(j3.2006) IEEE and the GPU world

[Bill Long]

Keith Bierman wrote:
> The issues raised by GPUs are not new; we had them with attached array 
> processors (FPS, et al), and on non-homogeneous processors (e.g. the 
> Cydra5 which had a VLIW numeric engine, and multiple 68K processors, and 
> program execution was transparently split (well, transparent to the 
> user, the folks making the system work certainly had no automatic tools!).
> 
> So I suggest the language be made slightly more generic to handle any 
> combination of IEEE and nonIEEE fp hardware, irrespective of what the 
> device is otherwise intended to do.
> 

Since the question originally posed concerned GPU's, I think it is OK 
directly answer that question.  What ultimately matters are the Edits at 
the end, and those are written for "a processor that does not 
support..." rather than mentioning "GPU".   I assume John did this 
intentionally, and for the reasons you noted.

Cheers,
Bill




> If a /note/ has to be included to spell out what a GPU is, and that 
> contemporary GPUs can fit into this role so be it. But I would have 
> argued that it's such a common term, it would like spelling out CPU ... 
> which is sometimes the correct pedantic thing to do, but is hardly 
> necessary in the context of an interpretation.
> 
> 
> Keith Bierman
> khbkhb at gmail.com <mailto:khbkhb at gmail.com>
> kbiermank AIM
> 303 997 2749
> 
> 
> On Fri, Aug 13, 2010 at 8:19 AM, Whitlock, Stan <stan.whitlock at intel.com 
> <mailto:stan.whitlock at intel.com>> wrote:
> 
>     John,
> 
>     Please add the word "that" in the following paragraphs under
>     "Answer" and "Edits":
> 
>      A scoping unit that has no explicit USE statement for an IEEE module
>      ...............^^^^
>      and makes no explicit use of an IEEE feature might be compliled to
>      run fully or partially on a GPU that does not support any IEEE flags.
> 
>     Thanks                          /Stan
> 
>     -----Original Message-----
>     From: j3-bounces at j3-fortran.org <mailto:j3-bounces at j3-fortran.org>
>     [mailto:j3-bounces at j3-fortran.org
>     <mailto:j3-bounces at j3-fortran.org>] On Behalf Of John Reid
>     Sent: Friday, August 13, 2010 7:16 AM
>     To: fortran standards email list for J3
>     Subject: Re: (j3.2006) IEEE and the GPU world
> 
>     J3,
> 
>     Here is a draft interp that I constructed after following the "IEEE
>     and the GPU
>     world" thread. Does anyone have any comments before I send it in?
> 
>     Cheers,
> 
>     John.
> 
>     _______________________________________________
>     J3 mailing list
>     J3 at j3-fortran.org <mailto:J3 at j3-fortran.org>
>     http://j3-fortran.org/mailman/listinfo/j3
> 
> 

-- 
Bill Long                                           longb at cray.com
Fortran Technical Support    &                 voice: 651-605-9024
Bioinformatics Software Development            fax:   651-605-9142
Cray Inc./Cray Plaza, Suite 210/380 Jackson St./St. Paul, MN 55101