VectorCAST CUDA Example
To allow for an intuitive workflow, VectorCAST/QA has been extended to automatically build a project that has N separate coverage environments: one for the host, and one for each of the GPUs. Combining these N environments into a single project allows you to build an instrumented application that can collect coverage for all architectures. As test cases are run using VectorCAST/QA, the code coverage information is “split” into the N projects, and combined at the project level. This means that you can easily view code coverage for either the complete file or for the architecture-specific parts independently.
This folder has a CUDA example for VectorCAST.
- Download the KB_1476_cuda-demo.zip to the working directory
- Untar the demo
- Open the localization.sh file and edit to fit the current working directories
- The source files for the demo is currently already included with the release
- Edit VECTOR_LICENSE_FILE to the license file/server in use
- Edit VECTORCAST_DIR to the location of VectorCAST installed
- Optional: Edit MYPROJECTNAME, testdirname and CUDABIN if not satisfied with the default
- Save the changes
- Run the createdemo.sh file
- The script will automatically build and execute the CUDA example and open VectorCAST when done
- Open the Project via File | Open and then selecting the newly constructed environment
- Based on the capability of the machine on which VectorCAST is run, the system testing environments will have different coverage results.