Next: Performance modelling Up: Calculator Previous: Chain / Star
Mesh / Hypercube
Under this links you can find more about mesh or hypercube architectures and the communication algorithms. It is assumed that the results are not returned to the originator.
The Mesh / Hypercube tab contains two fields.
In the first one you have to fill in the fields described below.
For the Mesh the fields are:
Number of communication ports (p) - number of simultaneously used communication ports
Number of layers (k) - number of moves in Peters - Syska algorithm (note, that originator is layer 0 but mesh with the originator only has 1 layer)
For Hypercube:
Number of dimensions (d) - number of communication links per one processor, and simultaneously, the number of hypercube dimensions
Other fields must be filled in for both architectures:
Volume of data (V) - the size of problem (number of data units, e.g. bytes)
Processing rate (A) - inverse of processing speed (time needed to process one data unit)
Communication rate (C) - inverse of communication speed (time needed to send one data unit)
Startup (S) - startup time (communication delay incurred during the initialization of the connection)
After filling in all the fields you can click Find solution button to generate a solution for your data.
Then the second field will be shown. It contains local assignment for the processor of each layer (not for the whole layer) and the schedule length.