To help understand how parallel performance is affected by the use of the NCORES and MPICORES parameters we will examine several cases for each and try to explain why individual problems are able to achieve better parallel performance than others, and how differences in using the two parameters can affect overall speed. The goal of this section is to help the user better understand how to use these parameters to achieve maximum performance benefit for their own models.

The three cases that we will be studying vary in size and scope to give a feel for how different problems benefit from both the MPICORES and NCORES parameters. All three cases use multiple user classes since we can only make comparisons for the MPICORES parameter in such cases. While we won't go into specific details on the full model of each case, the important data aspects for each are:

1. Case 1 — Medium-size static estimation problem with three user classes, 4681 zones and 954 screenline counts

2. Case 2 — Medium-size dynamic estimation problem with two user classes, 57 Zones, 12 time intervals, 44 screenline counts (class 1) and 42 screenline counts (class 2).

3. Case 3 — Very large static estimation problem with nine user classes, 10,000+ Screenline counts, and 6000+ zones

We will first examine using NCORES and MPICORES individually, and finally bring the two together to show how the maximum parallel performance benefit can be achieved through a combination of the two.

You may continue to:

• NCORES case study

• MPICORES case studies

• Using MPICORES and NCORES together