Repository navigation
docs(quest): accept the log lock profile and delete its quest - #7742
Conversation
Quest-Claim: 105be10455aa5d77
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
|
Reduced profiling evidence, part 1/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.AW5tm7, base, 1 workers{
"source_revision": "7237beb17a9405642828d954ea743d5b663172a8",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1~1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5099026692,
"workers": {
"1439025": {
"comm": "moq-quic-0",
"switches": 5948,
"cpu_ns": 3870514450,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 1205015967,
"mutex_wait_per_cpu": 0.0
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15358.6,
"recv_fps": 30717.8,
"send_mbps": 147.44256,
"recv_mbps": 294.89088,
"loss_pct": 0,
"latency_p99_ms": 3984,
"cpu_cores": 0.763365269461078,
"cpu_us_per_frame": 24.850909552802545,
"throughput_window_start_ms": 1793712361199,
"throughput_window_end_ms": 1793712366199,
"host_window_start_ms": 1793712361199,
"host_window_end_ms": 1793712366199,
"threads": 4,
"udp_drops": 0
},
"profiled": {
"send_fps": 15360,
"recv_fps": 31488,
"send_mbps": 147.456,
"recv_mbps": 302.2848,
"loss_pct": 0,
"latency_p99_ms": 3985,
"cpu_cores": 0.7664151696606785,
"cpu_us_per_frame": 24.33991265436606,
"throughput_window_start_ms": 1793712371275,
"throughput_window_end_ms": 1793712376275,
"host_window_start_ms": 1793712371275,
"host_window_end_ms": 1793712376275,
"threads": 4,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 0,
"group_share_pct": null,
"wait_by_type_us": {},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full run.AW5tm7, base, 16 workers{
"source_revision": "7237beb17a9405642828d954ea743d5b663172a8",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-10",
"moq-quic-11",
"moq-quic-12",
"moq-quic-13",
"moq-quic-14",
"moq-quic-15",
"moq-quic-2",
"moq-quic-3",
"moq-quic-4",
"moq-quic-5",
"moq-quic-6",
"moq-quic-7",
"moq-quic-8",
"moq-quic-9"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1~1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5100340036,
"workers": {
"1439451": {
"comm": "moq-quic-0",
"switches": 13598,
"cpu_ns": 477916443,
"mutex_wait_ns": 279490,
"other_futex_wait_ns": 3491,
"other_off_cpu_ns": 4534460652,
"mutex_wait_per_cpu": 0.0005848093408244588
},
"1439452": {
"comm": "moq-quic-1",
"switches": 11842,
"cpu_ns": 391567862,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4620172966,
"mutex_wait_per_cpu": 0.0
},
"1439453": {
"comm": "moq-quic-2",
"switches": 11904,
"cpu_ns": 340960779,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 60950,
"other_off_cpu_ns": 4669967674,
"mutex_wait_per_cpu": 0.0
},
"1439454": {
"comm": "moq-quic-3",
"switches": 8660,
"cpu_ns": 237012313,
"mutex_wait_ns": 7131,
"other_futex_wait_ns": 202766,
"other_off_cpu_ns": 4771557941,
"mutex_wait_per_cpu": 3.008704446506963e-05
},
"1439455": {
"comm": "moq-quic-4",
"switches": 13103,
"cpu_ns": 400334485,
"mutex_wait_ns": 10630,
"other_futex_wait_ns": 546095,
"other_off_cpu_ns": 4610797783,
"mutex_wait_per_cpu": 2.6552796219890973e-05
},
"1439456": {
"comm": "moq-quic-5",
"switches": 13463,
"cpu_ns": 448533245,
"mutex_wait_ns": 16993,
"other_futex_wait_ns": 455848,
"other_off_cpu_ns": 4563890194,
"mutex_wait_per_cpu": 3.788570900692099e-05
},
"1439457": {
"comm": "moq-quic-6",
"switches": 12564,
"cpu_ns": 371581826,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 92893,
"other_off_cpu_ns": 4639891849,
"mutex_wait_per_cpu": 0.0
},
"1439458": {
"comm": "moq-quic-7",
"switches": 12449,
"cpu_ns": 388060020,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 66625,
"other_off_cpu_ns": 4624758687,
"mutex_wait_per_cpu": 0.0
},
"1439459": {
"comm": "moq-quic-8",
"switches": 12518,
"cpu_ns": 479194839,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 486044,
"other_off_cpu_ns": 4534741605,
"mutex_wait_per_cpu": 0.0
},
"1439460": {
"comm": "moq-quic-9",
"switches": 12736,
"cpu_ns": 411446384,
"mutex_wait_ns": 13551,
"other_futex_wait_ns": 21307,
"other_off_cpu_ns": 4600457122,
"mutex_wait_per_cpu": 3.293503242940154e-05
},
"1439461": {
"comm": "moq-quic-10",
"switches": 12045,
"cpu_ns": 386481105,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4625104278,
"mutex_wait_per_cpu": 0.0
},
"1439462": {
"comm": "moq-quic-11",
"switches": 13160,
"cpu_ns": 413069803,
"mutex_wait_ns": 22490,
"other_futex_wait_ns": 15773,
"other_off_cpu_ns": 4598957296,
"mutex_wait_per_cpu": 5.4446003645538813e-05
},
"1439463": {
"comm": "moq-quic-12",
"switches": 13419,
"cpu_ns": 421463988,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 25154,
"other_off_cpu_ns": 4590727782,
"mutex_wait_per_cpu": 0.0
},
"1439464": {
"comm": "moq-quic-13",
"switches": 12895,
"cpu_ns": 439529292,
"mutex_wait_ns": 47141,
"other_futex_wait_ns": 9220,
"other_off_cpu_ns": 4572504922,
"mutex_wait_per_cpu": 0.00010725337504923335
},
"1439465": {
"comm": "moq-quic-14",
"switches": 13222,
"cpu_ns": 408671008,
"mutex_wait_ns": 3659,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4603439711,
"mutex_wait_per_cpu": 8.95341222737288e-06
},
"1439466": {
"comm": "moq-quic-15",
"switches": 12696,
"cpu_ns": 367026764,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 29290,
"other_off_cpu_ns": 4643900062,
"mutex_wait_per_cpu": 0.0
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15359.8,
"recv_fps": 31339.8,
"send_mbps": 147.45408,
"recv_mbps": 300.86208,
"loss_pct": 0,
"latency_p99_ms": 3984,
"cpu_cores": 1.2505290136857758,
"cpu_us_per_frame": 39.902265288412046,
"throughput_window_start_ms": 1793712417304,
"throughput_window_end_ms": 1793712422304,
"host_window_start_ms": 1793712417304,
"host_window_end_ms": 1793712422304,
"threads": 19,
"udp_drops": 0
},
"profiled": {
"send_fps": 15363.072614522905,
"recv_fps": 31512.3024604921,
"send_mbps": 147.4854970994199,
"recv_mbps": 302.51810362072416,
"loss_pct": 0,
"latency_p99_ms": 3984,
"cpu_cores": 1.347653933296381,
"cpu_us_per_frame": 42.76596211863523,
"throughput_window_start_ms": 1793712427882,
"throughput_window_end_ms": 1793712432881,
"host_window_start_ms": 1793712427882,
"host_window_end_ms": 1793712432881,
"threads": 19,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 47,
"group_share_pct": 11.718214343593004,
"wait_by_type_us": {
"TrackState": 347,
