Explore chapters and articles related to this topic
Parallel Architectures
Published in Pranabananda Chakraborty, Computer Organisation and Architecture, 2020
The UMA model of multiprocessors can be again divided into two categories: Symmetric and Asymmetric. When all the processors in the system share equal access to m shared memory modules as well as to all shared I/O devices through the same channels or through different channels that provide paths to the same devices, the multiprocessor is called a symmetric multiprocessor (SMP). This is illustrated in Figure 10.16a and b using interconnection network in the form of a common (hierarchical) bus and crossbar network, respectively. However, other forms (types) of interconnection network can also be employed in place of crossbar network. In this category, all the processors are allowed to run all sorts of interrupt service routines and other supervisor-related (kernel) programs. In the asymmetric category, not all but only one or a selective number of processors in the multiprocessor system are permitted to additionally handle all I/O and supervisor-related (kernel) activities. Those are treated as master processor(s) that supervises the execution activities of the other remaining processors, known as attached processors. However, all the processors here also have uniform access to any of m shared memory modules as usual.
S
Published in Phillip A. Laplante, Dictionary of Computer Science, Engineering, and Technology, 2017
symmetric multiprocessor (SMP) a class of multiprocessor consisting of two or more CPUs configured together via a common memory interface bus and with each process maintaining its own cache, but configured and attached to the same RAM. The operating system operates across all processors within the symmetric multiprocessor and, hence, from the user’s perspective, is transparent as to which processor is supporting use functions. The term symmetric is used to refer to the fact that all processors within the multiprocessor have “equal rank”; that is, no one processor is responsible for maintaining cache coherency or for communication with external interface devices. Rather, the operating system itself holds the functionality — and the responsibility — for maintaining consistency among processor cache segments and for evenly distributing workloads across all processors (level-loading).
Distributed Systems
Published in Vivek Kale, Agile Network Businesses, 2017
Like shared-memory systems, distributed memory systems (Figure 4.1c) vary widely but share a common characteristic: they require a communication network to connect the interprocessor memory. In this type of architecture, processors have access to their own memory and do not share memory with another SMP. If a processor needs information from another data store, it has to communicate how and when it is to be accessed. This sharing generally comes about through a simple Ethernet connection. Hybrids of this technology have also been developed and are referred to as distributed shared-memory architecture. However, this type of architecture is generally used in supercomputers.Within parallel computing, there are devices that provide a niche market because of its capabilities. As previously discussed, a system may have multiple cores, but may continue to be limited because of the amount of memory available to those cores. General-purpose computing on graphics processing units (GPGPU) is an inexpensive means of adding additional processing ability to a computer. A very beneficial and generous amount of memory is attached to video cards that are only accessible to its GPU. GPGPU addresses the multiple cores and memory limitations as an add-on card to a single computer system. As long as a motherboard has an available PCI-Express x16 slot, it will be capable of hosting a very powerful GPGPU.
Recent Features and Industrial Applications of the Hybrid SPH-FE Method
Published in International Journal of Computational Fluid Dynamics, 2021
Paul Groenenboom, Bruce Cartwright, Damian McGuckin
VPS is a general-purpose software tool for structural mechanics and acoustics. The component utilised for the work discussed below is the explicit FE solver that is optimised for dynamic, strongly non-linear, structural mechanics. This is performed within a Lagrangian frame of reference. VPS contains FE formulations for thin shells, solid elements, membranes and beams employing material models for metals (allowing for plasticity and failure), plastics, rubbers, foams and composites. The software incorporates robust contact algorithms to enable various parts within a model to interact. Optimal numerical performance is achieved using an appropriate mix of Shared Memory Parallel (SMP) and Distributed Memory Parallel (DMP) programming techniques within the software. The VPS software is currently employed within a wide range of industries and academia.
Quasirelativistic two-component core excitations and polarisabilities from a damped-response formulation of the Bethe–Salpeter equation
Published in Molecular Physics, 2020
Max Kehry, Yannick J. Franzke, Christof Holzer, Wim Klopper
Our implementation is available not only at the GW-BSE level but also at the time-dependent HF (TD-HF) and TD-DFT levels including the local spin-density approximation (LSDA), the generalised gradient approximation (GGA), meta-GGA (mGGA) functionals, and (range-separated) global and local hybrid functionals (LHFs) [53], together with RI and seminumerical exchange [54] methods by appropriate modifications of the matrices and [5]. For LHFs, the construction of the appropriate non-collinear Kramers-restricted kernel is described in Appendix 1, which also enables standard linear-response calculations for 2c LHFs in Kramers-restricted closed-shell systems. An interface to LIBXC [55] is provided for broad functional support. All time-consuming steps employ the OpenMP standard [56–58] to allow for a shared memory parallel (SMP) version. The solver is used to calculate two-component static polarisabilites in a straightforward manner by neglecting the frequency-dependent terms, that is, by setting .
A self-adaptive energy absorber for improved pedestrian safety and low-speed damage requirements
Published in International Journal of Crashworthiness, 2020
James Wu, William Altenhof, Sudip Bhattacharjee, Srinivasan Sundararajan, John Magliaro
Several boundary conditions were applied to the impactors and base support depending upon the simulation environment. The impactors were constrained to only allow for translation in the z-axis as shown in Figure 10. The base support was fully constrained against translation and rotation. Impactor displacement was defined with a boundary-prescribed motion like the physical testing apparatus. The model was simulated with explicit analysis and time scaling, and the velocity of the impactors was greater in the simulation than in the physical test by a factor of 100, which was approximately 8.33 mm/s. The analysis was performed with LS-DYNA, shared memory processing (SMP) and double precision solver R8.0.0. Data from the simulation results including energy terms and contact forces were output at 50 kHz for the legform. Since the pendulum tests were simulated with a test rate 10 times lower than the complementary legform tests, the data from the pendulum tests were collected at a proportionate rate of 5 kHz. By carefully observing and comparing the load/displacement responses, it was determined that the data collected provide a reasonably accurate representation of the expected test results.