CUDA to x86 compiler, project Ocelot

http://code.google.com/p/gpuocelot/

by jacquesm • 17 years ago

26 20 17 years ago

code.google.com

newhouseb• 17 years ago

For those not familiar with CUDA vocabulary:

PTX = Parallel Thread Execution is a pseudo-assembly language used in nVidia's CUDA programming environment. The 'nvcc' compiler translates code written in CUDA, a C-like language, into PTX, and the graphics driver contains a compiler which translates the PTX into something which can be run on the processing cores.

(source - wikipedia)

andrewcooke6• 17 years ago

so what happens if you go ptx -> llvm -> opencl -> nvidia gpu? how does the speed change?

Raphael_Amiard• 17 years ago

this shouldn't happen but it would be an interresting comparison nevertheless

andrewcooke6• 17 years ago

it seems possible from the diagram on the linked page. why shouldn't it happen?

i am curious whether the analysis stages can improve the code.

profquail2• 17 years ago

nVidia's latest OpenCL driver actually uses LLVM for compiling device code.

If you want to read more: http://llvm.org/devmtg/2009-10/Grover_PLANG.pdf

CUDA to x86 compiler, project Ocelot

Related Stories