I’m trying to improve First Contentful Paint (FCP) and Largest Contentful Paint (LCP) with these two PRs: https://github.com/discourse/discourse/pull/15440 https://github.com/discourse/discourse/pull/15441 I’m really interested in the actual impact of these changes – so please have a try and give…

The first one looks like an easy optimisation that I think makes sense, @david and @eviltrout spent time optimising this area so I will be very curious to see what they think about it. The second feels a bit more fragile long term, totally get the desire to optimise but it worries me a bit cause i…

Hi @rrit - thanks for the PRs. The first one sounds like a good improvement. Were you able to measure the performance impact? How much time does it save? As @sam said, the maintainability of the second one is a little concerning. It looks like it’s a copy/paste from the Ember source code? Did you c…

lookupView-patch Implemented now via Map instead of Array. Time spent on application boot in lookupView - for development instance: without patch: 124 ms Firefox Profiler (callstack filtered to lookupView) with patch: 9 ms Firefox Profiler (callstack filtered to lookupView) Time saved: ~…

@david found an excellent way to fix this issue - see his comment on github . https://github.com/discourse/discourse/pull/15456 The time for calls to renderTopicListItem on webpage ‘latest’ drops from 348 ms to 201 ms in an Ember ‘production’-build. The earlier benchmarks were still running in dev…

[image] rrit: The time for calls to renderTopicListItem on webpage ‘latest’ drops from 348 ms to 201 ms in an Ember ‘production’-build. Unfortunately I haven’t been able to replicate this large speed increase. On Firefox and Chrome (macOS) I’m not seeing any measurable improvement. Chrome spe…

I can still reproduce the performance benefits using performance API in Firefox private-mode (Linux). Testing http://localhost:4200/latest Time spent in renderTopicListItem is down to ~190 ms from ~290 ms. My Discourse testing-instance has many topics with many answers and many different authors …

[image] rrit: Firefox private-mode (Linux). Would you mind sharing the Firefox and OS version? The 290ms number is almost 3x slower than a 2018 android device, which is a little surprising. [image] rrit: This results in lots of elements to be rendered. Maybe this is the difference in ou…

Proxy to meta.discourse.org Unfortunately running ember with a proxy fails for me: d/ember-cli --environment production --proxy https://meta.discourse.org http://localhost:4200/ Discourse Build Error Error [ERR_TLS_CERT_ALTNAME_INVALID]: Hostname/IP does not match certificate's altnames: Host: l…

The PR for the refactoring is now merged: https://github.com/discourse/discourse/commit/e6ab8f5b71efec9712b66806915a1f79d1e5998a Thanks for raising this one @rrit - it’s a nice improvement!

Optimizing FCP/LCP by caching raw-view module lookups

Development

eviltrout (Robin Ward) January 4, 2022, 6:02pm 6

Thanks for the patches. The second one does seem a little brittle.

Are your benchmarks running in development mode or production mode? Ember has quite a different profile in both.

Topic		Replies	Views
Defer javascript and show interim content on initial page load Development	19	3040	July 5, 2022
Upgrading Ember to 2.10 Development ember	71	11647	December 22, 2016
Google May 4th Core Update impact on Discourse forums Community Building	102	10534	July 5, 2022
Discourse not loading on legacy browsers Bug	50	6285	May 13, 2022
The State of JavaScript on Android in 2015 is... poor General	285	152545	December 18, 2022

Optimizing FCP/LCP by caching raw-view module lookups

Related topics