Compiler Driven Code Comments and

Abstract. Helping programmers write parallel software is an urgent problem given the popularity of multi-core architectures. Engineering compilers which automatically parallelize and vectorize code has turned out to be very challenging. Compilers are not powerful enough to exploit all opportunities for optimization in a code fragment – rather, they are selective with respect to the coding patterns they optimize. However, even minor code changes may enable optimizations to succeed where they previously would not. We present an interactive approach and a tool set which leverages ad-vanced compiler analysis and optimizations while retaining programmer control over the source code and its transformation. Our tool-set pro-vides comments describing why optimizations do not apply to a code fragment and suggests workarounds which can be applied automatically. We demonstrate the ability of our tool to transform code, and suggest code refactoring that increase its amenability to optimization. Experiments with an image processing benchmark shows that automatic loop parallelization of the original code result in minor speed-ups or sig-nificant slow- downs with the GNU C compiler, gcc. With our approach, automatic parallelization yields 8.6 best-case speedup over the sequen-tial version. The transformations and suggestions are based on the loop parallelization and matrix reorganization capabilities in the latest pro-duction release of gcc.