On 11/14/2013 11:23 AM, Michael Matz wrote:
Hi,
On Thu, 14 Nov 2013, Andrew MacLeod wrote:
I think if following through with the whole plan there would (and
should) be nothing remaining that could be called a gimple expression.
very possibly, i just haven't gotten to those parts yet. I can change
the name back to gimple-decl.[ch] or some such thing if you like that
better.
-object? -operand? -stuff? ;-) Will all of these splits land at trunk,
i.e. 4.9? Why the hurry when not even such high-level things are clear?
I mean how can you think about rearchitecting the gimple data structures
without having looked at the current details. It's clear that not every
detail of the design can be fixated at this point, but basic questions
like "what's the operands?", "will there be expressions?", "how do we
iterate?", "recursive structures or not?" should at least get some answer
before really starting grind work, shouldn't they?
The splits are for header file cleanup and re-structuring into logical
components. As I mentioned in the original post, the file is needed to
break dependency cycles between gimple.h (the statements) , the
iterators, and gimplification. It is for the gimple stuff which doesn't
need any of those things but is consumed by them.
This really has nothing to do with my future plans, other than the fact
that I also said whatever is in this file is will eventually be split
into more things, but I'm not ready to do those splits yet, thus the
gimple-blah name doesn't matter to me. gimple-expr seemed convenient at
the time but clearly you don't like it, and I'll happily call it
whatever you want. It's a grab bag of all the gimple values which are
still trees...
maybe the suggested gimple-val.[ch] is ok?
Andrew
