As Intel gears up to sample Larrabee later this year, the chip maker continues to build hype over the architecture's x86 roots. Intel is quick to point out that developers will be able to program in C or C++ languages just as they're used to doing on x86 processors, giving them an easy way to port applications from other platforms over to Larrabee.
Meanwhile, Nvidia also wants to build hype, but over its competing CUDA architecture. DailyTech has posted Nvidia's comments on the issue, which read:
CUDA is a C-language compiler that is based on the PathScale C compiler. This open source compiler was originally developed for the x86 architecture. The NVIDIA computing architecture was specifically designed to support the C language - like any other processor architecture. Competitive comments that the GPU is only partially programmable are incorrect - all the processors in the NVIDIA GPU are programmable in the C language.
NVIDIA's approach to parallel computing has already proven to scale from 8 to 240 GPU cores. Also, NVIDIA is just about to release a multi-core CPU version of the CUDA compiler. This allows the developer to write an application once and run across multiple platforms. Larrabee's development environment is proprietary to Intel and, at least disclosed in marketing materials to date, is different than a multi-core CPU software environment.
Andrew Humber from Nvidia also went on to clarify that CUDA is a brand name for the C-compiler rather than being two different things. In addition, Nvidia has called upon C-programming gurus to "test your C coding talent on CUDA, the only C-language environment that taps in the massive processing power of GPUs."
Anyone else feel chilly when Nvidia and Intel are in the same room?
Image Credit: Nvidia
Comments are closed on this article
jwalch.hawk
August 09, 2008 at 1:28am
'Cause it's cold. Get it? Heh, yeah, it was lame... Anyways...
I'm inclined to take nVidia's side on this one. Kinda. From a pragmatic standpoint, they are in the right here. Compilers are ultimately a translating tool between the higher-level language itself (ie, C, Java, so on and so forth) and lower-level languages (like machine code or assembly). It's not like the x86 processor processes your C++ code. It processes the machine code that your C++ code got translated into. With that in mind, I see absolutely nothing preventing nVidia (well, besides maybe legal issues or something) from developing a special compiler that takes your C++ code and translates into something their GPUs like rather than something an x86 likes. It's just changing the target translation platform, right?
That all said, Intel does make a valid point to bring up that x86 is in fact an advantage because we already know how to make all the high-level languages that everyone loves work on it. This would allow applications that perhaps don't hit the GPU hard nowadays to run some more of their code on the GPU because it'd be so easy to port.
But if nVidia says that CUDA is basically just a glorified compiler for the C that so many already use... Then what's the difference?
Are Intel and nVidia both ultimately just changing different links in the equation underneath essentially the same high-level language?
Damn, I just talked myself in a circle and got nowhere fast.
Print
