Re: Performance of Integer Multiplication on PIII

[Kevin Atkinson <kevin at atkinson dot dhs dot org>]

On Sat, 3 Nov 2001, Jan Hubicka wrote:

> > I am confused myself as to which architecture is meant by -march=i686,
> > but I didn't think it was P-III.  I thought -march=pentiumpro was the one
> > for P-II/P-III.  However, wouldn't many people wonder whether you mean P4
> > when you talk about "faster Pentiums?" Some of us would not go out of our
> > way to slow down the P4 by asking for imul on code meant to run on a
> > variety of CPU's, although some would like to have -Os favor the use of
> > imul.  It might be interesting to have a separate table of the penalties
> > used in each version of gcc for each architecture selection, rather than
> > having to peruse source.
> Gcc do have set of costs for each supported x86 variant and for size
> optimization separately.  The model is an estimate (it does not handle leas and
> 2 address machines perfectly), as the algorithm to construct sequences is not
> trying all possibilities, so it is possible that the produced sequences are
> suboptimal, but in case it makes truly big differences, we probably should
> investigate it.

Yes I agree and this is one of the cases.