Let's land the ErrorRecovery work
Opened this issue · 0 comments
NullVoxPopuli commented
Steps:
- baseline bench in js-framework-benchmark
- needs to be updated to latest ember-source, as the VM has been upgraded
- link #1501 to ember-source (NOTE: additional followups here: #1528)
pnpm link:all
in glimmer-vm andpnpm link:glimmer-vm
in ember-source- do the upgrade and PR it, as draft
emberjs/ember.js#20602 - build ember-source and link to js-framework-benchmark
Diff of released, vs, w/glimmer-vm#main so far: https://github.com/krausest/js-framework-benchmark/compare/master...NullVoxPopuli:glimmer-vm-main?expand=1
Will add w/ the error-recovery PR - if the perf diff between current (not w/ glimmer-vm main) and post-error-recovery is "not significant" (in the wrong direction), we're good to move forward
- do the upgrade and PR it, as draft
Results:
Left to right:
- ember-source 5.7 (canary)
- ember-source 5.7 (canary + glimmer-vm's main branch)
- ember-source 5.7 (canary + glimmer-vm's error-recovery branch)
2323-12-12
ember-source is the same between both runs, (main)
only difference is glimmer-vm published vs main
still feels like margin of error though.
would need to compare with other folks' machines.
2023-12-11
NOTE: I need to use the same ember-source for both of these to reduce variables.
This bench had two variables:
- super old glimmer-vm vs glimmer-vm @ main
- ember-source 5.3 vs 5.6
- the current glimmer-vm on ember-source is much newer
ember-source 5.3 with glimmer-vm 0.84.2
vs
ember-source 5.6-alpha with glimmer-vm @ main