (heavy backfill) costs zero per-item disk hits. - Replace the front-loaded enumeration with a producer/consumer pipeline: the producer streams the catalog in chunks and feeds misses into a bounded channel; a fixed consumer pool keeps the download/encode pools saturated. - Make cover backfill parallelism runtime-tunable from the Performance Probe (threads slider + "Run full pass now"); HTTP download and CPU encode semaphores resize live. Not surfaced in app settings. - Add a "nothing changed" idle gate (catalog signature) so a settled pass is not re-run on every library:sync-idle, mirroring the analysis worker. - Cancel promptly on switch to lazy: consumers bail on enabled/focus change and the producer feeds via try_send so a full channel cannot deadlock. - Drop the per-item recursive disk walk from the ensure hot path. * fix(cover-backfill): cheap idle gate, settle on 404s, transient retries Follow-up to the snapshot-diff backfill: stop the periodic CPU spikes and the 89%-plateau wake storm on libraries whose covers can never reach 100%. - Idle gate is now disk-free: compare only the catalog COUNT(DISTINCT) instead of walking ~all cover dirs on every sync-idle. "Did the server change?" never touches the filesystem. Clear-cache commands re-arm the gate (rearm_idle_gate) since a clear leaves the catalog total unchanged, and the settings UI wakes the active server after a clear. - Settle the gate on any completed pass regardless of pending: remaining items are unfetchable-for-now (404), so the wake/sync-idle storm stops once the fetchable set is exhausted. - Stop auto-clearing .fetch-failed markers every pass (it defeated the 30-min backoff and re-attempted 404s forever). The manual "Run full pass now" sends force=true to clear them and retry; wake/sync-idle/configure stay opportunistic. - Rate-limit sync-idle passes (60s cooldown) as defence against chatty syncs. - Retry cover downloads up to 3x with backoff on transient failures (5xx / 429 / network), but never on a real 4xx so missing covers don't hammer the server. * fix(cover-cache): stop re-walking cover dirs from offline & cache menu The settings cover-cache section polled disk usage + progress every 15s for every server, each call doing a full recursive walk of the per-server cover directory. On a fully populated cache this caused periodic CPU spikes whenever that menu was open. - mod.rs: add a 10s TTL memo around the per-server cover dir walk (cached_dir_usage_for_server), shared by cover_cache_stats_server and library_cover_progress; invalidate on clear (per-server and clear-all). - CoverCacheStrategySection: recompute on entry only; rely on the cover:library-progress and cover:cache-cleared events for live updates; drop the per-cover cover:tier-ready refresh storm; turn the 15s loop into a 5-minute safety net. * fix(cover-backfill): keep emitting progress during the whole pass The producer finishes enumerating the worklist long before the consumer pool finishes downloading it, so progress was only emitted while feeding the channel — the "offline & cache" menu and overlay then froze through the entire drain phase. Replace the per-chunk emit with a 3s progress ticker that runs for the lifetime of the pass and is aborted once the consumers drain (final accurate emit still happens at settle). * docs(changelog): record cover-backfill idle-CPU fix (PR #943) Add [1.47.0] Fixed + Changed entries and a settingsCredits line for the cover-backfill idle CPU / offline & cache menu work.	2026-06-02 04:56:34 +03:00
