This is the mail archive of the crossgcc@sources.redhat.com mailing list for the crossgcc project.

See the CrossGCC FAQ for lots more information.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Floating point emulation


On Mon, Dec 02, 2002 at 10:49:57AM -0500, Peter Barada wrote:
> >(Incidently, if you are compiling with -msoft-float, I don't
> >expect that using doubles will "increase performance", as
> >you'll double the amount of work the processor has to do...)
> 
> It all depends on the hardware implementation of the FPU.  If the FPU
> can only operate in double precision, then to properly model single
> precision operations requires rounding instructions to be added after
> every operation to force the final result to appear as if it was
> calculated in single precision.  This can indeed slow it down.

Interesting.. But that has to be expected every time you use
hardware in a manner that it wasn't indended for (same
story, it's not easier to play at 8KHz on a DSP that only
supports 24KHz...). 

However, I was talking about code compiled with
-msoft-float... where arguably all operations are performed
with many cycles by the non-floating-point-enable CPU...
Then it should always be slower to use double than single?
(if double is indeed bigger than single, as pointed out in
Kai's other posting on the subject).


/Y


------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]