Label: ♦english

Happy new dragon year (which will start from this sunday)!

Actually, it was a busy month (I wish the christmas holiday was longer:) ), I was working on a new branch, so the changes didn't show up (also, googlecode does not display the change history very well, a better way is to run "hg serve" and then see it from a browser).

After the development in the past month, the next release has become a lot closer than what I previously thought:). The virtual machine kernel is more or less complete, with all the big features and design/structural improvements planned for the next release. Though there are still quite some issues need to be fixed or handled. The only time demanding parts left are the DaoJIT module and ClangDao tool.

The problem with the JIT module is that a simple and direct translation of Dao VM instructions to LLVM IR will not be very efficient for codes that are suitable for such translation, because Dao VM is already fast for such codes. The problem with direct translation is that it will generate a lot IR codes to synchronize data between the VM stack and JIT stack/registers, such synchronization is unnecessary for most intermediate data/variables. So to make a fast JIT, such redundant synchronization must be eliminated. But to do this safely and effectively, some static program analysis will be mandatory. I believe some basic analysis is sufficient for this, but I have no previous theoretical background and only a very little practical experience on program analysis, so I will need some time to catch up on this. Beside this, there isn't much left to do in the JIT module:)

There are more problems with ClangDao. The first one is that, the original implementation is based on the previous version of Clang, and needs some fixings to port it to the latest version. The second one is the support for C++ template, this is not yet complete and requires some serious work. The rest are, support for the Qt framework, and support for Objective-C (Clang supports Objective-C, there is no reason to not support it in ClangDao).

Maybe too many words for DaoJIT and ClangDao. Speaking of a roadmap, I think the development of the coming weeks will be on fixing the remaining issues of the VM kernel, adding some basic methods and algorithms to the graph module, finishing DaoJIT and fix a few things to port ClangDao to the latest version of Clang. Hopefully, the first release candidate will be out before the end of February. Depending on the development progress, the support for C++ template, Qt, Objective-C in ClangDao may or may not present in this release candidate.

I forgot to say something about the plan for the whole new year in my previous reply. Well, besides what I have mentioned in the previous reply, there aren't many things else that I really planned for this year. Actually there are a lot of things in my mind, but they are just not specifically for this year. I am sure it will still take a lot of time to add improvements to the VM kernel, existing modules, tools and documentations (I just remembered that I had planned to develop an online help system/module to make it easy to access documentations from interactive console:) ). I also planned to start to build a set of scientific modules/packages this year, but how much I will actually do for this will depend on many other things, we will see:)