Regarding UVA and peer-to-peer SDK code

Accelerated Computing CUDA CUDA Programming and Performance
1

Dear users,
I am facing the problem for uva and peer to peer.How can I check whether UVA is working or not? I run this sample code and got the result given below.

Related information are:
using CUDA toolkit and SDK version 4 and driver version is 280.13

…/…/bin/linux/release/simpleP2P
[simpleP2P] starting…
Checking for multiple GPUs…
CUDA-capable device count: 2
Checking for peer access…
Enabling peer access…
Checking for UVA…
Allocating buffers (64MB on GPU0, GPU1 and Host)…
Creating event handles…
cudaMemcpyPeer / cudaMemcpy between GPU0 and GPU1: 4.44GB/s
Preparing host buffer and memcpy to GPU0…
Run kernel on GPU1, taking source data from GPU0 and writing to GPU1…
Run kernel on GPU0, taking source data from GPU1 and writing to GPU0…
Copy data back to host from GPU0 and verify…
Verification error, element 6624
Verification error, element 6625
Verification error, element 6626
Verification error, element 6627
Verification error, element 6628
Verification error, element 6629
Verification error, element 6630
Verification error, element 6631
Verification error, element 6632
Verification error, element 6633
Verification error, element 6634
Verification error, element 6635
FAILED
Enabling peer access…
Shutting down…

Press ENTER to exit…

What can i conclude with this result? Is i am missing something?

Thanks in advance.

Deepti

2

I am not sure how to chekc oit but you need to have 64 bit system, TCC driver (for windows. whatever that means), Fermi class gpu, toolkit 4.0, latest drivers, if yolu have multiple gpus they must be ion the same IOH.

3

Yes, all the above condition for supporting UVA are satisfy. I am working on two Gpu Fermi based cards C2050.

Looking forward for more suggestions.

Deepti