"GroupState": 47
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full (Written by GPT-6.1 Sol) |
|
Reduced profiling evidence, part 2/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.AW5tm7, base, 4 workers{
"source_revision": "7237beb17a9405642828d954ea743d5b663172a8",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-2",
"moq-quic-3"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1~1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5100032948,
"workers": {
"1439230": {
"comm": "moq-quic-0",
"switches": 13636,
"cpu_ns": 1315505604,
"mutex_wait_ns": 486336,
"other_futex_wait_ns": 2704741,
"other_off_cpu_ns": 3711019040,
"mutex_wait_per_cpu": 0.00036969511837974656
},
"1439231": {
"comm": "moq-quic-1",
"switches": 14442,
"cpu_ns": 1210719532,
"mutex_wait_ns": 242212,
"other_futex_wait_ns": 1950579,
"other_off_cpu_ns": 3814726512,
"mutex_wait_per_cpu": 0.00020005624225776511
},
"1439232": {
"comm": "moq-quic-2",
"switches": 13234,
"cpu_ns": 1335022680,
"mutex_wait_ns": 273481,
"other_futex_wait_ns": 3546554,
"other_off_cpu_ns": 3689921886,
"mutex_wait_per_cpu": 0.00020485120147921383
},
"1439233": {
"comm": "moq-quic-3",
"switches": 14704,
"cpu_ns": 1156308644,
"mutex_wait_ns": 587704,
"other_futex_wait_ns": 1942549,
"other_off_cpu_ns": 3869576496,
"mutex_wait_per_cpu": 0.0005082587620957022
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15359.253499222394,
"recv_fps": 30714.285714285714,
"send_mbps": 147.44883359253498,
"recv_mbps": 294.85714285714283,
"loss_pct": 0,
"latency_p99_ms": 3985,
"cpu_cores": 0.9869308261947554,
"cpu_us_per_frame": 32.13263155052692,
"throughput_window_start_ms": 1793712386969,
"throughput_window_end_ms": 1793712391470,
"host_window_start_ms": 1793712386969,
"host_window_end_ms": 1793712391470,
"threads": 7,
"udp_drops": 0
},
"profiled": {
"send_fps": 15357.471494298861,
"recv_fps": 31481.496299259856,
"send_mbps": 147.43172634526906,
"recv_mbps": 302.2223644728946,
"loss_pct": 0,
"latency_p99_ms": 3985,
"cpu_cores": 1.0158359017112804,
"cpu_us_per_frame": 32.2677134547497,
"throughput_window_start_ms": 1793712397034,
"throughput_window_end_ms": 1793712402033,
"host_window_start_ms": 1793712397034,
"host_window_end_ms": 1793712402033,
"threads": 7,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 8,
"group_share_pct": 0.5032291586071371,
"wait_by_type_us": {
"TrackState": 1503,
"other": 16,
"GroupState": 8
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full run.AW5tm7, current, 1 workers{
"source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1~1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5099434551,
"workers": {
"1439673": {
"comm": "moq-quic-0",
"switches": 8956,
"cpu_ns": 3214809670,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 1849614646,
"mutex_wait_per_cpu": 0.0
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15363.6,
"recv_fps": 31284.4,
"send_mbps": 147.49056,
"recv_mbps": 300.33024,
"loss_pct": 0,
"latency_p99_ms": 2986,
"cpu_cores": 0.642874251497006,
"cpu_us_per_frame": 20.549355317570612,
"throughput_window_start_ms": 1793712452443,
"throughput_window_end_ms": 1793712457443,
"host_window_start_ms": 1793712452443,
"host_window_end_ms": 1793712457443,
"threads": 4,
"udp_drops": 0
},
"profiled": {
"send_fps": 15363.072614522905,
"recv_fps": 31251.250250050012,
"send_mbps": 147.4854970994199,
"recv_mbps": 300.01200240048007,
"loss_pct": 0,
"latency_p99_ms": 2986,
"cpu_cores": 0.6446574105240208,
"cpu_us_per_frame": 20.628211843236226,
"throughput_window_start_ms": 1793712463003,
"throughput_window_end_ms": 1793712468002,
"host_window_start_ms": 1793712463003,
"host_window_end_ms": 1793712468002,
"threads": 4,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 0,
"group_share_pct": null,
"wait_by_type_us": {},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full (Written by GPT-6.1 Sol) |
|
Reduced profiling evidence, part 3/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.AW5tm7, current, 16 workers{
"source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-10",
"moq-quic-11",
"moq-quic-12",
"moq-quic-13",
"moq-quic-14",
"moq-quic-15",
"moq-quic-2",
"moq-quic-3",
"moq-quic-4",
"moq-quic-5",
"moq-quic-6",
"moq-quic-7",
"moq-quic-8",
"moq-quic-9"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1~1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5098674978,
"workers": {
"1439959": {
"comm": "moq-quic-0",
"switches": 13057,
"cpu_ns": 312151867,
"mutex_wait_ns": 2591738,
"other_futex_wait_ns": 115667,
"other_off_cpu_ns": 4695901735,
"mutex_wait_per_cpu": 0.008302811144166566
},
"1439960": {
"comm": "moq-quic-1",
"switches": 14140,
"cpu_ns": 413434648,
"mutex_wait_ns": 6078633,
"other_futex_wait_ns": 351407,
"other_off_cpu_ns": 4591222457,
"mutex_wait_per_cpu": 0.01470276627613465
},
"1439961": {
"comm": "moq-quic-2",
"switches": 14209,
"cpu_ns": 405623508,
"mutex_wait_ns": 4131563,
"other_futex_wait_ns": 129125,
"other_off_cpu_ns": 4602497856,
"mutex_wait_per_cpu": 0.01018570896043826
},
"1439962": {
"comm": "moq-quic-3",
"switches": 9788,
"cpu_ns": 207983834,
"mutex_wait_ns": 1440708,
"other_futex_wait_ns": 17625,
"other_off_cpu_ns": 4799151526,
"mutex_wait_per_cpu": 0.006927019145151445
},
"1439963": {
"comm": "moq-quic-4",
"switches": 11282,
"cpu_ns": 295618025,
"mutex_wait_ns": 2884921,
"other_futex_wait_ns": 35937,
"other_off_cpu_ns": 4710180149,
"mutex_wait_per_cpu": 0.0097589482238101
},
"1439964": {
"comm": "moq-quic-5",
"switches": 12837,
"cpu_ns": 309358982,
"mutex_wait_ns": 4203357,
"other_futex_wait_ns": 847283,
"other_off_cpu_ns": 4696163544,
"mutex_wait_per_cpu": 0.013587311972729468
},
"1439965": {
"comm": "moq-quic-6",
"switches": 14486,
"cpu_ns": 485281227,
"mutex_wait_ns": 5402492,
"other_futex_wait_ns": 265905,
"other_off_cpu_ns": 4522963321,
"mutex_wait_per_cpu": 0.011132703470517724
},
"1439966": {
"comm": "moq-quic-7",
"switches": 12504,
"cpu_ns": 332216066,
"mutex_wait_ns": 3654192,
"other_futex_wait_ns": 214889,
"other_off_cpu_ns": 4674902218,
"mutex_wait_per_cpu": 0.010999443958258177
},
"1439967": {
"comm": "moq-quic-8",
"switches": 11867,
"cpu_ns": 314118378,
"mutex_wait_ns": 4006445,
"other_futex_wait_ns": 577540,
"other_off_cpu_ns": 4691839913,
"mutex_wait_per_cpu": 0.012754570507810275
},
"1439968": {
"comm": "moq-quic-9",
"switches": 12897,
"cpu_ns": 352184693,
"mutex_wait_ns": 2617438,
"other_futex_wait_ns": 18055,
"other_off_cpu_ns": 4656279998,
"mutex_wait_per_cpu": 0.007432003866221409
},
"1439969": {
"comm": "moq-quic-10",
"switches": 12533,
"cpu_ns": 293130016,
"mutex_wait_ns": 2405221,
"other_futex_wait_ns": 168912,
"other_off_cpu_ns": 4714726707,
"mutex_wait_per_cpu": 0.008205304365691434
},
"1439970": {
"comm": "moq-quic-11",
"switches": 13838,
"cpu_ns": 394295249,
"mutex_wait_ns": 4767638,
"other_futex_wait_ns": 41736,
"other_off_cpu_ns": 4612918285,
"mutex_wait_per_cpu": 0.012091543106571899
},
"1439971": {
"comm": "moq-quic-12",
"switches": 14274,
"cpu_ns": 429488658,
"mutex_wait_ns": 4846517,
"other_futex_wait_ns": 76804,
"other_off_cpu_ns": 4578043504,
"mutex_wait_per_cpu": 0.011284388795198405
},
"1439972": {
"comm": "moq-quic-13",
"switches": 10866,
"cpu_ns": 270156893,
"mutex_wait_ns": 2894176,
"other_futex_wait_ns": 49352,
"other_off_cpu_ns": 4736654137,
"mutex_wait_per_cpu": 0.010712945236603679
},
"1439973": {
"comm": "moq-quic-14",
"switches": 14395,
"cpu_ns": 456332944,
"mutex_wait_ns": 7024446,
"other_futex_wait_ns": 109072,
"other_off_cpu_ns": 4549574201,
"mutex_wait_per_cpu": 0.015393247610893506
},
"1439974": {
"comm": "moq-quic-15",
"switches": 9847,
"cpu_ns": 217682243,
"mutex_wait_ns": 2276747,
"other_futex_wait_ns": 3698,
"other_off_cpu_ns": 4789566817,
"mutex_wait_per_cpu": 0.010459038682360509
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15360,
"recv_fps": 31149.2,
"send_mbps": 147.456,
"recv_mbps": 299.03232,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 1.080877511908454,
"cpu_us_per_frame": 34.70000872922752,
"throughput_window_start_ms": 1793712508934,
"throughput_window_end_ms": 1793712513934,
"host_window_start_ms": 1793712508934,
"host_window_end_ms": 1793712513934,
"threads": 19,
"udp_drops": 0
},
"profiled": {
"send_fps": 15360,
"recv_fps": 30723.8,
"send_mbps": 147.456,
"recv_mbps": 294.94848,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 1.1929561752988045,
"cpu_us_per_frame": 38.828405838431586,
"throughput_window_start_ms": 1793712519998,
"throughput_window_end_ms": 1793712524998,
"host_window_start_ms": 1793712519998,
"host_window_end_ms": 1793712524998,
"threads": 19,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 937,
"group_share_pct": 1.530389784561624,
"wait_by_type_us": {
"TrackState": 57019,
"Waiters": 2371,
"GroupState": 937,
"other": 236
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full (Written by GPT-6.1 Sol) |
|
Reduced profiling evidence, part 4/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.AW5tm7, current, 4 workers{
"source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-2",
"moq-quic-3"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1~1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5096001824,
"workers": {
"1439789": {
"comm": "moq-quic-0",
"switches": 14585,
"cpu_ns": 1125106802,
"mutex_wait_ns": 448072,
"other_futex_wait_ns": 368531,
"other_off_cpu_ns": 3898504227,
"mutex_wait_per_cpu": 0.0003982484144647452
},
"1439790": {
"comm": "moq-quic-1",
"switches": 15488,
"cpu_ns": 1056182184,
"mutex_wait_ns": 480357,
"other_futex_wait_ns": 820992,
"other_off_cpu_ns": 3966620563,
"mutex_wait_per_cpu": 0.00045480505851819975
},
"1439791": {
"comm": "moq-quic-2",
"switches": 15449,
"cpu_ns": 1041876103,
"mutex_wait_ns": 857504,
"other_futex_wait_ns": 772572,
"other_off_cpu_ns": 3978996898,
"mutex_wait_per_cpu": 0.0008230383608289747
},
"1439792": {
"comm": "moq-quic-3",
"switches": 15126,
"cpu_ns": 1050983243,
"mutex_wait_ns": 507236,
"other_futex_wait_ns": 834966,
"other_off_cpu_ns": 3971085378,
"mutex_wait_per_cpu": 0.0004826299594959384
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15360.472094418885,
"recv_fps": 31135.027005401083,
"send_mbps": 147.46053210642128,
"recv_mbps": 298.8962592518504,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 0.8543265539335413,
"cpu_us_per_frame": 27.43940301658768,
"throughput_window_start_ms": 1793712477461,
"throughput_window_end_ms": 1793712482460,
"host_window_start_ms": 1793712477461,
"host_window_end_ms": 1793712482460,
"threads": 7,
"udp_drops": 0
},
"profiled": {
"send_fps": 15356.587425016662,
"recv_fps": 30710.064430126637,
"send_mbps": 147.42323928015995,
"recv_mbps": 294.8166185292157,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 0.8780154762203788,
"cpu_us_per_frame": 28.590479782876773,
"throughput_window_start_ms": 1793712488520,
"throughput_window_end_ms": 1793712493021,
"host_window_start_ms": 1793712488520,
"host_window_end_ms": 1793712493021,
"threads": 7,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 26,
"group_share_pct": 1.1338021750686496,
"wait_by_type_us": {
"TrackState": 2041,
"other": 146,
"GroupState": 26,
"Waiters": 9
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full run.nLyO6T, base, 1 workers{
"source_revision": "d854bf98b1a9daee5c65a841da2d655ed6613c6d",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5097329362,
"workers": {
"1447366": {
"comm": "moq-quic-0",
"switches": 5949,
"cpu_ns": 3830091889,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 1243132024,
"mutex_wait_per_cpu": 0.0
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15363.072614522905,
"recv_fps": 31489.89797959592,
"send_mbps": 147.4854970994199,
"recv_mbps": 302.30302060412083,
"loss_pct": 0,
"latency_p99_ms": 3985,
"cpu_cores": 0.7644562844704671,
"cpu_us_per_frame": 24.27623884224082,
"throughput_window_start_ms": 1793712881161,
"throughput_window_end_ms": 1793712886160,
"host_window_start_ms": 1793712881161,
"host_window_end_ms": 1793712886160,
"threads": 4,
"udp_drops": 0
},
"profiled": {
"send_fps": 15360,
"recv_fps": 30719.8,
"send_mbps": 147.456,
"recv_mbps": 294.91008,
"loss_pct": 0,
"latency_p99_ms": 3985,
"cpu_cores": 0.7613293413173653,
"cpu_us_per_frame": 24.783017510444903,
"throughput_window_start_ms": 1793712892235,
"throughput_window_end_ms": 1793712897235,
"host_window_start_ms": 1793712892235,
"host_window_end_ms": 1793712897235,
"threads": 4,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 0,
"group_share_pct": null,
"wait_by_type_us": {},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full (Written by GPT-6.1 Sol) |
|
Reduced profiling evidence, part 5/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.nLyO6T, base, 16 workers{
