On Mon, Sep 10, 2001 at 01:03:23PM +0200, Gacquer William wrote:
> Does FASTFPE emulate both single and double precision? ( if yes, was it true
> also for the 2.4.2 kernel ?)
> Other than precision uncertainties, what are FASTFPE limitations?
> Does FASTFPE compete against a fixed point math implementation from the
> performance point of view? I mean : is it as fast as a fixed point
> implementation?
You could try reading some of the comments in the fastfpe files:
| Inside the emulator the FP numbers are kept with 32 bit accuracy for both
| mantissa and exponent. The FP structure has 4 words reserved for each
| register, the first is used just for the sign in bit 31, the second is the
| mantissa (unsigned integer) and the third is the exponent (signed integer).
_______________________________________________
http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
Please visit the above address for information on this list.
