Floating point/math support implementation

[Harald Welte]

On Wed, Sep 17, 2008 at 09:50:49PM +1000, Carsten Haitzler wrote:

> > > > I have a few questions about the math support. I've noticed that
> > > > math functions such as sin() are very slow on the OpenMoko. How
> > > > is the math support implemented? Are all floating point operations
> > > > implemented in software, or is there hardware support for float
> > > > multiplication, and trigonometric functions?
> > >
> > > no fpu. so no hardware. software only. avoid floating point math where you
> > > can.

Which is true for the majority of all ARM based systems out there... so nothing
particular about the Openmoko devices.

> > Would it be thinkable to use the OpenRISC processors embedded into Glamo chip?
> 
> simple version: no. no specs on how to use it or even if it can do any fp. just
> avoid fp math. do it some other way. (MUCH easier than trying to use the
> glamo).

in fact, according to what I have read and been told by Smedia before, the 3372
doesn't have the OpenRISC.  Only the 3362 (for maximum QVGA resolution) and
older chips have it.

Also, given the slow interconnect, I think it is not a good solution
performance-wise

-- 
- Harald Welte <laforge at openmoko.org>          	        http://openmoko.org/
============================================================================
Software for the world's first truly open Free Software mobile phone