"source_revision": "d854bf98b1a9daee5c65a841da2d655ed6613c6d",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-10",
"moq-quic-11",
"moq-quic-12",
"moq-quic-13",
"moq-quic-14",
"moq-quic-15",
"moq-quic-2",
"moq-quic-3",
"moq-quic-4",
"moq-quic-5",
"moq-quic-6",
"moq-quic-7",
"moq-quic-8",
"moq-quic-9"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5100214009,
"workers": {
"1447609": {
"comm": "moq-quic-0",
"switches": 12582,
"cpu_ns": 367909430,
"mutex_wait_ns": 2030827,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4643126390,
"mutex_wait_per_cpu": 0.005519910158323476
},
"1447610": {
"comm": "moq-quic-1",
"switches": 13029,
"cpu_ns": 442406806,
"mutex_wait_ns": 2733298,
"other_futex_wait_ns": 426982,
"other_off_cpu_ns": 4569810799,
"mutex_wait_per_cpu": 0.006178245820205578
},
"1447611": {
"comm": "moq-quic-2",
"switches": 10641,
"cpu_ns": 276875531,
"mutex_wait_ns": 1397843,
"other_futex_wait_ns": 29331,
"other_off_cpu_ns": 4733831864,
"mutex_wait_per_cpu": 0.005048633206955366
},
"1447612": {
"comm": "moq-quic-3",
"switches": 13013,
"cpu_ns": 477870944,
"mutex_wait_ns": 3009241,
"other_futex_wait_ns": 37945,
"other_off_cpu_ns": 4534751106,
"mutex_wait_per_cpu": 0.006297183450433848
},
"1447613": {
"comm": "moq-quic-4",
"switches": 14033,
"cpu_ns": 466309275,
"mutex_wait_ns": 2841866,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4546718047,
"mutex_wait_per_cpu": 0.006094380172901343
},
"1447614": {
"comm": "moq-quic-5",
"switches": 13488,
"cpu_ns": 415669459,
"mutex_wait_ns": 2428825,
"other_futex_wait_ns": 203020,
"other_off_cpu_ns": 4596220338,
"mutex_wait_per_cpu": 0.005843164436095845
},
"1447615": {
"comm": "moq-quic-6",
"switches": 13360,
"cpu_ns": 411080022,
"mutex_wait_ns": 2107329,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4601282432,
"mutex_wait_per_cpu": 0.005126323069039828
},
"1447616": {
"comm": "moq-quic-7",
"switches": 14279,
"cpu_ns": 551859031,
"mutex_wait_ns": 2998012,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4462186956,
"mutex_wait_per_cpu": 0.0054325685212896335
},
"1447617": {
"comm": "moq-quic-8",
"switches": 13042,
"cpu_ns": 449841059,
"mutex_wait_ns": 1417434,
"other_futex_wait_ns": 234169,
"other_off_cpu_ns": 4563985174,
"mutex_wait_per_cpu": 0.0031509662616190843
},
"1447618": {
"comm": "moq-quic-9",
"switches": 11172,
"cpu_ns": 344089129,
"mutex_wait_ns": 2866963,
"other_futex_wait_ns": 176619,
"other_off_cpu_ns": 4666245849,
"mutex_wait_per_cpu": 0.008332035970831268
},
"1447619": {
"comm": "moq-quic-10",
"switches": 11885,
"cpu_ns": 348646788,
"mutex_wait_ns": 1294563,
"other_futex_wait_ns": 34302,
"other_off_cpu_ns": 4663559052,
"mutex_wait_per_cpu": 0.0037131074903234158
},
"1447620": {
"comm": "moq-quic-11",
"switches": 12264,
"cpu_ns": 417542766,
"mutex_wait_ns": 2436230,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4595220330,
"mutex_wait_per_cpu": 0.005834683769853649
},
"1447621": {
"comm": "moq-quic-12",
"switches": 11114,
"cpu_ns": 327880851,
"mutex_wait_ns": 513830,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 4684851268,
"mutex_wait_per_cpu": 0.0015671241502298041
},
"1447622": {
"comm": "moq-quic-13",
"switches": 8434,
"cpu_ns": 218222198,
"mutex_wait_ns": 809559,
"other_futex_wait_ns": 22278,
"other_off_cpu_ns": 4790937571,
"mutex_wait_per_cpu": 0.0037097921633068693
},
"1447623": {
"comm": "moq-quic-14",
"switches": 12514,
"cpu_ns": 417776679,
"mutex_wait_ns": 2294864,
"other_futex_wait_ns": 74032,
"other_off_cpu_ns": 4594525481,
"mutex_wait_per_cpu": 0.005493039978902221
},
"1447624": {
"comm": "moq-quic-15",
"switches": 14286,
"cpu_ns": 531607146,
"mutex_wait_ns": 2541566,
"other_futex_wait_ns": 724767,
"other_off_cpu_ns": 4481733595,
"mutex_wait_per_cpu": 0.004780910149766874
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15360.672134426886,
"recv_fps": 30725.745149029808,
"send_mbps": 147.4624524904981,
"recv_mbps": 294.96715343068615,
"loss_pct": 0,
"latency_p99_ms": 3984,
"cpu_cores": 1.2494674583619318,
"cpu_us_per_frame": 40.66516376744031,
"throughput_window_start_ms": 1793712940414,
"throughput_window_end_ms": 1793712945413,
"host_window_start_ms": 1793712940414,
"host_window_end_ms": 1793712945413,
"threads": 19,
"udp_drops": 0
},
"profiled": {
"send_fps": 15356.4,
"recv_fps": 31440.8,
"send_mbps": 147.42144,
"recv_mbps": 301.83168,
"loss_pct": 0,
"latency_p99_ms": 3984,
"cpu_cores": 1.3718747286303885,
"cpu_us_per_frame": 43.633582117197676,
"throughput_window_start_ms": 1793712950472,
"throughput_window_end_ms": 1793712955472,
"host_window_start_ms": 1793712950472,
"host_window_end_ms": 1793712955472,
"threads": 19,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 599,
"group_share_pct": 1.77627530784571,
"wait_by_type_us": {
"TrackState": 32128,
"GroupState": 599,
"other": 514
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full (Written by GPT-6.1 Sol) |
|
Reduced profiling evidence, part 6/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.nLyO6T, base, 4 workers{
"source_revision": "d854bf98b1a9daee5c65a841da2d655ed6613c6d",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-2",
"moq-quic-3"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5099898846,
"workers": {
"1447454": {
"comm": "moq-quic-0",
"switches": 13311,
"cpu_ns": 1352278449,
"mutex_wait_ns": 2429281,
"other_futex_wait_ns": 722260,
"other_off_cpu_ns": 3674915004,
"mutex_wait_per_cpu": 0.0017964354913712005
},
"1447455": {
"comm": "moq-quic-1",
"switches": 14540,
"cpu_ns": 1232099355,
"mutex_wait_ns": 2292139,
"other_futex_wait_ns": 183406,
"other_off_cpu_ns": 3792063839,
"mutex_wait_per_cpu": 0.0018603524064015114
},
"1447456": {
"comm": "moq-quic-2",
"switches": 13904,
"cpu_ns": 1121990828,
"mutex_wait_ns": 1652083,
"other_futex_wait_ns": 231686,
"other_off_cpu_ns": 3902942878,
"mutex_wait_per_cpu": 0.0014724567784078178
},
"1447457": {
"comm": "moq-quic-3",
"switches": 13076,
"cpu_ns": 1360513206,
"mutex_wait_ns": 1238948,
"other_futex_wait_ns": 902358,
"other_off_cpu_ns": 3666979118,
"mutex_wait_per_cpu": 0.0009106475369265912
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15362.4,
"recv_fps": 31682.6,
"send_mbps": 147.47904,
"recv_mbps": 304.15296,
"loss_pct": 0,
"latency_p99_ms": 3985,
"cpu_cores": 0.9860039920159681,
"cpu_us_per_frame": 31.121309236488422,
