@Bink Great test Bink! Wow OpenFL has great performance!!!
I was also running more precise tests as well but I have included several other languages.
I have tested with:
1) Java
java 12.0.1 2019-04-16
Java(TM) SE Runtime Environment (build 12.0.1+12)
Java HotSpot(TM) 64-Bit Server VM (build 12.0.1+12, mixed mode, sharing)
2) AIR 32 Windows Captive 32bit
3) AIR 32 Windows Captive 64bit
4) NodeJS 64bit for Windows v12.3.1
Test source code:
1) Java -> https://pastebin.com/SJtPGPhh
2) AIR -> https://pastebin.com/i1Jje48i
3) NodeJS -> https://pastebin.com/YKrTmjAe
Here is the link to the AIR source code project and release builds: http://bit.do/eVhZw
Test data results:
http://bit.do/eVhZE
Every column has 3 numbers separated with comma because I have run every test 3 times.
There are interesting points in this test:
1) It was pretty amazing that division with Number/float was nearly the same for all 3 platforms which is really incredible!
2) Modulo operator is really costly than I originally thought and it has really slow performance on Windows 64bit captive.
3) Number division is super slow when compared to other operations like add, power, subtract.
4) All the "pure" tests were without modulo and Java was the fastest, than NodeJS and AIR was the slowest one 🙁
5) AIR Windows 64bit is still slower than 32bit. Not by much but it is still slower. I think this can add up when doing a lot of calculations with physics engine. Also I would like to understand how is it possible that there is any chance for 64bit version to be slower than 32bit version in number crunching. Shouldn't 64bit always be faster than 32bit?
Also it would be cool if people that have AMD CPU could run this new test to see if there are differences with pure Number/integer addition, subtraction, power and division between 32bit and 64bit captive.