Panagiotis E. Hadjidoukas

Monte Carlo single-cell dosimetry of Auger-electron emitting radionuclides

A hybrid Monte Carlo transport scheme combining event-by-event and condensed-history simulation with a full account of energy-loss straggling was used to study the dosimetric characteristics of the Auger-emitting radionuclides 67Ga, 99mTc, 111In, 123I, 125I and 201Tl at the single-cell level. The influence of the intracellular localization of the Auger radionuclide upon cellular S-values, radial dose rate profiles and dose-volume-histograms (DVHs) was investigated. For the case where the radiopharmaceutical was either internalized into the cytoplasm or remained bound onto the cell surface (non-internalized), the dose to the cell nucleus was found to differ significantly from the MIRD values and other published data. In this case, the assumption of a homogeneous distribution throughout the cell is shown to significantly overestimate the nuclear dose. A dosimetric case study relevant to the radioimmunotherapy of single lymphoma B-cells with 125I and 123I is presented.

A microbenchmark study of OpenMP overheads under nested parallelism

In this work we present a microbenchmark methodology forassessing the overheads associated with nested parallelism in OpenMP.Our techniques are based on extensions to the well known EPCC microbenchmarksuite that allow measuring the overheads of OpenMPconstructs when they are effected in inner levels of parallelism. Themethodology is simple but powerful enough and has enabled us to gaininteresting insight into problems related to implementing and supportingnested parallelism. We measure and compare a number of commercialand freeware compilation systems. Our general conclusion is that whilenested parallelism is fortunately supported by many current implementations,the performance of this support is rather problematic. Thereseem to exist issues which have not yet been addressed effectively, asmost OpenMP systems do not exhibit a graceful reaction when made toexecute inner levels of concurrency.

Subcellular S-factors for low-energy electrons: a comparison of Monte Carlo simulations and continuous-slowing-down calculations.

Purpose: To study the energy deposition by low-energy electrons in submicron tissue-equivalent targets by comparing two widely used methodologies, namely, the continuous-slowing-down-approximation (CSDA) convolution integral and the Monte Carlo (MC) simulation. Methods: An MC track-structure code that simulates collision-by-collision the complete slowing down process is used to calculate the energy deposition in spherical volumes of unit density water medium. Comparisons are made with calculations based on the CSDA convolution integral using both empirical and MC-based range-energy analytic formulae. Results: We present self-irradiation absorbed fractions and S-factors for monoenergetic electrons of initial energies from 0.1–10 keV distributed uniformly in spheres of 5, 10, 50, 100, 500, and 1000 nm radius. The MC and CSDA results were found, in some cases, to differ by a factor of 2 or more; differences generally increase with decreasing sphere size. Contrary to high energies, the uncertainties associated with the straight-ahead approximation implicit in the CSDA calculations are of the same order as those related to straggling and δ-ray effects. Conclusion: The use of the CSDA methodology may be unsuitable for the sub-micron scale where a more realistic description of electron transport becomes important.

Nested parallelism in the OMPI OpenmP/C compiler

This paper presents a new version of the OMPi OpenMP C compiler, enhanced by lightweight runtime support based on user-level multithreading. A large number of threads can be spawned for a parallel region and multiple levels of parallelism are supported efficiently, without introducing additional overheads to the OpenMP library. Management of nested parallelism is based on an adaptive distribution scheme with hierarchical work stealing that not only favors computation and data locality but also maps directly to recent architectural developments in shared memory multiprocessors. A comparative performance evaluation of several OpenMP implementations demonstrates the efficiency of our approach.

A Monte Carlo study of absorbed dose distributions in both the vapor and liquid phases of water by intermediate energy electrons based on different condensed-history transport schemes.

