On Oct 9, 7:42 pm, jmakov <[EMAIL PROTECTED]> wrote:
> Also I have recently found this
> paperhttp://www.mathematik.uni-dortmund.de/lsiii/static/showpdffile_Goedde....
Hello Jernej,
>
> Some citations:
> "The double-single emulation on the GPU is not always sufficient, as
> the double format has a clearly larger mantissa and exponent range
> than the emulated
> format: 52/46 = 1.13, 211/28 = 8, and three term emulations become
> truly expensive. Therefore, we perform the high precision corrections
> on the CPU rather than emulating them on the GPU ... Instead, the GPU
> accelerates the more regular, most demanding parts of the algorithms,
> and leaves the subsequent processing of the results to the well
> tested, higher abstraction code on the CPU"
>
I am currently involved in a project at the University of Dortmund
that is porting Dominik Göddeke's multigrid solvers to the Cell CPU
(well we are writing the underlying linear algebra to use a number of
SPUs in the Cell in parallel). The end result ought to be a GPLed
library to solved systems with a variety of algorithms that can either
use CPU, GPU or Cell to solve the system. It ought to be released in
march. Any guess in which computer algebra system it ought to end up
in as an optional component?
I look for arbitrary precision arithmetic on GPUs and except for the
mentioning on the CIS 665 website I found no actual code. Got any more
pointers there?
Cheers,
Michael
> Jernej
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-forum
URLs: http://sage.math.washington.edu/sage/ and http://sage.scipy.org/sage/
-~----------~----~----~----~------~----~------~--~---
