We are excited to announce the release of Codee 1.6, which further improves the capabilities of our platform for automatic code inspection specifically designed to improve the performance of C/C++/Fortran applications.
Codee 1.6 features an improved Codee Vectorization Engine that is integrated with the compiler vectorization report. Since this version, the Codee set-up requires the specification of the target compiler used to build the C/C++/Fortran code. When a Codee command-line tool is invoked, first it runs the compiler behind the scenes in order to generate the compiler’s vectorization report. Next, this report is parsed to select the loops not vectorized automatically by the compiler, which finally are analyzed with the Codee Vectorization Engine. Quickly assess the additional vectorization opportunities available in your C/C++/Fortran code using the screening report of the pwreport tool.
For the first time, Codee 1.6 brings integrations beyond the GNU and LLVM compilers, including the Intel and Microsoft compilers as well. The feedback from beta testers confirmed the interest of these integrations in the scope of the software development life cycle in Linux and Windows. Recent and old versions of the following compilers are now supported: GNU gcc, LLVM clang and clang-cl, Intel icc and Microsoft cl. Check the support for your compiler version in the Codee Product Datasheet’s technical specifications.
Ease-of-use of Codee “auto mode” is greatly improved through a new report for automatic generation of vectorized code and verification of its correctness. The pwdirectives tool in “auto mode” proceeds in two steps. First, it automatically modifies the source code by adding compiler pragmas explicitly for the vectorizable loops overlooked by the compiler, leveraging the new Codee Vectorization Engine integrated with the compiler. And second, it invokes the compiler again to produce its vectorization report for the optimized source code and verify the actual vectorization of the loops through the messages reported by the compiler. Take a look at the new report produced by the pwdirectives tool used in”auto mode”.
With all these new features, Codee 1.6 brings a wider coverage of MBedTLS by speeding up by 25% the AES encryption algorithms XTS, CMAC and CBC. For the details, please download the latest version of our leaflet about the optimization of the performance of the MBedTLS open-source projects.
Download MBedTLS use case >
Daniel Lopez says
Great news!