cucadmuh	08b6aeeb17	fix(perf): reduce idle Rust CPU and stabilize Performance Probe overlay (#939) ... * fix(perf): skip Performance Probe CPU snapshot poll on Windows Windows has no Rust CPU/RSS sampler, but the probe still invoked performance_cpu_snapshot every 2s when the modal or overlay pins were active. Skip the IPC on unsupported platforms and only poll JS-side metrics; do not start overlay polling for CPU/memory pins alone. * fix(analysis): park backfill coordinator until Advanced is configured #881 started run_coordinator_forever at app init with a 2s sleep even when disabled, waking tokio on every platform for no work. Park on Notify instead; wake on configure (enable/disable) and library sync-idle. Long sleeps use select with wake so sync-idle can interrupt COMPLETED_RECHECK waits. Investigation branch — not for merge until periodic CPU root cause is confirmed. * fix(perf): stop probe overlay flicker on live poll updates Publish CPU samples only after a valid jiffies baseline, skip no-op snapshot emits, and sync sparkline history atomically in the store. Overlay uses wall-clock sparkline time and auto-scales low CPU values. * fix(perf): cut idle Rust CPU from probe scan, cover prefetch, and storage poll Move performance_cpu_snapshot /proc work to spawn_blocking so tokio workers are not charged with probe sampling. Stop lazy cover strategy from running route prefetch disk stats every 1.5s, and slow hot-cache size refresh on Settings → Storage to 15s. * fix(perf): stabilize probe sparkline clock between live poll ticks Track sampleAt separately from updatedAt so CPU rate history and overlay sparklines only advance on real % changes, not FPS re-renders or RSS-only poll ticks. Hold CPU sparkline Y scale with a peak ref to avoid scale jumps. * fix(cover): restore lazy route prefetch without idle disk stats poll Re-enable lazy cover registry warm-up so cached WebP paths reach diskSrcCache before cells mount. Skip cover_cache_stats on every 1.5s tick — drain batches via ensure only, poll full disk usage every 30s when the registry is idle. * docs: CHANGELOG and credits for PR #939 idle CPU perf fix * fix(cover): peek before route prefetch ensure to match main responsiveness Route prefetch moved batch drain ahead of cover_cache_stats for idle CPU, which removed the accidental throttle and flooded ensure invoke slots. Use warmCoverDiskSrcBatch first (cached hits skip ensure), ensure misses only, and yield while high-priority viewport work is queued.	2026-06-01 15:50:17 +03:00
cucadmuh	77ecc8ddfe	fix(perf): keep probe monitor metrics visible on Windows (#933) ... Stop replacing the whole Monitor tab when CPU/RSS sampling is unsupported; show pipeline, UI rate, and analysis sections with an inline platform note. Also compute UI diagRates when the Rust snapshot returns supported: false.	2026-05-31 02:55:34 +03:00
cucadmuh	8ea0308dba	feat(perf): explicit toggle for live thread-group CPU polling (#891) ... * feat(perf): explicit toggle for live thread-group CPU polling Replace implicit thread-group collection (section open / pin) with a persisted checkbox so Linux /proc scans run only when the user opts in for diagnosis. Fix IPC: pass includeThreadGroups (camelCase) so Tauri maps the flag to Rust; reset the CPU baseline when the option changes so thread % deltas are valid. * docs: CHANGELOG and credits for PR #891 * fix(perf): gate CHILD_RESCAN_EVERY to Linux/macOS only Avoid dead_code warning on Windows where perf child-PID rescan is unused.	2026-05-29 19:18:21 +03:00
cucadmuh	9925771a86	feat(browse,cover,perf): lazy catalogs, cover pipeline, and Performance Probe (#890) ... * fix(cover): per-server cache stats and cover pipeline perf probe Stop count_cached_cover_ids from borrowing sibling bucket counts so Settings progress no longer attributes one server's disk cache to another. Add cover pipeline queue stats (ui ensure queue, ui vs lib HTTP/WebP semaphores) to Performance Probe overlay, with clearer ui/lib labels. * fix(browse): stabilize in-page infinite scroll and cap cover memory caches Extract useInpageScrollSentinel for album grids and song lists so sentinel reconnects do not spam loadMore during scroll. Harden useAlbumBrowseData with sync loading refs, tighter root margin, and hasMore termination when dedupe adds nothing. Pause middle-priority cover work during SQL pagination and bound diskSrc/resolve/ensure tail maps on long cold-cache sessions. * refactor(browse): unify in-page infinite scroll hooks and sentinel UI Extract shared transport (viewport ref, async pagination guards, client slice) and InpageScrollSentinel so Albums, New Releases, Artists, and song lists