As New Year resolutions appear I want to state publicly my interests and my willing to work into my free/OSS projects.
I had working for two projects in the last year: CodeRefractor and at the end of the year to a more interesting (details will follow later) Java game.
This other project is JHeroes 2 a port of FHeroes 2 to Java. FHeroes 2 is a C++ project on top of SDL library and the main advantage of using Java for JHeroes 2 was to make it run in high resolutions laptops with weak CPUs (like mine :) ). For this reason I had as advantage Java's JavaFX technology, which is similar with WPF from .Net world.
I succeed to load a level, graphics from original Heroes 2 game (by porting FHeroes 2 core code to Java) but when it comes to displaying, the frames are visibily higher (FHeroes 2 in 1920x1080P has like 1-2 FPS on Windows, when JHeroes 2 has in my specific testing around 20 FPS) but the CPU is to high! I mean it uses more than 100% of one CPU even I don't do any separate background thread. Profiling I found that Java does create a separate thread in JavaFx and all commands go to this thread and I think this thread creates OpenGL commands which are executed back on the main thread. The bad part is that CPU usage is really high, and there is no way to put sleeps or anything to reduce this CPU usage. Or at least not obvious way to do it.
So I did what every other developer would do if it doesn't want to spend another 1 year of his free time (to get here I spent my 2 months of free time) to see that it doesn't work as intended at the end, I did test similar technologies using a VM and the single two technologies that do seem to work well enough (on my machine) are:
- OpenGL using OpenTK
- Mono + Cairo on Linux (on Windows it runs at around the 10-20 FPS, when on Linux I get between 30-60 FPS but also I have the possibility to put sleeps as all code runs on the main thread)
So, based on this results as I don't want to rewrite all my code to use OpenGL and from Java to rewrite it into C# (maybe in future I will have this will, but for now not so much), I also suspended my work on JHeroes 2.
Also I want to thank for all contributors to CR and for JetBrains offering a free license of Resharper for supporting OpenSource.
So as New Year resolution, I hope to do something else, maybe I will look into small items to improve on CR (very unlikely) or I will want to continue a Mirah like language implementation for .Net (named Calcium, but more on this I will write later).
But what can I say about 2014 is that I found some reassuring things which make work on CR to not be necessary:
- Microsoft did opensource a lot of its stack
- it has a "native compiler" which will be always better supported than CR will ever be
- Mono matured also a lot in the last year. As for me it still doesn't have full optimizations (CR still have in my understanding some optimizations that neither .Net or Mono has: especially Escape Analysis, "pure function evaluation", and this like that) but using LLVM backend, especially on Linux is really a great platform