Manuel Schmidt said:
Hi,
does anybody know if there is an affordable computer with a processor whose
FPU can handle floating point units with higher precision than a normal AMD
or INTEL? Quad precision would be nice.
Thanx,
Manuel
There is this coprocessor approach, for use with multiple socket
Opteron boards. But I would think using an emulation library,
and enough computing horsepower, should do the job for an
overall lower price.
http://www.dailytech.com/article.aspx?newsid=1920
You would need to hire an FPGA designer, to code up
your quad precision FPU. Or, the FPGA designer could
purchase a pre-built IP (intellectual property) package
with the necessary FPU already constructed and tested.
The FPGA designer would then construct glue logic to
make the FPU usable (an API). (And no, devices like
this are not fast, so it is debatable what kind of
FLOP rate would result. The devices have thermal limits
and large numbers of node operating at GHz speeds
spells trouble.)
That is a lot of unnecessary expense, when a quad precision
floating point library turns the problem into a purely
software issue. Then you buy a fast enough processor
or multiple processors, to speed up the job.
"High-Precision Software Directory"
http://crd.lbl.gov/~dhbailey/mpdist/
In doing a quick search, IEEE 754R comes up a bit. A
revision to the 754 standard. But I don't see any references
to it being completed, or implemented in hardware.
http://home.earthlink.net/~mrob/pub/math/floatformats.html
http://www.research.ibm.com/journal/rd/435/schwarz.html
Paul
