This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [4.0 PATCH] Disable Tree-SRA for the Ada compiler
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: ebotcazou at adacore dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sat, 26 Feb 05 10:20:49 EST
- Subject: Re: [4.0 PATCH] Disable Tree-SRA for the Ada compiler
IIRC it's related to discriminated types and subtypes. Components may
be lost during the transition between them and SRA doesn't like that
at all.
Since you've investigated this, can you give me the results of your
investigation so I don't have to start from scratch? Specifically, show
me what was miscompiled.
My impression was that making SRA fully works with Ada would require
an overhaul of the approach used by gigi to build subtypes of
discrimated types, which is currently direct translation from the
front-end's types.
I don't see that. At worst, we'd have to add a flag to a type that says
"don't SRA this even though you think we can", but I'd still like to know
what goes wrong.