This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: More on VIEW_CONVERT_EXPR
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: rth at redhat dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Wed, 21 Jul 04 17:03:52 EDT
- Subject: Re: More on VIEW_CONVERT_EXPR
Though I strongly suspect that whatever Ada's doing could be
instead represented as a nesting of structures/unions instead
of a VIEW_CONVERT_EXPR. Kenner says they tried that once, but
I really don't see how it differs from the type/type-as-base
thing that c++ has.
I don't know what it was that I said we tried, but I don't understand what
you mean here ...
Anyway, I suspect that incentive to move away from using VCE
will come from the fact that nested structures optimize well
and VCE doesn't.
As I said a few weeks ago, now that I'm looking at trees a lot more than
I used to, I see that Ada generates *a whole lot* of VCE's. I don't think
they'll affect optimization that much, but certainly there's no reason to
use it more than necessary.
I think that a lot of the history here is that this node was needed to
support an Ada language feature, so it was there. Then the front end
discovered that using that node, it could do lots of other things simpler.
That simplicity certainly had advantages, but I think it's time to revisit to
see if some of them are no longer needed and that is indeed my plan once I
get things working. (There are currenly 146 ACATS failures, with 46 ICE's.
The latter look like at least five different problems.)