Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel computers.”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

Performing optimized collective operations in a irregular subcommunicator of compute nodes in a parallel computer

In a parallel computer, performing optimized collective operations in an irregular subcommunicator of compute nodes may be carried out by: identifying, within the irregular subcommunicator, regular neighborhoods of compute nodes; selecting, for each neighborhood from the compute nodes of the neighborhood, a local root node; assigning each local root node to a node of a neighborhood-wide tree topology; mapping, for each neighborhood, the compute nodes of the neighborhood to a local tree topology having, at its root, the local root node of the neighborhood; and performing a one way, rooted collective operation within the subcommunicator including: performing, in one phase, the collective operation within each neighborhood; and performing, in another phase, the collective operation amongst the local root nodes.

Davis, Kristan Suzanne D.↗

Increasing processor utilization during parallel computation rundown

Some parallel processing environments provide for asynchronous execution and completion of general purpose parallel computations from a single computational phase. When all the computations from such a phase are complete, a new parallel computational phase is begun. Depending upon the granularity of the parallel computations to be performed, there may be a shortage of available work as a particular computational phase draws to a close (computational rundown). This can result in the waste of computing resources and the delay of the overall problem. In many practical instances, strict sequential ordering of phases of parallel computation is not totally required. In such cases, the beginning of one phase can be correctly computed before the end of a previous phase is completed. This allows additional work to be generated somewhat earlier to keep computing resources busy during each computational rundown. The conditions under which this can occur are identified and the frequency of occurrence of such overlapping in an actual parallel Navier-Stokes code is reported. A language construct is suggested and possible control strategies for the management of such computational phase overlapping are discussed.

Jones, W. H.↗

Increasing processor utilization during parallel computation rundown

Some parallel processing environments provide for asynchronous execution and completion of general purpose parallel computations from a single computational phase. When all the computations from such a phase are complete, a new parallel computational phase is begun. Depending upon the granularity of the parallel computations to be performed, there may be a shortage of available work as a particular computational phase draws to a close (computational rundown). This can result in the waste of computing resources and the delay of the overall problem. In many practical instances, strict sequential ordering of phases of parallel computation is not totally required. In such cases, the beginning of one phase can be correctly computed before the end of a previous phase is completed. This allows additional work to be generated somewhat earlier to keep computing resources busy during each computational rundown. The conditions under which this can occur are identified and the frequency of occurrence of such overlapping in an actual parallel Navier-Stokes code is reported. A language construct is suggested and possible control strategies for the management of such computational phase overlapping are discsused.

Jones, William H.↗

Sassena — X-ray and neutron scattering calculated from molecular dynamics trajectories using massively parallel computers

Massively parallel computers now permit the molecular dynamics (MD) simulation of multi-million atom systems on time scales up to the microsecond. However, the subsequent analysis of the resulting simulation trajectories has now become a high performance computing problem in itself. Here, we present software for calculating X-ray and neutron scattering intensities from MD simulation data that scales well on massively parallel supercomputers. The calculation and data staging schemes used maximize the degree of parallelism and minimize the IO bandwidth requirements. The strong scaling tested on the Jaguar Petaflop Cray XT5 at Oak Ridge National Laboratory exhibits virtually linear scaling up to 7000 cores for most benchmark systems. Since both MPI and thread parallelism is supported, the software is flexible enough to cover scaling demands for different types of scattering calculations. The result is a high performance tool capable of unifying large-scale supercomputing and a wide variety of neutron/synchrotron technology.

Computer Science↗

Experiences with the Lanczos method on a parallel computer

A parallel computer implementation of the Lanczos method for the free-vibration analysis of structures is considered, and results for two example problems show substantial time-reduction over the sequential solutions. The major Lanczos calculation tasks are subdivided into subtasks, and parallelism is introduced at the subtask level. A speedup of 7.8 on eight processors was obtained for the decomposition step of the problem involving a 60-m three-longeron space mast, and a speedup of 14.6 on 16 processors was obtained for the decomposition step of the problem involving a blade-stiffened graphite-epoxy panel.

