Authored by: Anonymous on Thursday, March 28 2013 @ 02:33 PM EDT |
The surviving software implementations almost certainly follow
the IEEE guidelines now for floating point operations. The
standard has been in place for many years now and I doubt any
existing (generally available) libraries or hardware implement
anything else.
Besides, I can't imagine that rounding before (which would
mean rounding all terms) would be an improvement over rounder
after (rounding just the result.)[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Thursday, March 28 2013 @ 03:31 PM EDT |
The kernel emulates floating point
instructions for machines that provide a
FPU only optionally. That would include
80386 class CPUs (although 80386 support
has been yanked recently), for example. At
least I think that's still the case,
emulation in user space would be too
inefficient. [ Reply to This | Parent | # ]
|
|
Authored by: kenryan on Thursday, March 28 2013 @ 07:37 PM EDT |
Yes, at least the 2.6.32 source tree I have lying around
provides in-kernel FPU emulation for certain processors (e.g.
Motorola 68000). Not all of them are IEEE either, but I
haven't looked to see if any pre-round numbers.
---
ken
(speaking only for myself, IANAL)[ Reply to This | Parent | # ]
|
|
Authored by: Anonymous on Friday, March 29 2013 @ 01:43 AM EDT |
Regardless, how could anyone breach the patent even if it was valid? Hardware
uses IEEE. Emulation uses IEEE. If the patent deviates from IEEE as it states,
no one following IEEE could be impacted unless they deliberately implemented
this patent instead of following the IEEE standard.
Even if the patent were 100% valid, I still wouldn't pay a dollar for it if I
were a patent troll because there would be no one to sue.[ Reply to This | Parent | # ]
|
|