This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: GIMPLE issues and status of gimple-tuples
- From: nkavv at physics dot auth dot gr
- To: gcc at gcc dot gnu dot org
- Date: Tue, 21 Nov 2006 07:55:14 +0200
- Subject: Re: GIMPLE issues and status of gimple-tuples
- References: <1164081013.24283.ezmlm@gcc.gnu.org>
Hi there
i have setup gcc-4.1.1 and friends :)
It seems that the dumps (with -fdump-tree-all) are really low-GIMPLE (good!),
e.g. t03 and t18. And it is unstructured three-address-code.
I have only one question for now: it looks like that in if-then-else statements,
one of the two branch targets is ALWAYS a fall-through. Is this guaranteed to
hold for all cases? If yes, that's fine!
A minor objection (and only on the dump format and not the intrinsics) is that
global symbols (i.e. global variables that would map to a global symbol table)
are not noted explicitly. I'll modify my version for listing the global symbols
before the function list.
BTW Sebastian, it seems that the tree-data-ref.{c,h} infrastructure has been
developed for array dependencies in loops. I'll go deeper if there is something
that is actually bookkeeping scalar variables dependencies (e.g. SSA variables
in basic blocks) as well. The tree-dh.{c,h} files are what is closer to that
than anything else.
Kind regards
Nikolaos Kavvadias