[BUG]: P2P direction printing information checking error

[tianxluo]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this bug and that I agree to the Code of Conduct

Type of Bug

Silent Failure

Component

cuda.bindings

Describe the bug

cuda-python/examples/0_Introduction/simpleP2P_test.py: line

p2pCapableGPUs = [-1, -1]
    for i in range(gpu_n):
        p2pCapableGPUs[0] = i
        for j in range(gpu_n):
            if i == j:
                continue
            i_access_j = checkCudaErrors(cudart.cudaDeviceCanAccessPeer(i, j))
            j_access_i = checkCudaErrors(cudart.cudaDeviceCanAccessPeer(j, i))
            print(
                "> Peer access from {} (GPU{}) -> {} (GPU{}) : {}\n".format(
                    prop[i].name, i, prop[j].name, j, "Yes" if i_access_j else "No"
                )
            )
            print(
                "> Peer access from {} (GPU{}) -> {} (GPU{}) : {}\n".format(
                    prop[j].name, j, prop[i].name, i, "Yes" if i_access_j else "No"
                )
            )
            if i_access_j and j_access_i:
                p2pCapableGPUs[1] = j
                break
        if p2pCapableGPUs[1] != -1:
            break

For GPU[j]->GPU[i], I think we should use the variable 'j_access_i' rather than 'i_access_j'. Btw, I think the extra/isoFDmodelling_test.py has the same issue, please double check this .py file https://github.com/NVIDIA/cuda-python/blob/main/cuda_bindings/examples/extra/isoFDModelling_test.py#L645 when you feel free.

How to Reproduce

This is a logging issue when you disable the peer access from GPU[j] to GPU[i] and print this logging information

Expected behavior

It should be j_access_i for GPU[j]->GPU[i].

Operating System

No response

nvidia-smi output

No response