telsa processor performance

hai to all

i recently purchased nvidia telsa system. in that having four telsa processors and each processor having 240 cores, but my question is how to write the kernel function on all processor on parallelly .

1 . can i specify this kernel function should run one processor ?

  1. earlier i was having geforce 8800 GT , if i use 4 telsa processor then can i get better performance for my signal processing algorithmas(FFT) ?

  2. can i write singal kernel fucntion , it can run on all processor parallely?

please any one help me

thanking you

one host thread can only bind to one GPU context, cuFFT would not be executed in 4 Tesla processor except that

you invoke 4 cuFFT host threads binding to each Tesla processor respectively.

for example, if you want to execute 10000 3-D FFT, then you can distribute workload among four Tesla processor,

each with 2500 3-D FFT.

if you want to invoke four Tesla processor, you can survey simpleMultiGPU example in SDK

this should help…