> * IR work. Where do we stand? Is it possible we might turn it on for > 1.7, or is that too ambitious right now? >
We probably could get to green on the interpreter and compiler by 1.7 and we can definitely invite people to play with it and get us bug reports. 1. Getting to green: About 30 odd specs left. 2. Code clean up: I (and I think Tom as well) would like to clean up the code, instructions, and operands. 3. Adding 1.9 support? -- do we want to release the 1.8 version or attempt to get to 1.9. I haven't yet looked at what all the changes might be, but I suspect it may not be very significant given that most of the hard work was in the runtime libraries which you guys already have in place. 4. Bytecode generation: You probably have a better sense of this but this should not be very hard given that a lot more of the execution semantics is explicitly specified, when compared to the AST. But, I am sure we'll need to make changes to accommodate this code generation pass. 5. Performance tweaking: In order to get to green, there are places in the code that were done in a quick and naive way (from a performance standpoint) which we'll have to go in and start fixing up. The goal here is for the IR interpreter to be at least as fast as the AST interpreter. As it stands now, on some benches, we do better and on others, we do worse. Tom, you, and I have discussed this previously, but we have couple of low-hanging-fruit fixes that should be easy to do. After this is done, I will start reworking the optimization code that I had previously written (that code has fallen behind the various IR changes in the past few months). So, quite a bit of work left to do. My sense is that 1, and 5 should be reasonable. 4. might be possible, and 5. .. I am not sure yet. 2. will be an ongoing thing and Tom and I both want to get most of the hacks and ugliness out sooner than later. I am sure Tom has thoughts on the matter as well. But, I don't know what you meant by "turn on the IR code", since making it the default implementation is probably not what you have in mind. Subbu.
