Re: [PATCH, middle-end]: Optimize a/sqrt(b/c) into a*sqrt(c/b).

["Richard Guenther" <richard dot guenther at gmail dot com>]

On 6/11/07, Kaveh R. GHAZI <ghazi@caip.rutgers.edu> wrote:
> On Mon, 11 Jun 2007, Uros Bizjak wrote:
>
> > Hello!
> >
> > This patch optimizes a/sqrt(b/c) into a*sqrt(c/b), converting one
> > division into the multiplication. This patch also converts cases like
> > sqrt(a)/sqrt(1.0/b) into sqrt (b*a).
> >
> > Patch was bootstrapped on x86_64-pc-linux-gnu, regression test for all
> > default languages is in progress. OK for mainline if it passes?
> >
> > 2007-06-11  Uros Bizjak  <ubizjak@gmail.com>
> >
> >         * fold-const (fold_binary) [RDIV_EXPR]: Optimize a/sqrt(b/c)
> >         into a*sqrt(c/b) if flag_unsafe_math_optimizations is set.
>
> Is it possible to do the same with cbrt, pow (possibly others?) along the
> lines of negate_exp/negate_expr_p/negate_mathfn_p?

We could certainly introduce a new predicate like reciprocal_mathfn_p
that tells we can compute the reciprocal of the function by computing
the function of the reciprocal argument.  And a corresponding
reciprocal_expr_p/reciprocal_expr set of
functions.

Richard.