On Wed, Feb 18, 2009 at 7:59 PM, nathan binkert <[email protected]> wrote:

> Hi Steve/everyone,
>
> right now, we have code in src/python/swig.  I'm thinking that for the
> swig stuff, we should distribute the .i files along with the code they
> wrap.  Mostly this means that swig code goes in sim.  I'm thinking
> about this because I was going to add a .i file or two and wanted to
> do the right thing.


What a coincidence... I've never really looked at this code before about 4
hours ago.  Brad and I were wondering where Stats::dump() gets called, and I
poked around and found src/python/swig/stats.i, and the first thing I
thought was "shouldn't this really go in src/base/stats?"  I was going to
send an email to you about it later this evening.


> Along these lines, should I rename .i files to
> .ii to indicate C++?


Does it really matter?  Is that a swig convention?  If not, I'd say don't
bother.  It's not like there's any plain C code that we wrap.


> Additionally, there's a question about src/python/m5. I'm inclined to
> leave it alone because it is actually a python package that we build.
> Arguably, we might want to rename python to lib, but I could be
> convinced that this is a bad idea.  I guess it just feels odd to have
> a directory called python.


I think it's fine the way it is... it is a little asymmetric since we don't
have a "c++" directory, but I can deal with that.  "lib" is a little too
generic I think.


> Finally, right now, SimObjects are compiled into a flat namespace
> called m5.objects.  It seems that we should have some package
> hierarchy here to potentially mirror the C++ file hierarchy (and to
> get rid of the practice of prepending the ISA to the name of many
> objects).  It would also help prevent name clashes with any EXTRAS
> people might create.


That makes sense.


> These things come up because I'm actively working on the build system
> which would touch many of these things.
>

Cool.

Steve
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev

Reply via email to