да, флинк и спарк очень много кодогенерации в рантайме, поэтому не запустится
причем заметно больше чем в спринге с их DI

разобраться запрос - построить план - на основе статистики сделать его оптимизацию - сгенерировать код java - janino compiler - выполнить на воркерах

p.s. у спарка janino compiler сразу проверить на драйвере что код скомпилировался, отослать сорс код на воркеры, они повторно компилируют и выполняют его
у флинка байткод или сорскод на воркеры уходят я не помню

да, посмотрел, тикет висит на emil compiler events, сейчас только самому логи парсить

Да, джанино в этом плане гениален, практически пикл. Но для ni смерть, хотя казалось бы

если интересно кстати - вот ишшуе https://github.com/dotnet/runtime/issues/43805

what's the disassembler of both programs?

Didn't check. Didn't run the .net version at all. Do you think it's important?

absolutely, otherwise it's impossible to know what's happening behind scenes

it would be cool to have the intermediate language disassembly of both and the final resulting assembler for some target architecture (for example, the one that the benchmarks are being run)

also it would be interesting to have the telemetry of the GC (or running the same tests with the GC completely disabled in order to avoid noise)

there are many factors that can influence in the performance

Yes and it all depends on the level of details one needs. Sometimes the wallclock measurements are enough

I also think that it is a microbenchmark

It's an example I wouldn't even call it a benchmark

exactly

it is cool if you want to understand how each runtime JITs the code

but it's not something relevant for measurements

Sorry I don't understand your point here

I mean that code is a toy, it can be used for learning, but it isn't a valid benchmark to me

you can investigate this and learn a lot about how both compilers/runtimes work

Literally nobody said otherwise right? I don't think looking into this example will teach a lot about either runtime. But I guess it depends on the tenacity of the investigation