We compute the running QCD coupling on the lattice by evaluating two-point and three-point off-shell gluon Green's functions in a fixed gauge and imposing non-perturbative renormalisation conditions on them. Our exploratory study is performed in the quenched approximation at β = 6.0 on 16 4 and 24 4 lattices. We show that, for momenta in the range 1.8 − 2.3 GeV, our coupling runs according to the two-loop asymptotic formula, allowing a precise determination of the corresponding Λ parameter. The role of lattice artifacts and finite-volume effects is carefully analysed and these appear to be under control in the momentum range of interest. Our renormalisation procedure corresponds to a momentum subtraction scheme in continuum field theory, and therefore lattice perturbation theory is not needed in order to match our results to the M S scheme, thus eliminating a major source of uncertainty in the determination of α M S . Our method can be applied directly to the unquenched case.2

SUMMARYIncreasing interest is being shown in the use of Java for large scale or Grande applications. This new use of Java places specific demands on the Java execution environments that could be tested and compared using a standard benchmark suite. We describe the design and implementation of such a suite, paying particular attention to Java-specific issues. Sample results are presented for a number of implementations of the Java Virtual Machine (JVM).

The current trend in HPC hardware is towards clusters of shared-memory (SMP) compute nodes. For applications developers the major question is how best to program these SMP clusters. To address this we study an algorithm from Discrete Element Modeling, parallelised using both the message-passing and shared-memory models simultaneously ("hybrid" parallelisation). The natural loadbalancing methods are different in the two parallel models, the shared-memory method being in principle more efficient for very load-imbalanced problems. It is therefore possible that hybrid parallelism will be beneficial on SMP clusters. We benchmark MPI and OpenMP implementations of the algorithm on MPP, SMP and cluster architectures, and evaluate the effectiveness of hybrid parallelism. Although we observe cases where OpenMP is more efficient than MPI on a single SMP node, we conclude that our current OpenMP implementation is not yet efficient enough for hybrid parallelism to outperform pure message-passing on an SMP cluster. MotivationFor a long time the HPC market was dominated by large, special-purpose MPPs with single-CPU nodes. More recently the shared-memory (SMP) architecture has grown in importance for HPC as advances in technology have allowed increasing numbers of CPUs to have access to a single memory space. In the future it appears that high-end HPC machines will have both MPP and SMP-like features. MPP manufacturers are replacing the single processors in their existing architectures with more powerful SMP nodes, while large SMPs are being clustered to go beyond the architectural limits on the number of CPUs in a single box. The fact that the machines being developed under the US ASCI program are of this type is clear evidence of the trend. Although large SMP clusters are straightforward to construct, the major question for the applications programmer is how to write portable parallel codes that run and scale efficiently.Codes written in a parallel language like HPF must rely on the compiler for efficiency, whereas those purely using multi-threading will require special OS support in order to see a single system image and memory space across the whole SMP cluster. Although both these approaches are possible, we do not envisage them being widespread in the immediate future and they are not considered here.Pure message-passing codes should port easily to any SMP cluster. It is necessary to use message-passing to communicate between processors in different boxes, but it is not immediately apparent that it is the most efficient parallelisation technique among processors in the same box. In this paper we examine the effectiveness of using a hybrid of both message-passing and shared-memory techniques within the same application. Although this approach clearly maps most closely to the underlying target architecture it remains to be shown that it is effective in practice.Situations where this hybrid parallelisation might be of benefit include:• where the MPI implementation is unoptimised for communications within an SMP• rep...

Increasing interest is being shown in the use of Java for large scale or Grande applications.This new use of Java places specific demands on the Java execution environments that could be tested and compared using a standard benchmark suite. EPCC has taken a leading role in the Java Grande Forum work to develop a framework and methodology for such a suite. Initial results presented here show interesting differences between JVMs, demonstrating the validity of the approach. Future work will concentrate on parallel benchmarks.

We present an improved study of spectroscopic observables in the SU(2) Yang-Mills theory with two adjoint fermions. We make an improvement on the precision of previous results which clarify the scale of finite volume effects present. This analysis adds to the evidence for near-conformal dynamics of this theory, while indicating a preference for a low anomalous mass dimension of the massless theory. CERN-PH-TH Minimal Walking TechnicolorThe mechanism of dynamical electro-weak symmetry breaking (often referred to as technicolor) [1] remains a possible explanation for the breaking of the electro-weak symmetry observed in nature. Gauge theories which possess an approximate infra-red fixed point have been proposed as preferred candidates for the technicolor sector in models of extended technicolor [2] as it has been argued that they would allow suppression of flavour changing neutral currents while permitting generation of large fermion masses. This is achieved through the non-trivial renormalisation dynamics between the technicolor and extended technicolor scales. This property of the theory has been coined "walking" in reference to the slow running of the coupling between the two scales 1 .We can hope to generate an infra-red fixed point in a gauge theory, while minimising contributions to electroweak precision constraints, by adding a low number of fermion flavours in higher gauge representations to a gauge theory with a low number of colours. For these reasons the gauge theory theory with symmetry group SU(2) and two flavours of Dirac fermion in the adjoint representation has been termed minimal walking technicolor (MWT) [4].Conformal or near-conformal dynamics have not been conclusively identified in any theory as yet, although there are encouraging hints from a number of sources. Due to the non-perturbative nature of the problem, much of the investigation of this problem has arisen from lattice simulations, and this activity is growing. Some recent lattice studies of MWT [5,6,7] have attempted to identify a near-conformal behaviour directly from the behaviour of the coupling and anomalous dimensions of the theory under renormalisation flow. Others [8,9,10,11,12,13], including this work, perform measurements of physical observables in the theory and from their behaviour attempt to identify signals of near-conformal dynamics.This study builds on previous work in [9,13]. In particular we seek to support these results and establish their reliability by investigating the effect of performing measurements and analysis using alternative methods. In addition, through this we expect to ascertain the scale of systematic uncertainties present, which have as yet been largely unexplored. Signals of ConformalityThe question of whether MWT in the chiral limit posesses an actual infra-red fixed point, i.e. it lies within the conformal window, or an approximate fixed point, has not been conclusively answered. Nevertheless, it is clear that MWT with a non-zero fermion mass and defined in a finite volume, as simulated on the lat...