How to Run Code Coverage Analysis
ICC
- Build the code as normal with the
-prof-gen=srcpos -prof-dir=PATH_TO_PROF_DIR/prof-infocompiler flag. WherePATH_TO_PROF_DIRis the path to a directory that you have created to store the profiling information. - Run the test/simulation as normal and this should generate a
.dynfile inprof-info, which contains dynamic profile information. - Navigate to the
prof-infodirectory. - Use the
profmergetool to merge all.dynfiles into one.dpifile. This merges multiple.dynfrom successive runs of the application with different runtime parameters (e.g. runningtest27cellswith varying number of particles per cell):
profmerge -a(Note: If you only want code coverage from the last test run remove old .dyn files and the .dpi file in the prof-info directory. As old test runs might cover code that was not covered in the latest test run.)
- Run the code coverage tool:
codecovThis will generate an HTML file called CODE_COVERAGE.HTML, which contains the code coverage report and show how much of the application was actually executed, line by line.
- If you didn't specify the
-prof-dirflag you can run the code coverage tool and specify the path to SWIFT library.spifile instead:
codecov -spi swiftsim/src/pgopti.spi(However this will not contain code coverage for the test source code itself.)
- A more detailed guide can be found at https://software.intel.com/en-us/node/522743
GCC
- Build the code as normal with
-Wall -fprofile-arcs -ftest-coveragecompiler flags. - Run the test/simulation as normal and this should generate a
.gcdafile, which contains dynamic profile information. - Run the
gcovtool on the executable source file:
gcov test27cells.cThis will generate annotated versions of each original source file appended with .gcov.
- Each
.gcovfile contains counts of the number of times each line was executed. Lines which were not executed are marked with hashes ‘######’. - A more detailed guide can be found at http://www.network-theory.co.uk/docs/gccintro/gccintro_81.html.