Abstract: A computation system for computing interactions in a multiple-body simulation includes an array of processing modules arranged into one or more serially interconnected processing groups of the processing modules. Each of the processing modules includes storage for data elements and includes circuitry for performing pairwise computations between data elements each associated with a spatial location. Each of the pairwise computations makes use of a data element from the storage of the processing module and a data element passing through the serially interconnected processing modules. Each of the processing modules includes circuitry for selecting the pairs of data elements according to separations between spatial locations associated with the data elements.

Abstract: A generalized approach to particle interaction can confer advantages over previously described method in terms of one or more of communications bandwidth and latency and memory access characteristics. These generalizations can involve one or more of at least spatial decomposition, import region rounding, and multiple zone communication scheduling. An architecture for computation of particle interactions makes use various forms of parallelism. In one implementation, the parallelism involves using multiple computation nodes arranged according to a geometric partitioning of a simulation volume.

Abstract: Distributed computation of multiple body interactions in a region uses multiple processing modules, where each of the processing modules is associated with a respective corresponding portion of the region. In some examples, the approach includes establishing multiple coordinate frames of reference, each processing module corresponding to one the coordinate frames of reference. In some examples, efficient techniques are used for selecting elements for computation of interactions according at least in part to a separation-based criterion.

Abstract: A computer-implemented method for determining computational units for computing interactions among sets of bodies located in a computation region includes, for each computation associated with one of the sets of bodies, determining, according to an assignment rule that provides a mapping from a location of each of the bodies to a determined computation unit from the plurality of computation units, a computation unit from a plurality of computation units for performing the computation.

Abstract: A generalized approach to particle interaction can confer advantages over previously described method in terms of one or more of communications bandwidth and latency and memory access characteristics. These generalizations can involve one or more of at least spatial decomposition, import region rounding, and multiple zone communication scheduling. An architecture for computation of particle interactions makes use various forms of parallelism. In one implementation, the parallelism involves using multiple computation nodes arranged according to a geometric partitioning of a simulation volume.

Abstract: Distributed computation of multiple body interactions in a region uses multiple processing modules, where each of the processing modules is associated with a respective corresponding portion of the region. In some examples, the approach includes establishing multiple coordinate frames of reference, each processing module corresponding to one the coordinate frames of reference. In some examples, efficient techniques are used for selecting elements for computation of interactions according at least in part to a separation-based criterion.

Abstract: An approach to introducing adaptive routing into a communication approach for passing messages between nodes over links between the nodes includes forming virtual channels over the links of the system and defining a deterministic routing function over the virtual channels such that the deterministic routing function is deadlock free. Adaptive routing is then permitted at nodes using the existing virtual channels by introducing a constraint on the available virtual channels used to forward communication that arrives at a node for a particular destination. The constraint on the virtual channels is such that the adaptive system is also deadlock free.

Abstract: A method for dynamics simulation involves maintaining quantities according to a floating point binary format quantized to a first precision lower than the precision supported by the floating point format. For example, although an IEEE floating point number can represent numbers with a precision of one part in 2^24, the quantities are quantized to a lower precision, such as one part in 2^22. Operations are applied to sets of the quantities by quantizing the intermediate results of the operations to the lower precision than the precision supported by the floating point format.

Abstract: An approach to data communication makes use of a protocol for encoding data on a serial link that provides both a run length limiting function and a frame marking function, while minimizing communication overhead over the data bearing portions of the signal, and while limiting latency introduced into the communication. In some examples, a single bit is added as a frame marker in such a way that a single bit frame marker also limits run length.

Abstract: An improved constraint approach reduces the energy drift rate to acceptable levels. In an embodiment of this approach, massively parallel constrained velocity Verlet NVE (constant particle number, constant volume, constant energy) MD simulations can be run using single precision arithmetic with very low energy drift (e.g., ˜1 Kelvin per microsecond simulated time) using large timesteps (e.g., 2.5 fs) for typical systems and MD force fields.

Abstract: In one aspect, a method of processing time-ordered multi-element data uses a set of computational nodes. In some examples, hundreds or thousands of nodes are used. A set of portions of the data are accepted, for example, from a MD simulation system. Each portion of the data is associated with a corresponding computational node in the plurality of computational nodes, and each portion representing a distinct range of time. Instructions for processing the data are accepted. These instructions include one or more instruction specifying a set of times, a set of elements, an analysis function, and an aggregation function. The accepted data is redistributed from within the portions at each computational node to multiple computational nodes in the plurality of computational nodes, such that data for any element of the specified set of elements is localized to a particular computational node.

Abstract: A method and a corresponding software and a system for processing a charge distribution according to a spreading function and a predefined operation are described.

Abstract: An approach to introducing adaptive routing into a communication approach for passing messages between nodes over links between the nodes includes forming virtual channels over the links of the system and defining a deterministic routing function over the virtual channels such that the deterministic routing function is deadlock free. Adaptive routing is then permitted at nodes using the existing virtual channels by introducing a constraint on the available virtual channels used to forward communication that arrives at a node for a particular destination. The constraint on the virtual channels is such that the adaptive system is also deadlock free.

Abstract: Distributed computation of multiple body interactions in a region uses multiple processing modules, where each of the processing modules is associated with a respective corresponding portion of the region. In some examples, the approach includes establishing multiple coordinate frames of reference, each processing module corresponding to one the coordinate frames of reference. In some examples, efficient techniques are used for selecting elements for computation of interactions according at least in part to a separation-based criterion.

Abstract: An inter-processor communication approach is applicable to a message passing pattern called iterative exchange. In such patterns, two processors exchange messages, then perform a computation, and then this process is repeated. If two sets of send and receive buffers are used, then it is possible to guarantee that a receive buffer on the receiver's side is always available to receive the message. A message passing system controls which buffers are used for sending and receiving. These buffers are registered beforehand, thereby avoiding repeated registration at the time messages are sent. The sender is initially informed of all the possible receive buffers that the receiver will use, and the sender then uses these receive buffers alternately. Examples of this approach can avoid the use of multiple-step rendezvous protocols, memory copies, and memory registrations when a message needs to be sent.

Abstract: A method for dynamics simulation involves maintaining quantities according to a floating point binary format quantized to a first precision lower than the precision supported by the floating point format. For example, although an IEEE floating point number can represent numbers with a precision of one part in 2ˆ24, the quantities are quantized to a lower precision, such as one part in 2ˆ22. Operations are applied to sets of the quantities by quantizing the intermediate results of the operations to the lower precision than the precision supported by the floating point format.

Abstract: An improved constraint approach reduces the energy drift rate to acceptable levels. In an embodiment of this approach, massively parallel constrained velocity Verlet NVE (constant particle number, constant volume, constant energy) MD simulations can be run using single precision arithmetic with very low energy drift (e.g., ˜1 Kelvin per microsecond simulated time) using large timesteps (e.g., 2.5 fs) for typical systems and MD force fields.