Bostic, Susan W.↗

Performing a local reduction operation on a parallel computer

A parallel computer including compute nodes, each including two reduction processing cores, a network write processing core, and a network read processing core, each processing core assigned an input buffer. Copying, in interleaved chunks by the reduction processing cores, contents of the reduction processing cores' input buffers to an interleaved buffer in shared memory; copying, by one of the reduction processing cores, contents of the network write processing core's input buffer to shared memory; copying, by another of the reduction processing cores, contents of the network read processing core's input buffer to shared memory; and locally reducing in parallel by the reduction processing cores: the contents of the reduction processing core's input buffer; every other interleaved chunk of the interleaved buffer; the copied contents of the network write processing core's input buffer; and the copied contents of the network read processing core's input buffer.

97 MATHEMATICS AND COMPUTING↗

Performing a local reduction operation on a parallel computer

A parallel computer including compute nodes, each including two reduction processing cores, a network write processing core, and a network read processing core, each processing core assigned an input buffer. Copying, in interleaved chunks by the reduction processing cores, contents of the reduction processing cores' input buffers to an interleaved buffer in shared memory; copying, by one of the reduction processing cores, contents of the network write processing core's input buffer to shared memory; copying, by another of the reduction processing cores, contents of the network read processing core's input buffer to shared memory; and locally reducing in parallel by the reduction processing cores: the contents of the reduction processing core's input buffer; every other interleaved chunk of the interleaved buffer; the copied contents of the network write processing core's input buffer; and the copied contents of the network read processing core's input buffer.

97 MATHEMATICS AND COMPUTING↗

Highly parallel computation

Highly parallel computing architectures are the only means to achieve the computation rates demanded by advanced scientific problems. A decade of research has demonstrated the feasibility of such machines and current research focuses on which architectures designated as multiple instruction multiple datastream (MIMD) and single instruction multiple datastream (SIMD) have produced the best results to date; neither shows a decisive advantage for most near-homogeneous scientific problems. For scientific problems with many dissimilar parts, more speculative architectures such as neural networks or data flow may be needed.

Denning, Peter J.↗

Optimized data communications in a parallel computer

A parallel computer includes nodes that include a network adapter that couples the node in a point-to-point network and supports communications in opposite directions of each dimension. Optimized communications include: receiving, by a network adapter of a receiving compute node, a packet--from a source direction--that specifies a destination node and deposit hints. Each hint is associated with a direction within which the packet is to be deposited. If a hint indicates the packet to be deposited in the opposite direction: the adapter delivers the packet to an application on the receiving node; forwards the packet to a next node in the opposite direction if the receiving node is not the destination; and forwards the packet to a node in a direction of a subsequent dimension if the hints indicate that the packet is to be deposited in the direction of the subsequent dimension.

97 MATHEMATICS AND COMPUTING↗

Optimized data communications in a parallel computer

A parallel computer includes nodes that include a network adapter that couples the node in a point-to-point network and supports communications in opposite directions of each dimension. Optimized communications include: receiving, by a network adapter of a receiving compute node, a packet--from a source direction--that specifies a destination node and deposit hints. Each hint is associated with a direction within which the packet is to be deposited. If a hint indicates the packet to be deposited in the opposite direction: the adapter delivers the packet to an application on the receiving node; forwards the packet to a next node in the opposite direction if the receiving node is not the destination; and forwards the packet to a node in a direction of a subsequent dimension if the hints indicate that the packet is to be deposited in the direction of the subsequent dimension.

97 MATHEMATICS AND COMPUTING↗

DMA shared byte counters in a parallel computer

A parallel computer system is constructed as a network of interconnected compute nodes. Each of the compute nodes includes at least one processor, a memory and a DMA engine. The DMA engine includes a processor interface for interfacing with the at least one processor, DMA logic, a memory interface for interfacing with the memory, a DMA network interface for interfacing with the network, injection and reception byte counters, injection and reception FIFO metadata, and status registers and control registers. The injection FIFOs maintain memory locations of the injection FIFO metadata memory locations including its current head and tail, and the reception FIFOs maintain the reception FIFO metadata memory locations including its current head and tail. The injection byte counters and reception byte counters may be shared between messages.

97 MATHEMATICS AND COMPUTING↗

Configuring compute nodes of a parallel computer in an operational group into a plurality of independent non-overlapping collective networks

Methods, apparatus, and products are disclosed for configuring compute nodes of a parallel computer in an operational group into a plurality of independent non-overlapping collective networks, the compute nodes in the operational group connected together for data communications through a global combining network, that include: partitioning the compute nodes in the operational group into a plurality of non-overlapping subgroups; designating one compute node from each of the non-overlapping subgroups as a master node; and assigning, to the compute nodes in each of the non-overlapping subgroups, class routing instructions that organize the compute nodes in that non-overlapping subgroup as a collective network such that the master node is a physical root.

