Distributed fast and accurate simulation platform for advanced ARM- and RISC-V-based HPC systems

Transient simulation of wind farms provides an effective approach to study dynamic performance of power system with high penetration of wind power. To accelerate the transient simulations, sophisticated modeling methods for the DFIGs are required. In this paper, a general framework is proposed to develop and verify the fast simulation models of the DFIGs, which is based on practical indices designed for evaluation the accuracy and efficiency of the fast simulation. Then, a new model of the DFIG is presented, which can not only describe the nonlinear characteristics of PWM converter but also achieve high computational efficiency at the same time. Test results of a single machine infinity bus system are used to validate the proposed modeling method.

With the emergence of versatile storage systems, multi-level checkpointing (MLC) has become a common approach to gain efficiency. However, multi-level checkpoint/restart can cause enormous I/O traffic on HPC systems. To use multilevel checkpointing efficiently, it is important to optimize check-point/restart configurations. Current approaches, namely modeling and simulation, are either inaccurate or slow in determining the optimal configuration for a large scale system. In this paper, we show that machine learning models can be used in combination with accurate simulation to determine the optimal checkpoint configurations. We also demonstrate that more advanced techniques such as neural networks can further improve the performance in optimizing checkpoint configurations.

Performance simulation of software for Multiprocessor System-on-a-Chips (MPSoC) suffers from poor tool support. Cycle accurate simulation at Instruction Set Simulation level is too slow and inefficient for any design of realistic size. Behavioral simulation, though useful for functional analysis at high level, does not provide any performance information that is crucial for design and analysis of MPSoC implementations. As a consequence, designers are often reduced to manually annotate performance information onto behavioral models, which contributes further to inefficiency and inaccuracy. In this paper, we use structural performance models to provide fast and accurate simulation of software for MPSoC. We generate structural models automatically using GCC with accurate performance annotation while considering optimizations for instruction selection, branch prediction, and pipeline interlock. Our structural models are able to simulate at several orders of magnitude faster than ISS and provide less than 1% error on performance estimation. These models allow realistic MPSoC design space explorations based on performance characteristics with simulation speed comparable to behavioral simulation. We validate our simulation models with several benchmarks and demonstrate our approach with a design case study of an MPEG-2 decoder.

Main concerns related to post-layout simulation, today, are about the format of the netlist coming out from the parasitic extractor. In fact, such a netlist is usually flat so that readability, whether compared to the pre-layout hierarchical one, is very poor due to device and net names which often change and to the difficulty to compare pre-layout and post-layout output signals. Furthermore, simulating such large flat netlists is frequently time consuming because it is not possible to exploit algorithms like hierarchical array reduction (HAR) and isomorphic matching (IM), strength points of state-of-the-art full chip simulators. In this paper, we present a new approach that, starting from a flat netlist with parasitic components and a pre-layout hierarchical one, allows to create a fully hierarchical post-layout netlist containing device parameters and parasitic components directly extracted from the layout. In this way, a fast and accurate post-layout simulation is made possible by the use of look-up table simulators, taking advantages from the HAR and IM algorithms as mentioned before. This methodology has been integrated in a complete design flow to guarantee first silicon success, cut down time-to-design, improve time-to-market and streamline design quality.

In this paper we develop a framework for fast and accurate multiprocessor system performance simulation. Our simulation model generator generates simulation modules with accurate time deltas for software processes based on the intermediate representations generated by a compiler. The simulation modules are simulated as concurrent tasks in multiprocessor system performance simulation environment in SystemC. We use aggregated waits to reduce overhead in the simulation kernel and triple the speed of the simulation. Our study shows that we can obtain overall system performance results with less than 6% error while simulating at 150times faster than using an Instruction Set Simulator. This opens up system design space explorations that were not possible before.

To accurately reflect the switching characteristics of GaN High Electron Mobility Transistor (GaN HEMT), this article proposes a fast and accurate Spice circuit simulation model of GaN HEMT. The model mainly imports the saturation current modulation factor $\theta_{(V g s)}$) and focuses on the modeling accuracy of the nonlinear capacitance $C_{\text{g}\text{d}}$, which would affect the switching process. In addition, to address the present problems of relying on commercial fitting software with high copyright fees for parameter extraction and extracting model parameters very intricately by traditional modeling methods, a combined optimization method is proposed to extract the model parameters for modeling. The combined algorithm is based on the Genetic algorithm of global optimization and the Levenberg-Marquardt algorithm of the local optimum and it has the advantages of the global optimal solution and fast convergence. The proposed modeling method is verified to be fast and accurate by building a Double-Pulse Test experimental platform and comparing the simulated and experimental waveforms of the dynamic and static characteristics of GaN HEMT.

