Thanks to type inference the jroc Javascript dotnet compiler out performs other native dotnet solutions by a magnitude of 10 to 1 in many cases. The prime benchmark is a good example of this.
CLI
https://t.co/5p2FVFuYie
source code
https://t.co/TddbldjiQ5
ECMA 262 compliance status
https://tomacox74/js2il/blob/master/docs/ECMA262/Index.md
#javascript #dotnet #csharp
@romxdev COGS do matter (in regard to comparing Bun and Nodejs) but agree that if you have something critical that is compute intensive maybe Javascript should not have been your first choice.
@davepl1968@MustangMan_TX 1997 Mitsubishi Eclipse GST.. Also shared the same platform with the Eagle Talon. What did happen to diamond-star motors? Just sort of disappeared. The green car below is the Eclipse that was in the movie Fast and Furious.
I played with hydrofusion. I wanted to see live output while the tasks are being run, not just a overly generic 5 point checklist. It also felt slower for smaller size tasks though i haven't actually done any controlled comparisons to what the passed clock time is compared to me manually choosing the LLM.
JROC compiled javascript is currently executed in a seperate thread, not the host thread that invoked the script. This favors stability over perf. The downside is that it adds the overhead of message passing and whatever additional memory is needed for the script execution thread.
I have a tracking item to be more smart about it and not spawn the extra thread when it is not needed. (i.e. if the script is %100 synchronous the extra thread is not needed). Plus I think the host should be able choose as well by providing a additional, optional setting.
@jrysana As I remember there are circular relationships in the data structures. I doubt that would specifically account for such a size difference though. Like everything else now you could ask AI to do a analysis to answer that question.