[experiment] updated DOM sync approach

[lifeart]

Testing benchmark here, related to: #1540

This is not pure-js implementation, we using "same" reactive primitives as we have at the moment in glimmer-vm.

In scope of this POC we have:

• Autotracking
• Derived values
• Minimal VM
• Destructors

    Benchmark Results Summary    

duration phase estimated improvement -7297ms [-7425ms to -7173ms] OR -45.74% [-46.54% to -44.96%]
renderEnd phase estimated improvement -45ms [-46ms to -44ms] OR -64.53% [-66.13% to -62.92%]
render1000Items1End phase estimated improvement -776ms [-780ms to -765ms] OR -62.56% [-62.92% to -61.66%]
clearItems1End phase estimated improvement -14ms [-57ms to -13ms] OR -22.64% [-88.68% to -20.31%]
render1000Items2End phase estimated improvement -474ms [-482ms to -468ms] OR -47.44% [-48.2% to -46.84%]
clearItems2End phase estimated improvement -7ms [-10ms to -5ms] OR -11.49% [-15.75% to -7.05%]
render5000Items1End phase estimated improvement -2400ms [-2422ms to -2377ms] OR -52.54% [-53.03% to -52.03%]
clearManyItems1End phase estimated improvement -19ms [-21ms to -17ms] OR -8% [-8.97% to -7.12%]
render5000Items2End phase estimated improvement -1886ms [-1901ms to -1872ms] OR -46.2% [-46.55% to -45.85%]
clearManyItems2End phase estimated improvement -30ms [-33ms to -27ms] OR -12.35% [-13.46% to -11.13%]
render1000Items3End phase estimated improvement -26ms [-35ms to -18ms] OR -3.18% [-4.28% to -2.23%]
append1000Items1End phase estimated improvement -524ms [-530ms to -518ms] OR -57.12% [-57.78% to -56.51%]
append1000Items2End phase estimated improvement -470ms [-482ms to -467ms] OR -49.47% [-50.78% to -49.17%]
updateEvery10thItem1End phase estimated improvement -47ms [-49ms to -38ms] OR -28.27% [-29.69% to -23.03%]
updateEvery10thItem2End phase estimated improvement -34ms [-34ms to -33ms] OR -22.35% [-22.54% to -22.21%]
selectFirstRow1End phase estimated improvement -118ms [-118ms to -108ms] OR -78.51% [-78.97% to -72.49%]
selectSecondRow1End phase estimated improvement -81ms [-82ms to -81ms] OR -60.97% [-61.42% to -60.58%]
removeFirstRow1End phase estimated improvement -36ms [-44ms to -33ms] OR -15.45% [-18.91% to -14.42%]
removeSecondRow1End phase estimated improvement -37ms [-46ms to -35ms] OR -16.53% [-20.78% to -15.67%]
swapRows1End phase estimated improvement -32ms [-33ms to -26ms] OR -20.85% [-21.56% to -16.46%]
swapRows2End phase estimated improvement -26ms [-32ms to -17ms] OR -17.09% [-21.65% to -11.45%]
clearItems4End phase estimated improvement -19ms [-32ms to -13ms] OR -12.68% [-21.46% to -8.92%]
paint phase no difference [0ms to 0ms]

artifact-1.pdf

[lifeart]

JS Perf benchmark:

[lolmaus]

This is a holiday miracle. 😳

[NullVoxPopuli]

but how do we make glimmer faster?

[NullVoxPopuli]

in your screenshot, is gxt glimmer? or the experimental other thing? 😅

[lifeart]

@NullVoxPopuli it's glimmer-js replacement based on proposed runtime ideas krausest/js-framework-benchmark#1554 (https://github.com/lifeart/glimmer-next)