[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: atlas on arm

In message <E1CIZeQ-0001p9-00@intech19.enhanced.com>
          Camm Maguire <camm@enhanced.com> wrote:

> Greetings!  Here is a piece of a post regarding difficulties compiling
> atlas on arm I recently submitted to the atlas mailing list.  I'd
> appreciate any thoughts.
> > 4) (longer term) I've been unable to compile atlas on arm for years.
> >    The rusage cpu timer is dysfunctional, and -DWALL gives
> >    unrepeatable timings which can't be boiled down within tolerance.
> >    The floating point performance by default is ridiculous when
> >    compared to the speed of the cpu (possibly the real source of these
> >    problems), and I can only surmise that in its minimal configuration
> >    floating point is emulated.

Floating point is emulated on most ARM CPUs.

> >    Later generations of the cpu with
> >    vector floating point instructions and respectable performance
> >    appear to exist as well.  ARM chips underly (most?)  Linux pda's.

Some of the XScale processors support vector floating point, yes - but I
don't think there's any support for this in ARM Linux, and in any case,
most of the XScale variants in use, including in PDAs, don't have this.

Peter Naulls - peter@chocky.org        | http://www.chocky.org/
Drobe - http://www.drobe.co.uk/        | The Premier RISC OS News Site

Reply to: