Re: The correctness of floating-point square-root on the PPC603e.
- From: andrew@xxxxxxxxxxxxxxxxxxx (Andrew Klossner)
- Date: Thu, 13 Oct 2005 16:59:47 GMT
> You don't know how many correct bits it will give. The spec say you
> have at least 5 correct bits. From various tests, the result may be
> more precise, depending on the implementation.
If the chip vendor will stand up and promise more precision, great,
but otherwise depending on more than 5 bits from FRSQTE would be as
foolish as depending on overclocking the chip. I've been using
Motorola/Freescale chips long enough to know that undocumented
behavior can come and go without notice during a minor revision.
-=- Andrew Klossner
.
- References:
- The correctness of floating-point square-root on the PPC603e.
- From: joa
- Re: The correctness of floating-point square-root on the PPC603e.
- From: l'indien
- Re: The correctness of floating-point square-root on the PPC603e.
- From: Andrew Klossner
- Re: The correctness of floating-point square-root on the PPC603e.
- From: l'indien
- The correctness of floating-point square-root on the PPC603e.
- Prev by Date: Re: Gearing up the iSeries 520 express edition
- Next by Date: Re: Gearing up the iSeries 520 express edition
- Previous by thread: Re: The correctness of floating-point square-root on the PPC603e.
- Next by thread: Re: The correctness of floating-point square-root on the PPC603e.
- Index(es):
Relevant Pages
|
Loading