Ok, so I finished the python translation of slicc. With the first patch applied, you can go into src/mem/pyslicc and run ./pyslicc. The output files can be diffed against the C++ slicc with -wBur. There's basically only minimal differences. The second diff changes the generated code output to be more inline with style and generally a bit easier to read. It also brings the comments back.
Neither diff actually replaces the current slicc with python slicc in the build. Basically, I'd like people to more or less say that they're cool with this and then I'll generate a new diff that replaces slicc with this python version of slicc. A few things to determine. Where should the python slicc package live? It can stay in src/mem/slicc, or we can put it in src/python/slicc, or src/python/m5/slicc, though I'm not sure there's any utility to including it in the m5 binary (we could do that if we wanted though.) I also created a pyslicc executable that imports and runs slicc. Where should that go? I'm thinking util/slicc since it's only for convenience. The eventual SCons code will just import slicc and call it directly. Let me know what you guys think. Hopefully you agree that it's a whole lot easier to read the slicc code. I generate both html and c++. I do not do the mif code. I can pretty if it is worth it. I thought mif was for Framemaker, but someone said that it was for a model checker. Which is correct? I think we should keep something that is for a model checker, but I'm not convinced after looking at it that this is it. Oh yeah, I more or less kept the AST and grammar the way it was, though I did fix some shift/reduce conflicts that the existing grammars have. I would like to talk about removing some of the unused stuff from slicc, like check_stop_slots. _______________________________________________ m5-dev mailing list m5-dev@m5sim.org http://m5sim.org/mailman/listinfo/m5-dev