Monte Carlo transport calculations of dose point kernels (DPKs) and depth dose profiles (DDPs) in both the vapor and liquid phases of water are presented for electrons with initial energy between 10 keV and 1 MeV. The results are obtained by the MC4 code using three different implementations of the condensed-history technique for inelastic collisions, namely the continuous slowing down approximation, the mixed-simulation with delta-ray transport and the addition of straggling distributions for soft collisions derived from accurate relativistic Born cross sections. In all schemes, elastic collisions are simulated individually based on single-scattering cross sections. Electron transport below 10 keV is performed in an event-by-event mode. Differences on inelastic interactions between the vapor and liquid phase are treated explicitly using our recently developed dielectric response function which is supplemented by relativistic corrections and the transverse contribution. On the whole, the interaction coefficients used agree to better than approximately 5% with NIST/ICRU values. It is shown that condensed phase effects in both DPKs and DDPs practically vanish above 100 keV. The effect of delta-rays, although decreases with energy, is sizeable leading to more diffused distributions, especially for DPKs. The addition of straggling for soft collisions is practically inconsequential above a few hundred keV. An extensive benchmarking with other condensed-history codes is provided.

Design and Implementation of OpenMP Tasks in the OMPi Compiler

In this paper we present the design and implementation of tasks in the context of the \ompi\ \openmp\ compiler. The modular architecture of \ompis runtime system allows a wide range of choices for experimenting with \openmp\ structures. We present two fully-fledged implementations of tasks: one based on \posix\ threads, with the addition of a tasking layer, and another one based on an almost unmodified user-level threading library. Both allow the tuning of their scheduling parameters so as to optimize memory consumption and execution times, resulting in highly competitive performance.

A Monte Carlo study of cellular S-factors for 1 keV to 1 MeV electrons

A systematic study of cellular S-factors and absorbed fractions for monoenergetic electrons of initial energy from 1 keV to 1 MeV is presented. The calculations are based on our in-house Monte Carlo codes which have been developed to simulate electron transport up to a few MeV using both event-by-event and condensed-history techniques. An extensive comparison with the MIRD tabulations is presented for spherical volumes of 1-10 microm radius and various source-to-target combinations relevant to the intracellular localization of the emitted electrons. When the primary electron range is comparable to the sphere radius, we find significantly higher values from the MIRD, while with increasing electron energy the escape of delta-rays leads gradually to the opposite effect. The largest differences with the MIRD are found for geometries where the target region is at some distance from the source region (e.g. surface-to-nucleus or cytoplasm-to-nucleus). The sensitivity of the results to different transport approximations is examined. The grouping of inelastic collisions is found adequate as long as delta-rays are explicitly simulated, while the inclusion of straggling for soft collisions has a negligible effect.

A numerical differentiation library exploiting parallel architectures

We present a software library for numerically estimating first and second order partial derivatives of a function by finite differencing. Various truncation schemes are offered resulting in corresponding formulas that are accurate to order O (h), O (h 2 ) ,a ndO (h 4 ), h being the differencing step. The derivatives are calculated via forward, backward and central differences. Care has been taken that only feasible points are used in the case where bound constraints are imposed on the variables. The Hessian may be approximated either from function or from gradient values. There are three versions of the software: a sequential version, an OpenMP version for shared memory architectures and an MPI version for distributed systems (clusters). The parallel versions exploit the multiprocessing capability offered by computer clusters, as well as modern multi-core systems and due to the independent character of the derivative computation, the speedup scales almost linearly with the number of available processors/cores.

A Runtime Library for Platform-Independent Task Parallelism

With the increasing diversity of computing systems and the rapid performance improvement of commodity hardware, heterogeneous clusters become the dominant platform for low-cost, high-performance computing. Grid-enabled and heterogeneous implementations of MPI establish it as the de facto programming model for these environments. On the other hand, task parallelism provides a natural way for exploiting their hierarchical architecture. This hierarchy has been further extended with the advent of general-purpose GPU devices. In this paper we present the implementation of an MPI-based task library for heterogeneous and GPU clusters. The library offers an intuitive programming interface for multilevel task parallelism with transparent data management and load balancing. We discuss design and implementation issues regarding heterogeneity support and report performance results on heterogeneous cluster computing environments.

Parallelization of a hierarchical data clustering algorithm using OpenMP

This paper presents a parallel implementation of CURE, an efficient hierarchical data clustering algorithm, using the OpenMP programming model. OpenMP provides a means of transparent management of the asymmetry and non-determinism in CURE, while our OpenMP runtime support enables the effective exploitation of the irregular nested loop-level parallelism. Experimental results for various problem parameters demonstrate the scalability of our implementation and the effective utilization of parallel hardware, which enable the use of CURE for large data sets.