"throughput_window_start_ms": 1793712906090,
"throughput_window_end_ms": 1793712911090,
"host_window_start_ms": 1793712906090,
"host_window_end_ms": 1793712911090,
"threads": 7,
"udp_drops": 0
},
"profiled": {
"send_fps": 15362.872574514904,
"recv_fps": 30726.14522904581,
"send_mbps": 147.48357671534308,
"recv_mbps": 294.9709941988398,
"loss_pct": 0,
"latency_p99_ms": 3984,
"cpu_cores": 1.028002007587146,
"cpu_us_per_frame": 33.456914296407184,
"throughput_window_start_ms": 1793712917173,
"throughput_window_end_ms": 1793712922172,
"host_window_start_ms": 1793712917173,
"host_window_end_ms": 1793712922172,
"threads": 7,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 0,
"group_share_pct": 0.0,
"wait_by_type_us": {
"TrackState": 7368,
"other": 44
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full run.nLyO6T, current, 1 workers{
"source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5104341396,
"workers": {
"1447751": {
"comm": "moq-quic-0",
"switches": 7756,
"cpu_ns": 3288776605,
"mutex_wait_ns": 0,
"other_futex_wait_ns": 0,
"other_off_cpu_ns": 1780610410,
"mutex_wait_per_cpu": 0.0
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15354.143523661407,
"recv_fps": 30717.840479893355,
"send_mbps": 147.3997778271495,
"recv_mbps": 294.8912686069762,
"loss_pct": 0,
"latency_p99_ms": 2986,
"cpu_cores": 0.658682634730539,
"cpu_us_per_frame": 21.442999391890385,
"throughput_window_start_ms": 1793712983964,
"throughput_window_end_ms": 1793712988465,
"host_window_start_ms": 1793712983964,
"host_window_end_ms": 1793712988465,
"threads": 4,
"udp_drops": 0
},
"profiled": {
"send_fps": 15363,
"recv_fps": 31128,
"send_mbps": 147.4848,
"recv_mbps": 298.8288,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 0.6548143712574848,
"cpu_us_per_frame": 21.03618514705361,
"throughput_window_start_ms": 1793712994043,
"throughput_window_end_ms": 1793712999043,
"host_window_start_ms": 1793712994043,
"host_window_end_ms": 1793712999043,
"threads": 4,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 0,
"group_share_pct": null,
"wait_by_type_us": {},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full (Written by GPT-6.1 Sol) |
|
Reduced profiling evidence, part 7/7. Full acquisition stacks and validated lock/runtime summaries from the unchanged profiling recipe. run.nLyO6T, current, 4 workers{
"source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"current_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"harness_source_revision": "b5a2e8fb7716144abd756ae81cd9a780b200be0b",
"runtime": {
"runtime": "tokio-workers",
"workers": [
"moq-quic-0",
"moq-quic-1",
"moq-quic-2",
"moq-quic-3"
]
},
"rustc": "rustc 1.98.1 (48a229cea 2026-09-01)\nbinary: rustc\ncommit-hash: 48a229ceaefd4985c50990b14116b6d856af0985\ncommit-date: 2026-09-01\nhost: x86_64-unknown-linux-gnu\nrelease: 1.98.1\nLLVM version: 22.1.8",
"cargo": "cargo 1.98.1 (797e8a9bc 2026-08-05)",
"affinity": [
0,
1,
2,
3,
4,
5,
6,
7
],
"profile": "profiling",
"features": "default + moq-relay/io-uring,moq-bench/uring; heap adds moq-relay/jemalloc",
"environment": {
"BENCH_PROFILE_TRANSPORT": "quic",
"BENCH_PROFILE_MODES": "lock",
"RUSTFLAGS": "-C force-frame-pointers=yes",
"BENCH_PROFILE_WORKERS": "1 4 16",
"BENCH_PROFILE_WORKLOAD": "video-heavy",
"BENCH_BASE": "d854bf98b1",
"BENCH_PROFILE_RUNTIME": "tokio-workers"
},
"source_dirty_patch_bytes": 0,
"untracked_sources": [],
"profiler": {
"perf": "perf version 7.3.0",
"jeprof": "jeprof (part of jemalloc 5.4.0-0-g0000000000000000000000000000000000000000)"
},
"lock": {
"version": 3,
"window_ns": 5098959050,
"workers": {
"1447910": {
"comm": "moq-quic-0",
"switches": 15357,
"cpu_ns": 1080233290,
"mutex_wait_ns": 1094352,
"other_futex_wait_ns": 455116,
"other_off_cpu_ns": 3943538391,
"mutex_wait_per_cpu": 0.0010130700563764333
},
"1447911": {
"comm": "moq-quic-1",
"switches": 14223,
"cpu_ns": 1149499578,
"mutex_wait_ns": 772379,
"other_futex_wait_ns": 717408,
"other_off_cpu_ns": 3874390646,
"mutex_wait_per_cpu": 0.0006719263014814261
},
"1447912": {
"comm": "moq-quic-2",
"switches": 14539,
"cpu_ns": 1071131502,
"mutex_wait_ns": 592793,
"other_futex_wait_ns": 58850,
"other_off_cpu_ns": 3952046518,
"mutex_wait_per_cpu": 0.0005534269124688669
},
"1447913": {
"comm": "moq-quic-3",
"switches": 16010,
"cpu_ns": 960720128,
"mutex_wait_ns": 800453,
"other_futex_wait_ns": 71264,
"other_off_cpu_ns": 4060059615,
"mutex_wait_per_cpu": 0.0008331802120835757
}
},
"idle": [],
"scope": "exact off-CPU time blocked in a futex under std Mutex::lock_contended; spinning is excluded"
},
"summaries": {
"control": {
"send_fps": 15361.030882026214,
"recv_fps": 30709.397911575204,
"send_mbps": 147.46589646745167,
"recv_mbps": 294.81021995112195,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 0.8582834331337323,
"cpu_us_per_frame": 27.948559447667385,
"throughput_window_start_ms": 1793713007742,
"throughput_window_end_ms": 1793713012243,
"host_window_start_ms": 1793713007742,
"host_window_end_ms": 1793713012243,
"threads": 7,
"udp_drops": 0
},
"profiled": {
"send_fps": 15361.253054876694,
"recv_fps": 30720.506554099087,
"send_mbps": 147.46802932681626,
"recv_mbps": 294.9168629193512,
"loss_pct": 0,
"latency_p99_ms": 2985,
"cpu_cores": 0.8759818731787702,
"cpu_us_per_frame": 28.514564746390437,
"throughput_window_start_ms": 1793713018303,
"throughput_window_end_ms": 1793713022804,
"host_window_start_ms": 1793713018303,
"host_window_end_ms": 1793713022804,
"threads": 7,
"udp_drops": 0
}
},
"attribution": {
"method": "Innermost lock<T> frame identifies the mutex. Waiters locks are separate even when outer frames mention GroupState.",
"group_wait_us": 18,
"group_share_pct": 0.5521511348086198,
"wait_by_type_us": {
"TrackState": 3148,
"GroupState": 18
},
"rounding": "Folded wait weights truncate each interval to microseconds; lock.json totals retain nanoseconds."
}
}Full Attempt preflight and limitations{
"utc": "2026-11-03T13:16:33Z",
"hostname": "desktop",
"cpus": 24,
"load": [
0.212890625,
1.0322265625,
2.3623046875
],
"disk": {
"total": 1998293323776,
"used": 1986059730944,
"free": 12233592832
},
"perf_event_paranoid": "-1",
"performance_cpus": "0-7",
"nix_shell": "impure",
"t3_running_threads": "Only this implementation child, verified via t3_thread_list allProjects=true includeSubagents=true."
