2 Commits

Author	SHA1	Message	Date
fanyang	4b654fc56e	perf(mpsc): sync send via noop_waker — +90% pps (249K → 474K) ... The async fn Future state machine overhead (~1.9us) dominated MpscTunnelSender::send, while RingSink operations were only ~40ns. Breakthrough: make send() an async fn that completes synchronously on the first poll for the direct (ring tunnel) path. Uses futures::task::noop_waker() to construct a dummy Context, then calls Sink trait methods (poll_ready, start_send, poll_flush) directly. RingSink always returns Ready immediately, so the waker is never invoked and the async fn completes without yielding. Channel mode (TCP/UDP/WG tunnels) still uses async send_async() with proper backpressure. Ring tunnels detected via tunnel_info() type check in PeerConn. Results (4 threads, 1400B, 15s): pps: 249K → 474K (+90%) send_msg_by_ip: 3.53us → 1.67us (-53%) send_msg_internal: 2.40us → 502ns (-79%) MpscTunnelSender::send: 1.97us → 144ns (-93%) All 207 peers:: tests pass. Netns-requiring tests (three_node, credential) unchanged (require root).	2026-06-28 21:15:05 +08:00
fanyang	79035ea972	perf(hotpath): add cpu_hotspot_ring bench and send-chain optimization plan ... - Add measure_all to PeerMap and CidrSet impl blocks (hotpath::measure_all) - Add [profile.hotpath] for samply-compatible builds (strip=false, debug=line-tables-only) - Add cpu_hotspot_ring example: 2-node ring tunnel with data-plane flooding (~234K pps) - Add plans/006-send-chain-cpu-optimization.md based on hotpath+samply 423M sample analysis Key findings: dashmap redundancy (14.9%), metrics overhead (8.3%), mpsc (14.1%) Target: reduce send_msg_internal from 3.26us to ~2us per packet	2026-06-28 12:45:05 +08:00