Citation Context

...hese systems indicate that implementations of parallel programs with MPI perform often better than implementations of the same programs with OpenMP. This is true especially for large industrial codes =-=[14]-=-. The most prominent problem that OpenMP are faced with on scalable sharedmemory multiprocessors is the non-uniformity of memory access latencies (NUMA). Although the shared-memory communication abstr...

Citation Context

... each other in non-trivial ways. But recent developments show that these issues can also be addressed in HPF [2, 4]. Nevertheless, dynamic scheduling strategies show non-negligible overhead, e.g. see =-=[10]-=-, and non-trivial task interaction reduces scalability. In contrast to other approaches where the HPF and OpenMP programming model are combined, we rely on a uniform programming model that is appropri...

Citation Context

...scribe their advantages and disadvantages. A number of talks and presentations have been given on comparing OpenMP and MPI and on mixed mode programming styles. For further information see references =-=[1, 2, 3, 4, 5, 6, 7]-=-. 3.1 MPI The message passing programming model is a distributed memory model with explicit control parallelism. This uses an SPMD model. Processes are only able to read and write to their respective ...

Citation Context

...xperience with real codes suggests that it is difficult to scale an OpenMP program to tens or hundreds of processors and that complex OpenMP codes tend to scale worse than the corresponding MPI codes =-=[2, 13, 31]-=-. Some researchers have pin-pointed this effect as a problem of the overhead of managing parallelism in OpenMP, which includes thread creation and synchronization. This overhead is an important perfor...

Citation Context

...nMP has exhibited performance inferior to that of message-passing and data parallel paradigms such as MPI and HPF, primarily due to the inability of the programming model to control data distribution =-=[1,12]-=-. OpenMP provides no means to the programmer for distributing data among processors. Although automatic page placement schemes at the operatingsystem level, such as first-touch and round-robin, are of...

Citation Context

...easonably [12]. Several application studies with the OpenMP programming standard for shared memory multiprocessors have reported significant performance losses due to poor page placement and locality =-=[5; 21]-=-. Nevertheless, none of these studies leveraged dynamic page migration to alleviate these problems. On the contrary, researchers have reverted to programming models based on message-passing such as HP...

Citation Context

...within the same application (mixed mode programming) may provide a more efficient parallelization strategy than pure MPI. Whilst SMP clusters offer the greatest reason for developing a mixed mode code=-=[8]-=-[9][10][11][12], both the OpenMP and MPI paradigms have different advantages and disadvantages and by developing such a model these characteristics might even be exploited to give the best performance...

Citation Context

...operations must be properly scheduled by the programmer such that no two threads try towrite to the same memory location. Critical sections can be used but being highly expensive they are discouraged =-=[50]-=-. The tasks that require threading are the inner list update, the electron-density computation, and the force computation. Each of the above steps comprises an outer-loop which runs through all the at...