}The preflight establishes the starting conditions only. No continuous independent host-idle trace was collected. The historical observations are candidate evidence, not proof that the host stayed quiet throughout the captures. Two all-project T3 checks showed only this implementation child running; other agent systems were not independently audited throughout. Refused second-current 16-worker captureThe recipe exited 1. No accepted profile or lock summary is reported for this capture. Build-tool automatic GC incidentThe first recipe invoked the installed Cargo shim; it automatically collected managed build output and shared cache data before the workload. This violated the no-shared-GC constraint despite no explicit GC command. The second recipe and local checks bypassed the shim using direct pinned Nix Cargo/rustc and BASH_ENV=/dev/null. The GC log lines are retained here to disclose the side effect: (Written by GPT-6.1 Sol) |
The repeat found no quiet window. Its one completed pre-epoch run shows the 16-worker GroupState wait on the same stack as the epoch capture, so that wait predates #6782; no valid epoch capture exists on that host yet. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Reduced profiling evidence, nixos2 repeat, part 1/2. Run {"run": "r1A", "label": "base", "workers": 1, "window": [1793732480.0, 1793732508.3], "samples": 16, "load1_max": 3.67, "load1_at_start": 2.4, "busy_cores_max": 3.595, "foreign_cores_mean": 0.07, "foreign_cores_max": 0.138, "kernel_thread_cores_max": 0.108, "agent_processes_max": 4, "agent_tools": [".gh-wrapped", "bash", "codex", "codex-code-mode", "git", "git-remote-http", "quest"], "agent_cores_max": 0.044, "foreign_builds": [], "top_foreign": [".claude-wrapped", "MainThread", "codex", "nsncd"], "control_fps": 22069.8, "control_loss_pct": 37.13474025974026, "profiled_fps": 23220.644128825767, "profiled_loss_pct": 37.2093023255814, "mutex_ms": 0.0, "idle_workers": 0, "groupstate_us": 0, "groupstate_share_pct": null, "by_lock_us": {}, "groupstate_stacks": {}}
{"run": "r1A", "label": "base", "workers": 4, "window": [1793732508.4, 1793732553.7], "samples": 25, "load1_max": 5.21, "load1_at_start": 2.85, "busy_cores_max": 6.549, "foreign_cores_mean": 0.061, "foreign_cores_max": 0.162, "kernel_thread_cores_max": 0.074, "agent_processes_max": 4, "agent_tools": ["bash", "codex", "codex-code-mode", "quest"], "agent_cores_max": 0.088, "foreign_builds": [], "top_foreign": [".claude-wrapped", ".kgx-wrapped", "MainThread", "codex", "nsncd"], "control_fps": 31483.296659331867, "control_loss_pct": 0, "profiled_fps": 30681.18195956454, "profiled_loss_pct": 0, "mutex_ms": 0.822, "idle_workers": 0, "groupstate_us": 9, "groupstate_share_pct": 1.1, "by_lock_us": {"lock<kio::State<moq_net::model::track::TrackState>>": 786, "lock<kio::State<moq_net::model::group::GroupState>>": 9}, "groupstate_stacks": {"<moq_net::lite::publisher::SubscribeServe as moq_net::lite::publisher::Request>::poll;poll_step;<kio::task::Tasks<moq_net::lite::publisher::GroupServe>>::poll;poll;poll_serve;<moq_net::model::group::Consumer>::poll_read_objects::<8>;poll_read_object;<moq_net::model::group::Consumer>::poll_read_frame;poll_resumed<moq_net::model::frame::Frame, fn(&mut moq_net::model::group::Consumer, &kio::waiter::Waiter) -> core::task::poll::Poll<core::result::Result<core::option::Option<moq_net::model::frame::Frame>, moq_net::error::Error>>>;call_mut<fn(&mut moq_net::model::group::Consumer, &kio::waiter::Waiter) -> core::task::poll::Poll<core::result::Result<core::option::Option<moq_net::model::frame::Frame>, moq_net::error::Error>>,;<moq_net::model::group::Consumer>::poll_read_frame_once;poll_expired_if_blocked<core::option::Option<moq_net::model::frame::Frame>>;<moq_net::model::group::Consumer>::poll_expiry;expiry_pending;expiry_pending;read<moq_net::model::group::GroupState>;lock<kio::State<moq_net::model::group::GroupState>>;lock<kio::State<moq_net::model::group::GroupState>>;lock;<std::sys::sync::mutex::futex::Mutex>::lock_contended;std::sys::pal::unix::futex::futex_wait;syscall": 9}}
{"run": "r1A", "label": "base", "workers": 16, "window": [1793732553.7, 1793732638.6], "samples": 43, "load1_max": 12.77, "load1_at_start": 4.88, "busy_cores_max": 8.817, "foreign_cores_mean": 0.21, "foreign_cores_max": 1.027, "kernel_thread_cores_max": 0.108, "agent_processes_max": 3, "agent_tools": ["bash"], "agent_cores_max": 0.163, "foreign_builds": [], "top_foreign": [".claude-wrapped", "MainThread", "nsncd"], "control_fps": 30295.267718284824, "control_loss_pct": 0, "profiled_fps": 30940, "profiled_loss_pct": 0, "mutex_ms": 38.819, "idle_workers": 0, "groupstate_us": 709, "groupstate_share_pct": 1.83, "by_lock_us": {"lock<kio::State<moq_net::model::track::TrackState>>": 37906, "lock<kio::State<moq_net::model::group::GroupState>>": 709, "lock<kio::State<moq_net::model::resume::Route>>": 67}, "groupstate_stacks": {"<moq_net::lite::publisher::SubscribeServe as moq_net::lite::publisher::Request>::poll;poll_step;<kio::task::Tasks<moq_net::lite::publisher::GroupServe>>::poll;poll;poll_serve;<moq_net::model::group::Consumer>::poll_read_objects::<8>;poll_read_object;<moq_net::model::group::Consumer>::poll_read_frame;poll_resumed<moq_net::model::frame::Frame, fn(&mut moq_net::model::group::Consumer, &kio::waiter::Waiter) -> core::task::poll::Poll<core::result::Result<core::option::Option<moq_net::model::frame::Frame>, moq_net::error::Error>>>;call_mut<fn(&mut moq_net::model::group::Consumer, &kio::waiter::Waiter) -> core::task::poll::Poll<core::result::Result<core::option::Option<moq_net::model::frame::Frame>, moq_net::error::Error>>,;<moq_net::model::group::Consumer>::poll_read_frame_once;poll_read_frame;poll<moq_net::model::group::GroupState, moq_net::model::group::{impl#24}::poll_read_frame::{closure_env#1}, core::result::Result<core::option::Option<moq_net::model::group::Inflate>, moq_net::error::Error>>;lock<kio::State<moq_net::model::group::GroupState>>;lock<kio::State<moq_net::model::group::GroupState>>;lock;<std::sys::sync::mutex::futex::Mutex>::lock_contended;std::sys::pal::unix::futex::futex_wait;syscall": 709}}
{"run": "r1A", "label": "current", "workers": 1, "window": [1793732638.6, 1793732665.0], "samples": 15, "load1_max": 11.83, "load1_at_start": 11.83, "busy_cores_max": 7.698, "foreign_cores_mean": 0.032, "foreign_cores_max": 0.099, "kernel_thread_cores_max": 0.02, "agent_processes_max": 3, "agent_tools": ["bash"], "agent_cores_max": 0.015, "foreign_builds": [], "top_foreign": ["nsncd"], "control_fps": 20632.926585317065, "control_loss_pct": 34.5475910693302, "profiled_fps": 21438, "profiled_loss_pct": 32.49903735078937, "mutex_ms": 0.0, "idle_workers": 0, "groupstate_us": 0, "groupstate_share_pct": null, "by_lock_us": {}, "groupstate_stacks": {}}
{"run": "r1A", "label": "current", "workers": 4, "window": [1793732665.0, 1793732727.4], "samples": 32, "load1_max": 9.07, "load1_at_start": 8.22, "busy_cores_max": 6.635, "foreign_cores_mean": 0.054, "foreign_cores_max": 0.707, "kernel_thread_cores_max": 0.236, "agent_processes_max": 3, "agent_tools": ["bash"], "agent_cores_max": 0.02, "foreign_builds": [], "top_foreign": ["MainThread", "nsncd"], "control_fps": 31239.6, "control_loss_pct": 0, "profiled_fps": 31665.4, "profiled_loss_pct": 0, "mutex_ms": 3.489, "idle_workers": 0, "groupstate_us": 0, "groupstate_share_pct": 0.0, "by_lock_us": {"lock<moq_net::model::track::TrackState>": 3322, "lock<kio::channel::Waiters>": 57, "lock<moq_net::model::resume::Route>": 16, "lock<moq_net::model::expiry::Parked>": 13}, "groupstate_stacks": {}}
{"run": "r1A", "label": "current", "workers": 16, "window": [1793732727.4, 1793732786.9], "samples": 31, "load1_max": 7.99, "load1_at_start": 7.07, "busy_cores_max": 7.542, "foreign_cores_mean": 0.046, "foreign_cores_max": 0.272, "kernel_thread_cores_max": 0.158, "agent_processes_max": 4, "agent_tools": ["bash"], "agent_cores_max": 0.025, "foreign_builds": [], "top_foreign": [".claude-wrapped", "MainThread", "nsncd"], "control_fps": 31216.52991795077, "control_loss_pct": 0, "refused": true, "refusal": ["profile: worker moq-quic-7 (2953414) has nonalternating switch records"]}r1A, base (pre-epoch), 16 workers: full
|
|
Reduced profiling evidence, nixos2 repeat, part 2/2. The pre-epoch 4-worker fold from r1A, base (pre-epoch), 4 workers: full
|
The maintainer accepted no lock-contention regression attributable to #6782: the pre-epoch run shows the same GroupState wait stack at a similar wait (709 vs 599 us). Record that in the unbounded-groups epic. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 9 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. CodeRabbit Verify
|
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 37d59f3.
