hi everbody,

i am having trouble within a DEFINE_ADJUST macro, with the use of node_to_host and host_to_node macros. I'm using an interpreted UDF.

The serial version of the UDF runs normally when it is run parallely - of course generating incorrect results due to lack of communication. The problem with the parallel UDF also disappears once I delete the lines containing node_to_host and host_to_node macros. So the problem is definately caused by the use of these macros.

Should I use compiled UDF instead? Or is there any other rule that I am breaching?

The error I get is "Floating point error: invalid number". And while interpreting the UDF there is no problem. I get the error when I start running the simulation.

To experiment with this error, I typed a much simpler DEFINE_ADJUST macro containing almost only a single host_to_node macro, and i am still getting the same error.

Thanks for any help.

Peace.

Karem,

Are you using a cluster of several machines, or a large SMP (shared memory) machine? I have a similiar issue, although using a compiled UDF that doesn't throw an error, but rather stops executing and seems to enter an infinite loop. I'm afraid I can't help yet, but I'd like to know if you have the same problem I do.

Many thanks,

Tom

Hi Tom,

i am using a SMP with a double core processor running on WinXP. I have also access to a Linux cluster but i haven't tried it there.

Is yours also a host<->node communication issue?

I've actually put this thing aside for a while, but will be back to it soon. Please drop a line if you find something.

Regards,

Kerem