Integration and System Testing

We noted above that the testing of scientific applications is done at the system level (a) because the focus is on results and speed and (b) because the software will in all probability use well maintained libraries for the major computational work. At the system level testing can be broken down into four parts which we will describe in the remainder of this section.

Validation

Do the results tally with experiments or expert user expectations?

Debugging

What to do if the software doesn't work or the results are not as expected.

Code coverage analysis

Are all parts of the software exercised by test cases?

Performance analysis

Where does the code spend its time, where should it be changed to speed things up?