Re: [PATCH] PR optimization/11741: PRE multiple sets in a PARALLEL

[Richard Henderson <rth at redhat dot com>]

On Wed, Sep 03, 2003 at 09:57:15PM -0600, Roger Sayle wrote:
> Before dropping this patch [which solves a PR, because there might
> be a bug elsewhere in the compiler, which hasn't yet caused a PR]

This is not the right way to think.  I know for a fact the patch
will cause other problems; I havn't bothered to find a test case,
but it would only be a matter of time.

> Unfortunately, many of the comments in gcse.c appear to be out
> of date, so following the logic isn't easy.  Could you or Jeff
> explain which part of the LCM code in gcse.c would need fixing?

You'd have to treat parallel patterns as a unit.  You could not
allow the parts of the parallel to be broken up into separate pieces,
not allow to be unified with insns that only do one one of the
sets, compute properties wrt the entire parallel, etc.

My guess is that this is a mostly pointless thing to try.  Such
insns will be movable basically never.

Indeed, for the test case in question, the insn can't be moved,
since the memory move part of the parallel aliases other memory
uses in the function.


r~