The high performance Fortran handbook
The high performance Fortran handbook
Application and architectural bottlenecks in large scale distributed shared memory machines
ISCA '96 Proceedings of the 23rd annual international symposium on Computer architecture
Operating system support for improving data locality on CC-NUMA compute servers
Proceedings of the seventh international conference on Architectural support for programming languages and operating systems
Data distribution support on distributed shared memory multiprocessors
Proceedings of the ACM SIGPLAN 1997 conference on Programming language design and implementation
The SGI Origin: a ccNUMA highly scalable server
Proceedings of the 24th annual international symposium on Computer architecture
Scaling application performance on a cache-coherent multiprocessor
ISCA '99 Proceedings of the 26th annual international symposium on Computer architecture
A high-level abstraction of shared accesses
ACM Transactions on Computer Systems (TOCS)
A case for user-level dynamic page migration
Proceedings of the 14th international conference on Supercomputing
Automation of Data Traffic Control on DSM Architectures
ICCS '01 Proceedings of the International Conference on Computational Science-Part II
LCR '00 Selected Papers from the 5th International Workshop on Languages, Compilers, and Run-Time Systems for Scalable Computers
HiPC'06 Proceedings of the 13th international conference on High Performance Computing
| Hi-index | 0.00 |
This paper describes transparent mechanisms for emulating some of the data distribution facilities offered by traditional data-parallel programming models, such as High Performance Fortran, in OpenMP. The vehicle for implementing these facilities in OpenMP without modifying the programming model or exporting data distribution details to the programmer is user-level dynamic page migration [9,10]. We have implemented a runtime system called UPMlib, which allows the compiler to inject into the application a smart user-level page migration engine. The page migration engine improves transparently the locality of memory references at the page level on behalf of the application. This engine can accurately and timely establish effective initial page placement schemes for OpenMP programs. Furthermore, it incorporates mechanisms for tuning page placement across phase changes in the application communication pattern. The effectiveness of page migration in these cases depends heavily on the overhead of page movements, the duration of phases in the application code and architectural characteristics. In general, dynamic page migration between phases is effective if the duration of a phase is long enough to amortize the cost of page movements.