[P2] Qualify the cross-host wait figures in the durable conclusion (quest/m1/unbounded-groups/README.md:117–120). The 709 µs pre-epoch wait was measured on nixos2, while the 599 µs epoch wait was on desktop. The PR's own evidence says those hosts' totals are not comparable, the quiet same-host repeat was not completed, and the original share goal has no valid measurement. Placing the two values side by side without that qualification can make the retained epic read like a controlled before/after result. Keep the maintainer's accepted decision, but state alongside it that the figures are cross-host and the comparison was inconclusive.
Static GitHub-only review. I inspected the recorded evidence and did not rerun profiles or independently verify the host activity.
There was a problem hiding this comment.
Important
The added durable conclusion pairs two host-specific wait figures as a controlled before/after, but the PR body itself says those hosts' measurements are not comparable. Qualify the figures before this lands in the epic.
Reviewed changes
- Delete the log-lock quest — removes
quest/m1/unbounded-groups/log-lock-profile.md, whose goal was ajust bench-profile lock video-heavybefore/after #6782 comparison. - Drop the epic link — removes the quest's line from the "Required" list in
quest/m1/unbounded-groups/README.md. - Record the accepted conclusion — adds four lines to the epic's storage decision stating no lock-contention regression is attributable to #6782, citing #7742.
No dangling references remain: no other tracked file mentions the deleted quest or its link, and the net diff is only these two files.
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) | 𝕏
The 709 and 599 us waits come from different hosts, so the epic now says they are not comparable while keeping the accepted conclusion. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 686ba58.
The prior cross-host qualification issue is addressed: the retained epic now names each host and says the figures are not comparable and no quiet same-host repeat was completed. No additional actionable findings in this update.
Verification: static GitHub-only inspection; I did not rerun profiles or independently verify host activity.
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
This run re-reviewed the single commit since the prior pullfrog review (686ba58), which qualifies the durable conclusion added by the quest deletion.
- Qualified the cross-host wait figures — the epic's storage decision now names
nixos2beside the 709 µs pre-epoch wait anddesktopbeside the 599 µs epoch wait, and states the figures are not comparable with no quiet same-host repeat completed.
The prior concern (cross-host figures reading as a controlled before/after) is addressed, and no dangling references to the deleted quest remain.
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) | 𝕏
Resolve the modify/delete conflicts on quests main finished (m0/ci/shared-target.md in #7715, m1/resume-model-7542.md in #7705), and drop or repoint links and Required edges to quests main deleted: m1/ietf-resume-behind.md (#7708), m1/subscription-delay.md (#7697), m1/unbounded-groups/log-lock-profile.md (#7742) and m1/moq-crypto.md (#7702). #7705 did not replay #7739's seed 17893651, which this PR had added to the resume-model quest, so it moves to m1/resume-seed-17893651.md. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Problem
The log-lock quest needed evidence that the epoch-based
kio::logchange in moq-dev/moq#6782 causes no lock-contention regression. The first attempt ondesktopwas mixed: theGroupStateshare dropped, but absolute wait rose at 16 workers (47 -> 599 us). A quiet-window repeat onnixos2found no quiet window. Its one completed pre-epoch run shows the contrary wait on the same stack before #6782.Conclusion (accepted by the maintainer, 2026-11-03): no lock-contention regression attributable to moq-dev/moq#6782. The pre-epoch run on
nixos2shows the sameGroupStatewait stack (Consumer::poll_read_frame_once -> poll_read_frame -> poll<GroupState>, 709 µs) as the epoch run ondesktop(599 µs). The figures come from different hosts and are not comparable; no quiet same-host repeat was completed. This PR deletes the quest withquest rm, which removes its link from the unbounded-groups epic. It also records the conclusion in that epic's storage decision.Approach (desktop attempt)
Ran the unchanged current
just bench-profile lock video-heavy "1 4 16" tokio-workers <base>recipe twice throughquest gate, using the pre-epoch parent and the epoch merge. Both used the current generator and harness at the claim commit, whose tree equals its main parent. The quest now records the evidence, refusal, and remaining maintainer choice.desktop: 24 logical CPUs,perf_event_paranoid=-1, initial one-minute load 0.21, no other running T3 agents in two all-project checks. The recipe pins to performance CPUs 0-7, so 16 workers oversubscribe that allowed set as the documented recipe specifies. No continuous independent host-idle trace was collected, so this is a quiet-start, no-other-agent attempt, not proof that every background activity stayed absent. All builds and captures ran sequentially; no reviewer ran during profiling.Nix dev shell, rustc 1.98.1, profiling profile (release, fat LTO, one codegen unit, line-table debug info, no stripping),
-C force-frame-pointers=yes, system allocator, default features plusmoq-relay/io-uring,moq-bench/uring; the selected runtime was verified as Tokio workers, with the requested exact count. All source dirty patches were empty and untracked-source lists empty.The workload remains 256 connections, one broadcast each, two subscriptions, 60 fps, 1200-byte frames,
group_size=59. Each control precedes its capture; capture duration is 5 s. Valid profile summaries report zero lifetime loss. This is one observation per historical revision/worker count, not a statistical estimate of a small difference.Results on desktop (first attempt)
Durable reduced evidence: part 1, part 2, part 3, part 4, part 5, part 6, part 7. These comments retain the full folded stacks, capture/runtime summaries, tool and source context, failure diagnostic, and build-tool incident log.
A =
run.AW5tm7, pre-epoch base; B =run.nLyO6T, epoch base.currenthas the same source in both runs, rebuilt through different Cargo entrypoints as disclosed below. Waits exclude spinning. The denominator is all worker std-mutex blocked time, not CPU time.Folded stacks truncate each interval to microseconds; the total mutex denominator retains nanoseconds from
lock.json. Attribute by the innermostlock<T>frame:kio::channel::Waitersis separate even if its callers mentionGroupState. Full folds and lock/runtime/control summaries are retained in the evidence comments below, so scratch cleanup does not erase the reduced evidence. The multi-gigabyteperf.datafiles remain local scratch, not durable attachments.Historical wait-by-stack diff
Every nonzero
GroupStateacquisition is below. Common leaf suffix:lock<GroupState>(orlock<kio::State<GroupState>>) ->Mutex::lock_contended->futex_wait->syscall. Full generic and scheduler frames are preserved in the evidence comments.GroupServe::poll_serve -> Handle::cache_refresh -> read<GroupState>Consumer::poll_read_frame_once -> Cursor::poll_read_frame -> poll<GroupState>GroupServe::poll_serve -> Handle::cache_refresh -> read<GroupState>Consumer::poll_read_frame_once -> Cursor::poll_read_frame -> poll<GroupState>At 4 workers the group-state wait falls to none. At 16 workers its share drops 11.72% -> 1.78%, while its absolute wait rises 47 -> 599 us. The denominator grew 0.401 -> 33.722 ms; most folded wait is
TrackState(347 -> 32128 us). Calling this an unqualified no-contention-regression result from share alone would hide the contrary absolute measurement. It also does not refute the earlier Criterion per-reader cost numbers: blocked time under this live workload measures something different.The post-epoch source still takes the state lock when the log cannot serve a frame, including its live edge: Cursor::poll_read_frame. A blocked stack alone does not identify which branch or a root cause, and one capture cannot establish that the epoch change caused the increase. Recommendation at the time (since decided below): keep the quest open and authorize alternating historical captures with warm direct-Cargo builds, recording host load per capture and verifying no other agents run throughout, preserving the existing workload, worker matrix, and validity checks. The 16-worker contrast is the unresolved signal. Make no production change from these samples.
