Tuesday, February 22, 2011

CUDA - Programming the future

Cuda is short for Compute Unified Device Architecture. It is a parallel comouting architecture for nVidia's GPUs. It is a computing engine provided by nVidia in its GPU so that programmers can exploit the real power of massively parallel softwares.


Due to its speed, some programmers has a different name for CUDA and that is "C with NVIDIA extensions and certain restrictions".Through Cuda ,now developers do not have to worry about how the parallelism is achieved, they just need to tell the GPU the arrangement of thread in the GPU and their execution. Threads are nothing but small processes working together to carry out a same objective. A nVidia basic GPU is capable of running million of threads for any problem.

In the not too far future where we'll be using our smartphones just to calculate a solution to some of the complex problem, CUDA will prove to be the most powerful as well as the most efficient programming ever made.

I'll discuss about some of it's basics in the posts to come.

Below is the nVidia Tesla series GPU having 448 Cuda cores and a 3GB DDR3 RAM...

3 comments:

cuda_shiva said...

Gud work.......i really like ur post about cuda.....

cuda_shiva said...

i really like ur post....

Vivek said...

thanks bro

It's just the truth.
CUDA is THE future.

I'll be posting the memory optimed codes soon.