Nvidia says Larrabee is Intel's reaction to CUDA

http://www.custompc.co.uk/news/60449...-to-cuda..html

Nvidia: Larrabee is a reaction to CUDA

Posted at: 4:26pm 21st July 2008 by Ben Hardwidge

Nvidia responds to Pat Gelsinger’s comments about CUDA being just a
‘footnote’ in computing history

Intel may have put the wind up the graphics business with the
development of its Larrabee graphics chip, but Nvidia reckons that
Larrabee is just a reaction to what Nvidia has already achieved with
its GPGPU CUDA technology. What’s more, the comments from Intel’s Pat
Gelsinger earlier this month have also stirred up a debate about the
future of multi-core programming.

Nvidia’s general manager of its GPU computing group, Andy Keane, told
Custom PC that the high level of interest in CUDA 'is causing
Larrabee. Larrabee’s the reaction.’ He then added that ‘these comments
from Gelsinger; if we were not making a lot of headway do you think
he’d even give us a moment’s notice? No. It’s because he sees a lot of
this activity. The strategy is to try to position it [CUDA] as
something scary and unique, and it’s really not; it’s something that’s
very accessible.’

Gelsinger said that CUDA would end up in the ‘interesting footnotes in
the history of computing annals – they had great promise and there
were a few applications that were able to take advantage of them.’ He
then added that ‘generally an evolutionary compatible computing model,
such as we’re proposing with Larrabee, we expect will be the right
answer long term.’

However, Nvidia says that Gelsinger’s comments were misleading. ‘We
use common languages,’ says Keane, ‘and this is where the Gelsinger
information is totally misinformed, because it [CUDA] is standard C.
It is actually the open 64 compiler which was originally designed for
the Itanium – that’s our compiler. We’re actually using a CPU
compiler, but we’ve given it a set of rules that basically say “if you
write your program this way, it will scale across a few cores, or
hundreds of cores.”’

‘Industry standard languages always live,’ added Keane, ‘that’s kind
of the misinformation from Gelsinger. We’re just C, and CUDA’s just a
set of rules around C.’ Keane was also keen to point out that Intel
was also behind the times when Anwar Ghuloum, a principal engineer
with Intel's Microprocessor Technology Lab, said: ‘developers should
start thinking about tens, hundreds, and thousands of cores now.’

‘There’s the Gelsinger point, which is inconsistent in itself,’ said
Keane, but there’s also the other guy who said that programmers should
get ready for thousands of cores. Well, we already have a programming
language; that’s the goal of CUDA. We’re already here with hundreds of
cores - now we’re at 240 processor cores.’

Larrabee is a forthcoming discrete graphics chip from Intel that’s
based on multiple x86 cores rather than stream processors. It will be
compatible with standard 3D APIs, including DirectX and Open GL, but
Gelsinger also promised ‘a broad set of new programming models to go
with it.’ Many have speculated that Larrabee could also be used for
ray tracing and physics calculations, as well as other tasks. Little
is known about Larrabee yet, but more details will be released later
this year at Siggraph.