Current-source observations and refusal
At 4 workers both current captures still have group-state waits in
poll_read_frame_once<..., Prefetch::pop_cached> -> poll<GroupState>(26 and 18 us). At 16 workers the first current capture has 937 us in total group-state waits; the full stack evidence is below. These candidate observations identify the Plan's possible epoch-versus-main investigation, but that suggestion remains conditional on reproducing the waits with established per-capture quietness; they do not establish a later regression. The 16-worker second-current capture was refused:It has no
lock.json, no folded-stack result, and no accepted profiled throughput. The command exited 1; it was not retried or relaxed. The first command exited 0. Successful captures passed the existing switch-window, symbol-resolution, and dropped-event validation; no invalid capture enters the tables.Repeat on nixos2 (2026-11-03)
Evidence: part 1 (reduced rows with per-capture host fields, and the full pre-epoch 16-worker fold) and part 2 (4-worker fold, void runs, run log).
Host.
desktopwas running many agents (load 25 to 31) andnixosran the land worker and CI, so the repeat usednixos2: i7-8700, 6 cores / 12 threads, no hybrid cores (so no pinning; affinity 0-11),perf_event_paranoid=-1, kernel 6.18. That replaces the quest's decideddesktophost for this repeat, per the maintainer's newest brief. Host configuration was not changed. Its~/.cargo/config.tomljobs = 2was left in force. The scratch checkout lived under~/moq-7742-profileat the claim commit b5a2e8fb77, the same harness and current source as thedesktopattempt. It was removed afterwards, and nothing of this run is left running.Builds. Direct Nix Cargo/rustc 1.98.1 and perf 7.3.0 from the dev shell, with no shim (
type -a cargoresolved to the Nix store) andBASH_ENV=/dev/null. Warm-up built current and both bases into the recipe's own target directories with the recipe's flags, before any capture. Each run still rebuilds its base's workspace crates (about 5 minutes), because the harness checks the base out at a fresh path. Within a run, all builds finish before the first capture.Quietness. Each run started only after no agent process had a running tool command for one minute (
uptimeload 0.10 to 0.18 at each start). A 2 s monitor recorded load, whole-host busy cores, CPU used outside the run's process tree, other agents' tool commands, and foreign build processes. Each capture's window runs from its control's first file to its profiled run's last file.d854bf98b1~1d854bf98b1~1codex,git,gh,quest,bash); 0.06 to 0.21 cores mean, 1.03 peak; no foreign buildsd854bf98b1relay-cpu-bisect), 5.1 to 5.8 cores mean throughoutOnly r1A has usable historical captures, and even it fails the "no other agents running throughout" bar; its foreign CPU was low. The run log, pre and post
uptime, and every row are in the evidence comments. The alternating sequence stopped after r2B: three other agent sessions and a detached CPU-pinned bisect benchmark were active onnixos2, so further runs would neither be quiet nor leave that benchmark undisturbed.Load includes the run's own relay and generator. At 1 worker this CPU delivers only 20 to 23k of the 31k offered frames per second, so 1-worker captures on
nixos2fail the zero-loss check (they have no mutex wait either way). The pre-epoch 16-worker wait is entirelyConsumer::poll_read_frame_once -> poll_read_frame -> poll<GroupState>. That is the stack thedesktopepoch capture showed at 599 us, where thedesktoppre-epoch capture had none on it. The pre-epoch 4-worker wait isConsumer::poll_expiry -> read<GroupState>. Total mutex wait is again mostlyTrackState(37906 of 38819 us at 16 workers). The current 4-worker capture had noGroupStatewait, unlike bothdesktopcurrent 4-worker captures (26 and 18 us underPrefetch::pop_cached).Conclusion: still inconclusive, but the contrary signal is weaker. The 47 -> 599 us
desktopcontrast is not attributable to #6782: the same stack waits 709 us on the pre-epoch revision here. Different hosts' totals are not comparable (0.401 ms before ondesktop, 38.8 ms here), so no same-host before/after pair exists yet. The share goal ("drops or holds") therefore has no valid measurement. Repeated 16-worker refusals (nonalternating switch records twice, a dropped event once) also make 16-worker pairs expensive to complete.Impact
quest/m1/unbounded-groups/log-lock-profile.mdand its epic link, and records the accepted conclusion inquest/m1/unbounded-groups/README.md. Main was merged into the branch first (no rebase).Alternatives and choices
nixos2repeat: ran the unchanged recipe withTMPDIRset to a scratch directory beside (not inside) the scratch checkout, so the harness's base worktree inherited no checkout Cargo configuration and needed nobwrap. Stopped the sequence rather than run unquiet captures, and did not touch the other sessions on the host..tmp/. The first attempt failed before building because a nested historical checkout inherited the root Cargo alias with an incompatible type. The successful commands usedbwrap --bind / / --dev-bind /dev /dev --bind "$PWD/.tmp/profile-tmp" /tmp, so the recipe's temporary checkout stays in owned scratch but Cargo sees no parent checkout configuration. No source or harness patch was applied.Build-tool incident (desktop attempt)
The initial Nix-shell
cargoentrypoint was an installedmbxshim. Its automatic capacity GC ran during builds before the first workload, contrary to the task's no-shared-GC constraint. The log reports 7 managed target directories removed (7.4 GiB logical; 6 abandoned, 1 live), 5.4 GiB of other cache objects evicted, and about 30.6 GiB of explicitly shared cache objects evicted. I did not invoke GC or delete a checkout, but failed to inspect that entrypoint before building. This cannot be undone; affected build artifacts can be regenerated. The second recipe and local CI bypassed it using the pinned Nix Cargo/rustc directory first in PATH andBASH_ENV=/dev/null. No cache-manager output appeared in the second recipe. Initial free disk was 12.2 GB; no actual storage failure occurred.Checks
quest check --base origin/mainwith the flake-pinned quest (nix shell .#quest): passed, 580 documents. The globally installedquest 0.1.0rejects theWaitingheading in 19 other quests already on main, which is pre-existing tool-version drift. That is also whyquest rmexited 1 after it had deleted the file and fixed the link.quest gate nix develop --command just check origin/main: passed (biome, markdown, gh check and test, quest, shell, large files, text, toml, nix, justfile).desktophistorical 1/4/16-worker captures passed validation; onnixos2, only r1A's pre-epoch 1/4/16 and current 1/4 captures did (1 worker fails zero-loss on that CPU).Review
Skipped for head
686ba58acb: the change since37d59f306fis a one-line doc fix to the epic README. It names the host beside each wait figure and says they are not comparable, per Pullfrog's thread.37d59f306fwas itself a docs-only head with the review skipped. The last review round was ond59c481488(Claude Code, Claude Opus 5.5, medium effort,quest review --base origin/main, round 3), and it reported no findings.quest/m1/unbounded-groups/README.md:120, cross-host 709 vs 599 µs read as a controlled comparison): fixed in686ba58acb, mirrored in the Problem section above.Decisions
nixos2, which had no quiet window (see above).kixtopas a quiet host, but it is a macOS laptop without Linuxperf, so it cannot run this profile; accepting was the stated alternative.Follow-ups
relay-cpu-bisectinstruction benchmark started onnixos2in the same minute (19:13Z). r2B's captures (19:19 to 19:23Z) overlapped that benchmark's build phase. That session should check whether its own captures overlapped r2B. A host lock thatquest gateor the bench recipes take would prevent this.nixos2too (r1A current, A1 base under load, r2B base). Cluster route memory already plans that harness fix, so no new quest is needed.GroupStatewait, so thedesktopcandidate did not persist.## Waiting, but the installedquest 0.1.0does not; a quest pin bump will need to migrate those 19 sections.(Written by Claude Opus 5.5)