PASS

Apple M5 Pro · macOS 26.6 · 48d5eb6 · A/B vs reference · 10s

PASS env_ac_power PASS env_thermal PASS env_cpu_idle

tl;dr

  1. 36.2% lower finalize latency efficient-720p30-hevc · gate passed
  2. 21.3% lower p95 CPU efficient-720p30-hevc · gate passed
  3. 11.8% lower average CPU balanced-1080p30-h264 · gate passed

efficient-720p30-hevc

PASS
observed fps
29.37
Effective FPS measured from the decoded recording. Stimulus: 119.51 fps.
completeness
24.6%
Decoded frames captured divided by stimulus frames.
start
856 ms
Time from job submission until the first captured frame.
finalize
112 ms
Time from the stop request until the file is finalized.
cpu avg/p95
3.3/14.5
Average and p95 CPU percentage across the recorder process tree.
peak rss
45.9 MB
Highest resident memory used by the recorder process tree.
gate status measured threshold Δ vs ref notes
observed_fps PASS 29.37 >= 28.5 fps (95% of min(target 30, stimulus 119.506)) –
capture_completeness SKIPPED 0.25 capture rate >= display rate – profile captures slower than the stimulus displays; skipping every other frame is by design
self_report_drop_ratio PASS 0 <= 0.50% –
max_pts_gap_ms PASS 65 <= 600 ms –
pts_monotonic PASS true true –
codec_match PASS hevc hevc –
dimensions_match_request PASS 1280x720 1280x720 –
self_report_disagreement PASS 0 <= 5% –
start_latency_ms PASS 856 <= 4000 ms –
finalize_latency_ms PASS 112 <= 3000 ms –
regression_observed_fps PASS 29.37 candidate >= reference - 15% and >= reference - 1 fps -0.15 (-0.5%)
regression_cpu_avg PASS 3.32 candidate <= reference + 15% and <= reference + 3 % -0.06 (-1.6%)
regression_cpu_p95 PASS 14.50 candidate <= reference + 15% and <= reference + 3 % -3.30 (-21.3%)
regression_max_rss PASS 45.9 MB candidate <= reference + 15% and <= reference + 20 MB 4.0 MB (+9.6%)
regression_start_latency PASS 856 candidate <= reference + 30% and <= reference + 500 ms -53 (-5.9%)
regression_finalize_latency PASS 112 candidate <= reference + 15% and <= reference + 150 ms -38 (-36.2%)

balanced-1080p30-hevc

PASS
observed fps
29.48
Effective FPS measured from the decoded recording. Stimulus: 119.65 fps.
completeness
24.6%
Decoded frames captured divided by stimulus frames.
start
965 ms
Time from job submission until the first captured frame.
finalize
119 ms
Time from the stop request until the file is finalized.
cpu avg/p95
3.4/10.7
Average and p95 CPU percentage across the recorder process tree.
peak rss
46.0 MB
Highest resident memory used by the recorder process tree.
gate status measured threshold Δ vs ref notes
observed_fps PASS 29.48 >= 28.5 fps (95% of min(target 30, stimulus 119.652)) –
capture_completeness SKIPPED 0.25 capture rate >= display rate – profile captures slower than the stimulus displays; skipping every other frame is by design
self_report_drop_ratio PASS 0 <= 0.50% –
max_pts_gap_ms PASS 76.67 <= 600 ms –
pts_monotonic PASS true true –
codec_match PASS hevc hevc –
dimensions_match_request PASS 1920x1080 1920x1080 –
self_report_disagreement PASS 0 <= 5% –
start_latency_ms PASS 965 <= 4000 ms –
finalize_latency_ms PASS 119 <= 3000 ms –
regression_observed_fps PASS 29.48 candidate >= reference - 15% and >= reference - 1 fps 0.06 (+0.2%)
regression_cpu_avg PASS 3.43 candidate <= reference + 15% and <= reference + 3 % -0.07 (-2.1%)
regression_cpu_p95 PASS 10.70 candidate <= reference + 15% and <= reference + 3 % -1.50 (-10.3%)
regression_max_rss PASS 46.0 MB candidate <= reference + 15% and <= reference + 20 MB 4.1 MB (+9.7%)
regression_start_latency PASS 965 candidate <= reference + 30% and <= reference + 500 ms 135 (+16.7%)
regression_finalize_latency PASS 119 candidate <= reference + 15% and <= reference + 150 ms 7 (+6.3%)

