Re: Fixed point Vs Floating point

On Wed, 18 Jun 2008 09:32:01 -0400, Phil Hobbs
<pcdhSpamMeSenseless@xxxxxxxxxxxx> wrote:

JosephKK wrote:
On Tue, 17 Jun 2008 07:28:58 -0400, Phil Hobbs
<pcdhSpamMeSenseless@xxxxxxxxxxxxxxxxxx> wrote:

IEEE 754 floating point is great in theory, but there's no hardware
support on any CPU I know of for some of its features--especially
denormalized numbers.(*)

Actually both the 8087 and its successors, and the 68881 and its
successors, as well as later MIPS, SPARCs, and HP-PARISC, and IBM
POWER processors all do correct and complete IEEE 754 floating point.
Most of cause interrupts on each denormal result.

The unimplemented features have to be emulated
by the runtime handling the EM_DENORMAL exception in software on
_every_ _single_ _operation_, which takes *forever*. On my big EM
simulator, I have to fill the entire volume with noise of a few times
FLT_MIN before starting to avoid it running like molasses for several
hundred iterations until all the denormals get flushed out.

Now that is an inefficient implementation issue.

Do you know of any processor that handles denormals in hardware, other
than by flushing to zero? I haven't found one, and I don't think any
exists. That's what I mean by "no hardware support".

Cheers,

Phil Hobbs

Gosh, just after i named multiple ones. Go study the databooks before
you respond.

.

Relevant Pages

  • Re: Fixed point Vs Floating point
    ... POWER processors all do correct and complete IEEE 754 floating point. ... I have to fill the entire volume with noise of a few times FLT_MIN before starting to avoid it running like molasses for several hundred iterations until all the denormals get flushed out. ... That's what I mean by "no hardware support". ...
    (sci.electronics.design)
  • Re: Future Risc
    ... Add one to the exponent field before ... Hardware just does the operation assuming ... Denormals may result in traps, replays or are just flushed to zero on ... hardware on the special path including the extra shifts. ...
    (comp.arch)
Quantcast