(j3.2006) illusive exaflops

Bill Long longb
Tue Feb 22 17:08:11 EST 2011


If you look back at the Gordon Bell contest winners (real programs, as 
opposed to the artificial Linpack benchmark used for the Top 500 list), 
the pattern of passing metric thresholds has been quite regular:

1988 - Gigaflops

1998 - Teraflops

2008 - Petaflops

2018 - Exaflops (projected)

So, I agree that 2015 is optimistic, but 2018 is on the historical 
track.   It is not that difficult to gang together a lot of processors - 
the real issue for a useful Exaflops system is software - how do you 
program it?  We have parallel languages now (go F2008!). The next 
challenge is codes that are resilient to hardware failure, since that 
WILL happen with the number of components involved with such a system. 
One would prefer that the hardware/OS took care of recovery for the 
user, but I imagine there will be some programmer participation required 
as well.  Of course, the one thing that we already know - the Exaflops 
code will be written in the same language as the three milestones before 
- Fortran. :)

Cheers,
Bill



On 2/21/11 4:47 PM, Jerry Wagener wrote:
> To the Fortran standards community -
>
> After reading the supercomputer article in the latest issue of IEEE
> Spectrum
> (http://spectrum.ieee.org/computing/hardware/nextgeneration-supercomputers),
> I couldn't resist sending a sort of historical "wow" to y'all.
>
> The article deals with the the kinds of computers (as many flops as
> possible) used for Fortran's world of scientific computations, and so I
> suppose it's gist is old hat in this community. But as an old geezer,
> retired from the Fortran wars for over a decade, I found it most intriguing.
>
> In the 80's for instance, for most of the decade we tracked the fastest
> computers in megaflops, and it was big news when the first gigaflops
> computer appeared. As I recall in the late 90's we were anticipating the
> amazing advent of the first teraflops computer. And now, according to
> the article, the fastest computer is about 5 petaflops. So for the past
> 3 decades we've roughly, on-average, doubled the top computing speed
> each year (10**6 to 10**15). That's even faster than Moore's Law. Amazing.
>
> But, alas, the article points out that some tough physical limits are
> raising their ugly heads which look like this era may be coming to an
> end, and with it DARPA's hopes for an exaflops computer by 2015 may not
> be realized. Though I'm reminded of the saying that somebody saying
> something's impossible is usually interrupted by somebody doing it.
> Anyway, for an old out-of-touch Fortran geezer it was an interesting read.
>
> Best wishes to all. -Jerry

-- 
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





More information about the J3 mailing list