search

GoogleChrome / lighthouse

Automated auditing, performance metrics, and best practices for the web.
https://developer.chrome.com/docs/lighthouse/overview/
Apache License 2.0
28.4k stars 9.38k forks source link

Lantern EIL+TBT should use internally consistent FCP/TTI estimates #11677

Open patrickhulce opened 3 years ago

patrickhulce commented 3 years ago

Summary Currently in the optimistic/pessimistic lantern estimates for EIL and TBT we use the timing data computed by the FCP/FMP/TTI metric simulations as cutoffs but this leads to internally inconsistent data when the graphs are slightly different in EIL/TBT (e.g. TBT's graph has a TTI of 3.2s but the timing from the pessimistic simulation was 3.1s so we ignore the last 100ms of blocking time)

We should explore using the nodes of FCP that exist in EIL/TBT graphs and using those times within our EIL/TBT simulation results computation.

Discovered as part of #11666

paulirish commented 1 year ago

EIL part is probably obsolete now.

Related to the FCP-can-be-faster-than-TTFB issue.

Also related: should TBT not be dependent on TTI.

More useful to extract metrics from the graph, rather than give-me-the-endTime-of-the-graph.