This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/18595] [4.0 Regression] IV-OPTS is O(N^3)
- From: "rakdver at atrey dot karlin dot mff dot cuni dot cz" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 24 Jan 2005 23:09:35 -0000
- Subject: [Bug tree-optimization/18595] [4.0 Regression] IV-OPTS is O(N^3)
- References: <20041121145949.18595.pinskia@gcc.gnu.org>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Additional Comments From rakdver at atrey dot karlin dot mff dot cuni dot cz 2005-01-24 23:09 -------
Subject: Re: [4.0 Regression] IV-OPTS is O(N^3)
> > Other part is that scev tries to be too clever. Without need to
> > represent nonaffine induction variables (that we do not use anywhere)
> > we could use more memory efficient representation of ivs.
> > Without need to handle symbolic references (that we also do not use
> > anywhere, we could store evolutions in a fully instantiated form, and
> > we would not need instantiate_parameters/resolve_mixers functions atall.
>
> Uh, symbolic references are or will be used to do data dependence when
> MEM_REF and ARRAY_REF couldn't be generated from the pointers.
How? If the reference is left in symbolic form, it means that you know
nothing about its value, so the only thing you can do with it is to
check its equality/inequality, in code like
while (...)
{
i = something_weird ();
a[i] = ...; (a)
a[i+1] = ...; (b)
a[i] = ...; (c)
}
to find that (b) is independent on (a) and (c) and to find the
dependence between (a) and (c), but you do not need scev for this --
value numbering info is enough.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18595