We propose a matrix of application kernels and a questionnaire to study the productivity of parallel programming languages for high performance computing. The matrix presently defines ten kernels representing important algorithms or parallel processing paradigms. We ask anyone implementing to keep careful records and report key production statistics via the questionnaire. These statistics include, but are not limited to, design time, development time, data structures, debugging, maintainability, execution time, and scalability. Since the last two are functions of the target machine on which the code runs, the matrix spans three dimensions: kernel, language, and architecture.