simpleMPI problem Jetson Nano

[laserrapt0r]

In https://github.com/garyexplains/examples/blob/master/how_to_build_nvidia_jetson_gpu_cluster.md

you say:

Edit simpleMPI.cpp and alter int blockSize = 256; to:
int blockSize = 384 / commSize;

bit in simpleMPI.cpp block size is initialized after this.

if i change the order to

int commSize, commRank;
MPI_CHECK(MPI_Comm_size(MPI_COMM_WORLD, &commSize));
MPI_CHECK(MPI_Comm_rank(MPI_COMM_WORLD, &commRank));

int blockSize = 384 / commSize;
int gridSize = 10000;
int dataSizePerNode = gridSize * blockSize;

it compiles but i cannot execute it

Running on 1 nodes
CUDA error calling "cudaMemcpy(hostData, deviceOutputData, dataSize *sizeof(float), cudaMemcpyDeviceToHost)", code is 702
Test FAILED

MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 702.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.

any idea? the vanilla version is working fine