We are excited to announce the availability Parallelware Analyzer 0.15. This release has the following new features:
- Memory access patterns reporting. Now pwloops provides information not only about computational parallel patterns but also about memory access patterns. You can find more information about the patterns in our knowledge base.
- Use OpenMP teams to offload to GPU. When using pwdirectives to generate OpenMP offloading directives you can now select to use OpenMP teams which is the recommended practice for GPU accelerators.
- JSON Compilation databases are now accepted configuration files. We’ve received feedback that supporting compile_commands.json files generated by tools such as CMake is really useful, but having to create a minimal configuration file to tell Parallelware Analyzer to use it was a bit of a hassle. Now you can pass the path to the compile_commands.json file as an argument for –config.
- New options to choose what language to analyze. You can specify a supported programming language to only analyze source code files of that language. Additionally, you can now tell Parallelware Analyzer to analyze header files just like regular source files (by default, they are only analyzed when included in regular source code files).
- New defects and recommendations. Parallelware Analyzer checks catalog has been extended with the following defect and recommendations:
Take a look at our quickstart to cover the configuration file either in our blog post or in the PDF file included with Parallelware Analyzer!
Early Access Program is open!
Enter the Early Access Program to have access to all versions of Parallelware Analyzer until the official release.