Re: [Starpu-devel] Executing Cholesky with StarPU

[giorgos matheou <geomat88@gmail.com>]

Hello again Samuel,

Great! Thank you very much.

George Matheou

On Thu, May 11, 2017 at 3:15 PM, Samuel Thibault <samuel.thibault@inria.fr> wrote:

Hello,

giorgos matheou, on jeu. 11 mai 2017 15:00:44 +0300, wrote:
> Well I need the Cholesky benchmarks for comparison purposes. So it is OK since
> other frameworks have the same implementation.

Perhaps there is a misunderstanding: I'm saying that the implementation
shipped within StarPU may not be a so good StarPU implementation, while
Chameleon does provide a well-benchmarked StarPU implementation.

> The execution terminates with the following error:
> [n081:29365] too many retries sending message to 0x0075:0x00030d11, giving up
> [n082:25251] too many retries sending message to 0x0075:0x00030d11, giving up

This looks like an issue in the MPI implementation.

> Also, I am receiving these warnings:
> [starpu][check_bus_config_file] No performance model for the bus,
> calibrating...
> [starpu][check_bus_config_file] ... done

That is expected: the first time you run StarPU it checks the machine.

> For each node:
> [starpu][_starpu_mpi_print_thread_level_support] MPI_Init_thread level =
> MPI_THREAD_SERIALIZED; Multiple threads may make MPI calls, but only one at a
> time.

That is just informational actually, we have just removed it.

> For the execution I use these commands:
> export STARPU_CALIBRATE=1
>
> STARPU_SCHED=dmdas mpirun --npernode 1 ./starpu-1.1.7/mpi/examples/
> matrix_decomposition/mpi_cholesky_distributed -size $MATRIX_SIZE -nblocks $
> ((MATRIX_SIZE/BLOCK_SIZE))

That should be fine.

I however notice that you are using starpu 1.1.7.  The 1.1 branch is
very old, you should really used starpu 1.2.2 which has just been
released.

Samuel