[wasm] reduce eval in production code - #71932

Tagging subscribers to 'arch-wasm': @lewing
See info in area-owners.md if you want to be subscribed.

cc @kg

Sorry guys, I realized that there is still usage in the benchmark.
That also forced me to fix JSExport from nested class.

https://radekdoulik.github.io/WasmPerformanceMeasurements/?startDate=2022-07-11T19%3A31%3A11.000Z&endDate=2022-07-11T21%3A06%3A41.000Z&tasks=%2CAppStart&flavors=2

https://radekdoulik.github.io/WasmPerformanceMeasurements/?startDate=2022-07-11T19%3A31%3A11.000Z&endDate=2022-07-11T21%3A06%3A41.000Z&tasks=%2CAppStart&flavors=2

The time increase in this doesn't make any sense given the change in question, seems like variance?

cae954a...4a2ebf2 is the only commit in the diff and it changes the test mechanism. it was known and mostly fixed with the lazy init I think

Yes, I had to rewrite the benchmark to stop using eval code path and so I introduced dependency on JSExport into benchmark. The implementation at the time was triggering static constructors of the containing class during runtime init. That's now fixed in #77293

https://radekdoulik.github.io/WasmPerformanceMeasurements/?startDate=2022-07-11T19%3A31%3A11.000Z&endDate=2022-07-11T21%3A06%3A41.000Z&tasks=%2CAppStart&flavors=2

yup, it is #75598 and Pavel fixed it in #77293. the static constructors were doing a lot of work, which was not needed during startup