high-native60-hevc

PASS
observed fps
57.76
Effective FPS measured from the decoded recording. Stimulus: 119.95 fps.
completeness
48.2%
Decoded frames captured divided by stimulus frames.
start
775 ms
Time from job submission until the first captured frame.
finalize
139 ms
Time from the stop request until the file is finalized.
cpu avg/p95
5.6/17.7
Average and p95 CPU percentage across the recorder process tree.
peak rss
47.3 MB
Highest resident memory used by the recorder process tree.
gate status measured threshold Δ vs ref notes
observed_fps PASS 57.76 >= 54 fps (90% of min(target 60, stimulus 119.95)) –
capture_completeness SKIPPED 0.48 capture rate >= display rate – profile captures slower than the stimulus displays; skipping every other frame is by design
self_report_drop_ratio PASS 0 <= 1.00% –
max_pts_gap_ms PASS 55.00 <= 600 ms –
pts_monotonic PASS true true –
codec_match PASS hevc hevc –
dimensions_match_request PASS 2560x1440 2560x1440 –
self_report_disagreement PASS 0 <= 5% –
start_latency_ms PASS 775 <= 4000 ms –
finalize_latency_ms PASS 139 <= 5000 ms –
regression_observed_fps PASS 57.76 candidate >= reference - 15% and >= reference - 1 fps 0.12 (+0.2%)
regression_cpu_avg PASS 5.65 candidate <= reference + 15% and <= reference + 3 % -0.26 (-4.4%)
regression_cpu_p95 PASS 17.70 candidate <= reference + 15% and <= reference + 3 % -2.10 (-10.3%)
regression_max_rss PASS 47.3 MB candidate <= reference + 15% and <= reference + 20 MB 4.3 MB (+9.9%)
regression_start_latency PASS 775 candidate <= reference + 30% and <= reference + 500 ms -37 (-4.6%)
regression_finalize_latency PASS 139 candidate <= reference + 15% and <= reference + 150 ms 40 (+36.7%)

balanced-1080p30-h264

PASS
observed fps
29.53
Effective FPS measured from the decoded recording. Stimulus: 119.61 fps.
completeness
24.7%
Decoded frames captured divided by stimulus frames.
start
775 ms
Time from job submission until the first captured frame.
finalize
84 ms
Time from the stop request until the file is finalized.
cpu avg/p95
3.3/13.3
Average and p95 CPU percentage across the recorder process tree.
peak rss
46.2 MB
Highest resident memory used by the recorder process tree.
gate status measured threshold Δ vs ref notes
observed_fps PASS 29.53 >= 28.5 fps (95% of min(target 30, stimulus 119.606)) –
capture_completeness SKIPPED 0.25 capture rate >= display rate – profile captures slower than the stimulus displays; skipping every other frame is by design
self_report_drop_ratio PASS 0 <= 0.50% –
max_pts_gap_ms PASS 93.33 <= 600 ms –
pts_monotonic PASS true true –
codec_match PASS h264 h264 –
dimensions_match_request PASS 1920x1080 1920x1080 –
self_report_disagreement PASS 0 <= 5% –
start_latency_ms PASS 775 <= 4000 ms –
finalize_latency_ms PASS 84 <= 3000 ms –
regression_observed_fps PASS 29.53 candidate >= reference - 15% and >= reference - 1 fps 0.05 (+0.2%)
regression_cpu_avg PASS 3.29 candidate <= reference + 15% and <= reference + 3 % -0.42 (-11.8%)
regression_cpu_p95 PASS 13.30 candidate <= reference + 15% and <= reference + 3 % -1.80 (-9.4%)
regression_max_rss PASS 46.2 MB candidate <= reference + 15% and <= reference + 20 MB 4.1 MB (+9.8%)
regression_start_latency PASS 775 candidate <= reference + 30% and <= reference + 500 ms -44 (-5.7%)
regression_finalize_latency PASS 84 candidate <= reference + 15% and <= reference + 150 ms -3 (-2.0%)

history

run status suite commit mode
PASS release 824c8fe A/B
PASS release 8a62b7b A/B
PASS release 8c49193 A/B