Refine
Year of publication
Document Type
- ZIB-Report (18)
- Article (2)
- In Proceedings (1)
Is part of the Bibliography
- no (21)
Keywords
- QCD (5)
- parallel tempering (4)
- topological charge (3)
- Aoki phase (2)
- Wilson fermions (2)
- parity-flavor symmetry (2)
- phase diagram (2)
- Benutzertreffen (1)
- Grid computing (1)
- HLRN (1)
We study performance and scaling of the Berlin Quantum Chromodynamics Program (BQCD) on the SGI Altix 4700 at Leibniz Supercomputing Centre (LRZ). We employ different communication methods (MPI, MPI with two OpenMP threads per process, as well as the shmem library) and run the MPI version on the two types of nodes of that machine. For comparison with other machines we made performance measurements on an IBM p690 cluster and a Cray XT4.
The latest machine generation installed at supercomputer centres in Germany offers a peak performance in the tens of Tflop/s range. We study performance and scaling of our quantum chromodynamics simulation programme BQCD that we obtained on two of these machines, an IBM Blue Gene/L and an SGI Altix 4700. We compare the performance of Fortran/MPI code with assembler code. The latter allows to exploit concurrency at more levels, in particular in overlapping communication and computation as well as prefetching data from main memory.
In this letter we report on a numerical investigation of the Aoki phase in the case of finite temperature which continues our former study at zero temperature. We have performed simulations with Wilson fermions at $\beta=4.6$ using lattices with temporal extension $N_{\tau}=4$. In contrast to the zero temperature case, the existence of an Aoki phase can be confirmed for a small range in $\kappa$ at $\beta=4.6$, however, shifted slightly to lower $\kappa$. Despite fine-tuning $\kappa$ we could not separate the thermal transition line from the Aoki phase.
Efficient implementations of irregular problems on vector and parallel architectures are generally hard to realize. An important class of problems are Gauß-Seidel iteration schemes applied to irregular data sets. The unstructured data dependences arising there prevent restructuring compilers from generating efficient code for vector or parallel machines. It is shown, how to structure the data dependences by decomposing the underlying data set using graph coloring techniques and by specifying a particular execution order already on the algorithm level. Methods to master the irregularities originating from different types of tasks are proposed. An application is given and some open issues and future developments are discussed.
Hasenbusch has proposed splitting the pseudo-fermionic action into two parts, in order to speed-up Hybrid Monte Carlo simulations of QCD. We have tested a different splitting, also using clover-improved Wilson fermions. An additional speed-up between 5 and 20\% over the original proposal was achieved in production runs.
UMTS is a 3rd generation mobile telecommunication system which enables multi-service and multi-bit rate communication going beyond the possibilities of previous systems. The simulator MoDySim models UMTS in great detail. Characteristics of UMTS such as soft hand-over and the interdependency of load and capacity among neighbouring cells are challenges for the parallelisation of such a system. In this paper we explain how the software was parallelised and present performance results of a UMTS simulation for the city of Berlin.