This is the mail archive of the
mailing list for the GCC project.
Re: SSA usage question
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: "gcc at gcc dot gnu dot org" <gcc at gcc dot gnu dot org>, Richard Henderson <rth at redhat dot com>
- Date: Mon, 07 Jun 2004 11:24:19 -0400
- Subject: Re: SSA usage question
- Organization: Red Hat Canada
- References: <10406071515.AA10316@vlsi1.ultra.nyu.edu>
On Mon, 2004-06-07 at 11:15, Richard Kenner wrote:
> (1) I realized I made a mistake. "x" can't be TYPE_SIZE_UNIT, but has
> to be TYPE_SIZE_UNIT divided by the constant which is TYPE_ALIGN of the
> element type divided by BITS_PER_UNIT. The reason is that then the
> proper multiplication will be made by get_inner_reference for the use
> of highest_pow2_factor.
> (2) We don't need to do this (and probably shouldn't) if they are constant.
> (3) It isn't exactly TYPE_MIN_VALUE but the result of
> SUBSTITUTE_PLACEHOLDER_IN_EXPR with that as the first argument and
> TREE_OPERAND (array_ref, 0) as the second.
OK, but what's really important is that the computation needed to get
'v' and 'x' must be explicitly expanded in the GIMPLE code. Even if
it's simple enough to be trivially dead. Let the tree optimizers deal
with the additional fluff you may create in the FE.
> By the way, I'm going to check in my changes to the Ada front end early
> this afternoon. It's far from working, but does build, which is an
> advantage over what's there. It'll let people see for themselves what's
> going on.
Sounds good. Thanks.