No abstract

Directive-based programming models, such as OpenACC and OpenMP, allow developers to convert a sequential program into a parallel one with minimum human intervention. However, inserting pragmas into production code is a difficult and error-prone task, often requiring familiarity with the target program. This difficulty restricts the ability of developers to annotate code that they have not written themselves. This article provides a suite of compiler-related methods to mitigate this problem. Such techniques rely on symbolic range analysis, a well-known static technique, to achieve two purposes: populate source code with data transfer primitives and to disambiguate pointers that could hinder automatic parallelization due to aliasing. We have materialized our ideas into a tool, DawnCC, which can be used stand-alone or through an online interface. To demonstrate its effectiveness, we show how DawnCC can annotate the programs available in PolyBench without any intervention from users. Such annotations lead to speedups of over 100× in an Nvidia architecture and over 50× in an ARM architecture.

No abstract

After decades of advances in techniques of automatic parallelization, software developers can count, today, on many different tools that transform a program, so that it runs in parallel. Yet, most of these tools are still considered research artifacts. They contain latent bugs, often consequence of a sparsity of testing frameworks for autoparallelizers. This dissertation describes one such framework: AutoParBench - the product of a cooperation between UFMG's Compilers Lab, and Lawrence Livermore National Laboratory. AutoParBench is today publicly available, and has been successfully used to find three zero-day bugs in the Intel C Compiler. Its usage also uncovered problems in more research-oriented tools: 2 bugs in DawnCC, 4 in Rose AutoPar and 2 in Cetus. All these bugs have been confirmed, and some of them have been already fixed as an aftermath of this work.