Hi, Sure
- We need a way of testing the New compiler, It is very hard to now if the changes I am introducing work. I have a set of test but they are mainly oriented towards the closure implementation. - Not everything is done with visitors. - The ASTTranslator and subclasses have methods of considerable size which are hard to understand, mainly the emits and accepts - The Decompiler does not work - IRTranslator: needs some work, there are some abstractions that are not clear like sequences and pending instructions which make the code in every other method more complicated. - It is not completely modular, we do not have a way of configuring it from scratch - The phases of the compiler are not clearly defined If you ask me, I would say that the key point is to have a way of validating our compiler. This is what is holding me back from introducing more aggressive refactorings. This does not mean that we need all those point in order to have a running compiler, on the contrary, the compiler is working quite well, I am fixing special cases right now. But, if we what something we could improve and experiment with, we will need these points. Cheers, Jorge On Thu, Mar 4, 2010 at 1:16 PM, Stéphane Ducasse <[email protected]> wrote: > jorge > > could you make a list of things to do in 10 min? > So that we get an idea. I really believe that we should invest for the future. > > Stef > > On Mar 4, 2010, at 10:07 AM, Jorge Ressia wrote: > >> Yes, the New compiler is not better. However, I think the New Compiler >> is the right direction to pursue. >> >> After fighting with it for some months to make the Closures work in it >> I can tell you that I learn a great deal from it, painfully :) . >> But I now have a pretty good idea of what it is required to have a >> modularized compiler with more suitable abstraction for this >> particular domain. >> I implemented many tests and I have to implement a lot more. Then, I >> believe that by expressing ourselves through these tests we will be >> able to modify the compiler to make it look the way we want. >> >> At least that is my plan with the New compiler. >> >> Cheers, >> >> Jorge >> >> On Thu, Mar 4, 2010 at 9:53 AM, Lukas Renggli <[email protected]> wrote: >>>>> I'll try to build the tests and come up with a better solution. >>>> >>>> You what I would love to spend 2/3 weeks just coding in the new compiler >>>> and making sure >>>> that it has the right abstractions/messages to deal with issue like that. >>> >>> The problem is that the New Compiler is not really in a better state >>> than the old compiler. Of course it has more objects and is much >>> easier to extend, but all in all it is a mess too. This is mostly due >>> to the fact that the system below it changed quite a bit in the past >>> years (pragmas, properties, closures, primitives, new bytecodes, ...) >>> and that the New Compiler had to be patched and changed quite heavily >>> to accommodate these new requirements that it was not designed for. >>> Maintaining and fixing the New Compiler got so extremely expensive >>> that it is questionable if this is still a viable platform? Ask Jorge >>> and Marcus. >>> >>> Lukas >>> >>> -- >>> Lukas Renggli >>> http://www.lukas-renggli.ch >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> [email protected] >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>> >> >> _______________________________________________ >> Pharo-project mailing list >> [email protected] >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > _______________________________________________ > Pharo-project mailing list > [email protected] > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > _______________________________________________ Pharo-project mailing list [email protected] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