The tremendous challenge of comparing our theoretical models with the gravitational-wave observations in the new era of multimessenger astronomy requires accurate and fast numerical simulations of complicated physical systems described by the Einstein and the matter equations. These requirements can only be satisfied if the simulations can be parallelized efficiently on a large number of processors and advanced computational strategies are adopted. To achieve this goal we have developed Simflowny, an open platform for scientific dynamical models which automatically generates parallel code for different simulation frameworks, allowing the use of HPC infrastructures to non-specialist scientists. One of these frameworks is SAMRAI, a mature patch-based structured adaptive mesh refinement infrastructure, capable of reaching exascale in some specific problems. Here we present the numerical techniques that we have implemented on this framework by using Simflowny in order to perform fast, efficient, accurate and highly-scalable simulations. These techniques involve high-order schemes for smooth and non-smooth solutions, Adaptive Mesh Refinement with arbitrary resolution ratios and an optimal strategy for the sub-cycling in time. We validate the automatically generated codes for the SAMRAI infrastructure with some simple test examples (i.e. wave equation and Newtonian MHD) and finally with the Einstein equations.

Power consumption is crucial in embedded systems, mainly because of the limited battery capacity and the problem of heat dissipation. The energy efficiency of System-on-Chips (SoCs) is optimized at both hardware and software level using simulation platforms. The challenge of these platforms lies in the tradeoff between accuracy and simulation speed for early architecture exploration and HW/SW validation. In the context of many-core systems in which heavy software stacks are executed, fast simulation platforms are required. In this paper, we present our power modeling approach of a complex many-core system to estimate the power consumption of software applications executed on it. We propose in particular a light and accurate power model for VLIW processors, as this kind of processor is commonly used in such systems. The power estimator we set up is part of a practical power characterization framework fully automated that includes low level simulations which are then used to back-annotate fast simulation models. Our case study is the STHORM accelerator, a clustered many-core architecture comprising dual-instruction issue processors, a complex memory hierarchy and DMA engines. Experimental results show that we can perform fast power estimations with an estimation error lower than 4% for the VLIW cores, which are the main source of power consumption, and less than 10% for the overall SoC platform, and for a simulation time overhead lower than 1%.

Low- and medium-voltage connectors are designed for a service life of more than 30 years, during which they have to withstand extreme conditions, so it is primordial ensuring their thermal performance. Mandatory standardised short-circuit tests are required to homologate electrical connectors which are conducted in singular and scarce laboratories, so it is essential to dispose of fast and accurate simulation tools to predict the thermal performance of the equipment during the design stage. This study focuses on the application of a fast and accurate simulation method to reproduce the transient thermal behaviour and to estimate the transient temperature rise and the subsequent cooling of power connectors during short circuits. To minimise the computational burden, this study proposes a fast finite-difference method approach, based on one-dimensional reduction of the analysed geometry. To improve accuracy, key three-dimensional information is retained, such as the convective coefficients, the incremental resistance or the cross-section of each node. Results attained by means of the proposed method are validated against experimental results conducted in a high-current laboratory, thus corroborating the usefulness and accuracy of the proposed method. The methodology exposed in this study can be applied to many other hardware for power lines and substations.

In order to describe the voltage and current waveforms of a switch-mode power supply (SMPS) accurately by simulation the parasitic output capacitance of the MOSFET cannot be neglected. For this reason the paper describes and compares two different simulation methods which take this capacitance into account without extending the simulation time considerably, so that a fast and accurate transient simulation of the SMPS is obtained. The first method is an analytical solution with an equivalent linear capacitance. The second method really considers the non-linear dependency of the capacitance on the voltage. The measurement at the end shows, that the second method describes the SMPS very accurately. Furthermore a comparison with a calculation which neglects the parasitic capacitance shows the relevance of the proposed method.

Dark Energy Spectroscopic Instrument (DESI) will construct a large and precise three-dimensional map of our Universe. The survey effective volume reaches $\sim 20\, h^{-3}\, \mathrm{Gpc}^{3}$. It is a great challenge to prepare high-resolution simulations with a much larger volume for validating the DESI analysis pipelines. AbacusSummit is a suite of high-resolution dark-matter-only simulations designed for this purpose, with $200\, h^{-3}\, \mathrm{Gpc}^{3}$ (10 times DESI volume) for the base cosmology. However, further efforts need to be done to provide a more precise analysis of the data and to cover also other cosmologies. Recently, the CARPool method was proposed to use paired accurate and approximate simulations to achieve high statistical precision with a limited number of high-resolution simulations. Relying on this technique, we propose to use fast quasi-N-body solvers combined with accurate simulations to produce accurate summary statistics. This enables us to obtain 100 times smaller variance than the expected DESI statistical variance at the scales we are interested in, e.g. $k \lt 0.3\, h\, \mathrm{Mpc}^{-1}$ for the halo power spectrum. In addition, it can significantly suppress the sample variance of the halo bispectrum. We further generalize the method for other cosmologies with only one realization in AbacusSummit suite to extend the effective volume ∼20 times. In summary, our proposed strategy of combining high-fidelity simulations with fast approximate gravity solvers and a series of variance suppression techniques sets the path for a robust cosmological analysis of galaxy survey data.

