Epoch Release 14 Finalization Plan
Last night at some unholy hour I finished a first pass at marshaling Epoch data back and forth across C-ABI boundaries. In less annoying terminology, this means that Epoch programs can do things like call Windows API functions, C-compatible APIs in other DLLs, and so on. More interestingly, those APIs can be given Epoch functions as callbacks, so there is seamless interoperability between C-compatible programs and Epoch programs.
This marks a great milestone in Epoch Release 14's development. There are now 58 compiler tests in the suite and all of them pass as fully native JIT compiled machine code.
As I mentioned previously, the last piece remaining to get Epoch fully-featured in native code is to implement the garbage collector. There are also some minor bugs related to Win32 interop that mean the raytracer demo is currently non-functional, so I probably have some work to do on the marshaling layer to get that all squared away.
My plan for Release 14 therefore looks something like this:
- Fix whatever bugs remain in marshaling
- Get the realtime raytracer benchmark running as 100% native code
- Map out garbage collection implementation (will post on this later)
- Implement garbage collection at least in basic form
- Final compiler test suite and probably a test harness for ensuring garbage collection works properly
- Package and ship R14
Of course, if you've followed the Epoch project at all in the past, you probably know as well as I do that I rarely stick to my release plans, so we'll see how it goes :-)
Once R14 ships, there's a few pieces I need to go back and hit. Here's the gist of what I'd like to do in the future with the language, in no particular order:
- Better reference semantics for controlling object lifetime and interlinked data structures
- Implementation of tasks (green threads)
- Implementation of message passing model
- Separate compilation and module support
- Enhanced template support for better generic programming features
- Namespace support
- Native support for array types
Naturally one of my big goals for the language moving forward is to build a self-hosting compiler and a proper IDE. Most of the above features are geared around making my life easier so I can actually accomplish those two projects. Of course, it may not be worth writing too many compiler features in the current system before moving to the self-hosting model...
So (with the above caveat of not following my plans very well) I just might wind up working on self-hosting instead. It all kind of depends on how entertaining this stuff is and where my whimsy directs me.
Another major thing I need to do that I have lamented about repeatedly is build a proper debugging facility for the native code model. Right now any bug in the code leads to lots of hunting through cryptic disassembly listings and requires a lot of painstaking effort. A proper debugger would probably fall under the Era IDE project heading, though, and thereby might wait a while. Even though I really need it.
Anyways... lots of speculation but as always we'll just have to see what the future brings. I'm super excited about all this and really looking forward to getting the language to a point where I can write hardcore software in it.
After all, that was kind of the point from day one :-)