use one pagination pattern instead of duplicated IntersectionObserver wiring. * fix(browse): prioritize album SQL pagination over cover ensures Pause the entire webview ensure pump during grid page fetches, resume after SQL settles, add cover-queue backpressure before load-more, and re-probe the sentinel when pagination finishes so cold-cache scroll does not stall. * fix(browse): unblock covers, SQL spawn_blocking, and pagination retry Pair grid-pagination hold begin/end on stale fetches, resume the ensure pump after SQL, retry load-more when the cover backlog drains while the sentinel stays visible, and run album browse SQL on spawn_blocking so Tokio stays responsive during library_advanced_search. * feat(browse): All Albums client-slice scroll on local index (Artists-style) Load the filtered catalog once from SQLite when the library index is ready, then grow the visible grid with useClientSliceInfiniteScroll instead of offset SQL pagination per scroll. Network-only servers keep page mode. * fix(browse): lazy local catalog chunks instead of full 50k SQL fetch All Albums slice mode now loads 200 albums first, shows the grid immediately, then appends catalog chunks in the background as the user scrolls. Avoids the blocking library_advanced_search that hung the app on large libraries. * fix(browse): keep album covers loading during active grid scroll Pass high ensure priority and the in-page scroll root to AlbumCard on All Albums, stop pausing cover traffic for background catalog chunks, and never trim high-priority ensure jobs from the queue during scroll bursts. * fix(cover): viewport priority tiers and unstick ensure invoke pump All Albums uses IO-driven high/middle instead of blanket high; release only on unmount so scroll-ahead jobs are not dropped on reprioritize. Ensure queue shares one Rust flight per cover id, attaches duplicate waiters without consuming invoke slots, and times out wedged calls. Warm the first viewport slice on large grids; acquire CPU permits before spawn_blocking in cover_cache to avoid blocking-thread deadlocks. * fix(cover): wire in-page scroll root on New Releases and Lossless grids AlbumCard IO uses the same viewport id as VirtualCardGrid so cover ensure priority tracks visible in-page rows like All Albums. * fix(browse): lazy local artist catalog in 200-row chunks Replace runLocalBrowseAllArtists bulk fetch with paginated local-index chunks so large libraries do not hang on open; preserve text search, starred, letter filter, and client-slice scroll behavior. * feat(perf): add RSS and thread CPU groups to Performance Probe Extend performance_cpu_snapshot with process RSS (psysonic + WebKit children) and in-process thread CPU breakdown. Classify tokio-rt-worker and tokio-* workers separately from glib, audio/pipewire, reqwest, and other misc threads (Linux /proc only). * feat(perf): redesign Performance Probe with tabs, pins, and overlay layout Split the probe into Monitor (live metric cards, per-metric overlay pins, corner and opacity controls) and Toggles (diagnostic tree). Share live polling via perfLiveStore; label analysis/cover pipeline blocks in the HUD. * feat(perf): overlay sparklines, macOS CPU/memory, and sync fixes Add 1-minute pinned-metric sparklines with right-aligned growth and a shared poll clock. Enable macOS performance snapshots via sysinfo. Fix overlay infinite loop from unstable history snapshots, bar/sparkline tick jitter, and probe bar rescale flicker. * docs: CHANGELOG and credits for PR #890 * perf(probe): scoped CPU poll, adjustable interval, lazy thread groups Read only psysonic + WebKit children instead of the full process table; macOS uses sysctl host CPU and refreshes cached child PIDs. Add 0.5–10s poll slider (default 2s). Collect /proc thread groups only when the Monitor section is open or a thread metric is pinned. * feat(perf): three-way overlay mode switch (off / FPS / pinned) Add Monitor control for overlay visibility: hidden, FPS-only, or pinned metrics from Monitor. Live CPU poll runs only in pinned mode with live pins.	2026-05-29 04:40:31 +03:00