BackgroundBiochemical systems with relatively low numbers of components must be simulated stochastically in order to capture their inherent noise. Although there has recently been considerable work on discrete stochastic solvers, there is still a need for numerical methods that are both fast and accurate. The Bulirsch-Stoer method is an established method for solving ordinary differential equations that possesses both of these qualities.ResultsIn this paper, we present the Stochastic Bulirsch-Stoer method, a new numerical method for simulating discrete chemical reaction systems, inspired by its deterministic counterpart. It is able to achieve an excellent efficiency due to the fact that it is based on an approach with high deterministic order, allowing for larger stepsizes and leading to fast simulations. We compare it to the Euler τ-leap, as well as two more recent τ-leap methods, on a number of example problems, and find that as well as being very accurate, our method is the most robust, in terms of efficiency, of all the methods considered in this paper. The problems it is most suited for are those with increased populations that would be too slow to simulate using Gillespie’s stochastic simulation algorithm. For such problems, it is likely to achieve higher weak order in the moments.ConclusionsThe Stochastic Bulirsch-Stoer method is a novel stochastic solver that can be used for fast and accurate simulations. Crucially, compared to other similar methods, it better retains its high accuracy when the timesteps are increased. Thus the Stochastic Bulirsch-Stoer method is both computationally efficient and robust. These are key properties for any stochastic numerical method, as they must typically run many thousands of simulations.

A proof of concept methodology has been demonstrated for greatly speeding up the simulation of buried defects inside of EUVL multilayer mask blanks through exploiting the problem specific nature of EUV multilayer materials. The optical homogeneity and optical smoothness of EUV buried defects, along with the low numerical aperture of the EUV system, allow the development of fast and accurate simulations. The buried defect problem is broken into three separate components: a push inward to deliver the electromagnetic energy, a reflection calculation, and a push outward to propagate the reflections outside of the multilayer. This new method has been shown to give results as accurate as FDTD simulations, while maintaining the speed of the SSA method when used to simulate buried defects coated with the smoothing process developed at LLNL. The newly proposed method can be viewed as a generalization of the SSA method to incorporate the entire multilayer stack by using a dual mirror structure to approximate the resonance conditions. The new method has been shown to be about 400X faster and use about 20X less memory than FDTD simulations.

In this paper, we propose a flow-level simulator called FSIM (Fluid-based SIMulator) for performance evaluation of large-scale networks, and verify its effectiveness using our FSIM implementation. The notable features of our flow-level simulator FSIM are its accuracy and fast simulation execution compared with conventional flow-level simulators. For improving simulation accuracy, our flow-level simulator FSIM utilizes accurate fluid-flow models. For accelerating simulation execution speed, our flow-level simulator FSIM adopts an adaptive numerical computation algorithm for ordinary differential equations. Another notable feature of our flow-level simulator FSIM is its compatibility with the existing network performance analysis tool. In this paper, through several experiments using our FSIM implementation, we evaluate the effectiveness of our flow-level simulator FSIM in terms of simulation speed, accuracy and memory consumption. Consequently, we show that our flow-level simulator FSIM outperforms a conventional flow-level simulator; i.e., it realizes approximately 100% faster simulation with higher accuracy and less memory consumption than a conventional flow-level simulator.

This paper introduces MOCHY, a tool designed for the modeling of concurrent systems with variants of stochastic, timed and hybrid Petri nets. Beyond modeling, the tool serves as a platform for fast simulation, and can be used for statistical verification of properties, controller testing, and learning of control rules. The targeted models are variants of stochastic and timed nets where tokens can be continuous quantities depicting trajectories of moving objects. The architecture of the tool is designed to be as adaptive as possible, and allow the redefinition of objects behaviors or transitions firing through the refinement of a few semantic rules. The framework also allows for the integration of controllers. For any model variant, MOCHY can perform fast simulation, and perform statistical verification, evaluate some quantitative properties of a model, or learn control rules for reachability or quantitative objectives.