13.07.2015 Views

automatically exploiting cross-invocation parallelism using runtime ...

automatically exploiting cross-invocation parallelism using runtime ...

automatically exploiting cross-invocation parallelism using runtime ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Chapter 5EvaluationThe implementations of DOMORE and SPECCROSS systems are evaluated on a singleplatform: a 24-core shared memory machine which has four Intel 6-core Xeon X7460processors running at 2.66 GHz with 24 GB of memory. Its operating system is 64-bitUbuntu 9.10. The sequential baseline compilations are performed by the clang compilerversion 3.0 at optimization level three.The benchmark programs evaluated in this dissertation are from seven benchmark suites.Table 5.1 gives their details. These programs were <strong>automatically</strong> chosen because they sharetwo characteristics: their performance dominating loop nests cannot be successfully parallelizedby parallelization techniques implemented in Liberty parallelizing compiler infrastructure,including DOALL, LOCALWRITE, DSWP and PS-DSWP. Meanwhile, althoughthese loop nests contain parallelizable inner loops, inner loop parallelization introducesfrequent barrier synchronizations limiting overall scalability. These two characteristics arerequired for DOMORE and SPECCROSS to have a potential benefit.In section 5.1 and 5.2, we demonstrate the performance improvement achieved by DO-MORE and SPECCROSS and discuss the applicability and scalability of both techniques.In section 5.3, we compare DOMORE and SPECCROSS with previous works, showing thatthrough <strong>exploiting</strong> additional <strong>cross</strong>-<strong>invocation</strong> <strong>parallelism</strong>, DOMORE and SPECCROSS are77

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!