Archer, Charles J↗

Chaining direct memory access data transfer operations for compute nodes in a parallel computer

Methods, systems, and products are disclosed for chaining DMA data transfer operations for compute nodes in a parallel computer that include: receiving, by an origin DMA engine on an origin node in an origin injection FIFO buffer for the origin DMA engine, a RGET data descriptor specifying a DMA transfer operation data descriptor on the origin node and a second RGET data descriptor on the origin node, the second RGET data descriptor specifying a target RGET data descriptor on the target node, the target RGET data descriptor specifying an additional DMA transfer operation data descriptor on the origin node; creating, by the origin DMA engine, an RGET packet in dependence upon the RGET data descriptor, the RGET packet containing the DMA transfer operation data descriptor and the second RGET data descriptor; and transferring, by the origin DMA engine to a target DMA engine on the target node, the RGET packet.

Archer, Charles J↗

Pacing a data transfer operation between compute nodes on a parallel computer

Methods, systems, and products are disclosed for pacing a data transfer between compute nodes on a parallel computer that include: transferring, by an origin compute node, a chunk of an application message to a target compute node; sending, by the origin compute node, a pacing request to a target direct memory access (`DMA`) engine on the target compute node using a remote get DMA operation; determining, by the origin compute node, whether a pacing response to the pacing request has been received from the target DMA engine; and transferring, by the origin compute node, a next chunk of the application message if the pacing response to the pacing request has been received from the target DMA engine.

97 MATHEMATICS AND COMPUTING↗

Scheduling applications for execution on a plurality of compute nodes of a parallel computer to manage temperature of the nodes during execution

Methods, apparatus, and products are disclosed for scheduling applications for execution on a plurality of compute nodes of a parallel computer to manage temperature of the plurality of compute nodes during execution that include: identifying one or more applications for execution on the plurality of compute nodes; creating a plurality of physically discontiguous node partitions in dependence upon temperature characteristics for the compute nodes and a physical topology for the compute nodes, each discontiguous node partition specifying a collection of physically adjacent compute nodes; and assigning, for each application, that application to one or more of the discontiguous node partitions for execution on the compute nodes specified by the assigned discontiguous node partitions.

97 MATHEMATICS AND COMPUTING↗

Performing an allreduce operation on a plurality of compute nodes of a parallel computer

Methods, apparatus, and products are disclosed for performing an allreduce operation on a plurality of compute nodes of a parallel computer, each node including at least two processing cores, that include: performing, for each node, a local reduction operation using allreduce contribution data for the cores of that node, yielding, for each node, a local reduction result for one or more representative cores for that node; establishing one or more logical rings among the nodes, each logical ring including only one of the representative cores from each node; performing, for each logical ring, a global allreduce operation using the local reduction result for the representative cores included in that logical ring, yielding a global allreduce result for each representative core included in that logical ring; and performing, for each node, a local broadcast operation using the global allreduce results for each representative core on that node.

97 MATHEMATICS AND COMPUTING↗

Performing an allreduce operation on a plurality of compute nodes of a parallel computer

Methods, apparatus, and products are disclosed for performing an allreduce operation on a plurality of compute nodes of a parallel computer, each node including at least two processing cores, that include: establishing, for each node, a plurality of logical rings, each ring including a different set of at least one core on that node, each ring including the cores on at least two of the nodes; iteratively for each node: assigning each core of that node to one of the rings established for that node to which the core has not previously been assigned, and performing, for each ring for that node, a global allreduce operation using contribution data for the cores assigned to that ring or any global allreduce results from previous global allreduce operations, yielding current global allreduce results for each core; and performing, for each node, a local allreduce operation using the global allreduce results.

97 MATHEMATICS AND COMPUTING↗

Dynamic Load Balancing for Computational Plasticity on Parallel Computers

The simulation of the computational plasticity on a complex structure remains a formidable computational task, especially when a highly nonlinear, complex material model was used. It appears that the computational requirements for a such problem can only be satisfied by massively parallel architectures. In order to effectively harness the tremendous computational power provided by such architectures, it is imperative to investigate and to study the algorithmic and implementation issues pertaining to dynamic load balancing for computational plasticity on a highly parallel, distributed-memory, multiple-instruction, multiple-data computers. This paper will measure the effectiveness of the algorithms developed in handling the dynamic load balancing.

Pramono, Eddy↗