Abstract: Identifying a collective operation within an application executing on a parallel computer; identifying a call site of the collective operation; determining whether the collective operation is root-based; if the collective operation is not root-based: establishing a tuning session and executing the collective operation in the tuning session; if the collective operation is root-based, determining whether all compute nodes executing the application identified the collective operation at the same call site; if all compute nodes identified the collective operation at the same call site, establishing a tuning session and executing the collective operation in the tuning session; and if all compute nodes executing the application did not identify the collective operation at the same call site, executing the collective operation without establishing a tuning session.

Abstract: Data communications in a parallel active messaging interface (‘PAMI’) of a parallel computer, the PAMI composed of data communications endpoints, each endpoint including a specification of data communications parameters for a thread of execution on a compute node, including specifications of a client, a context, and a task, endpoints coupled for data communications through the PAMI and through data communications resources, including receiving in an origin endpoint of the PAMI a SEND instruction, the SEND instruction specifying a transmission of transfer data from the origin endpoint to a first target endpoint; transmitting from the origin endpoint to the first target endpoint a Request-To-Send (‘RTS’) message advising the first target endpoint of the location and size of the transfer data; assigning by the first target endpoint to each of a plurality of target endpoints separate portions of the transfer data; and receiving by the plurality of target endpoints the transfer data.

Abstract: Methods, apparatuses, and computer program products for endpoint-based parallel data processing with non-blocking collective instructions in a parallel active messaging interface (‘PAMI’) of a parallel computer are provided. Embodiments include establishing by a parallel application a data communications geometry, the geometry specifying a set of endpoints that are used in collective operations of the PAMI, including associating with the geometry a list of collective algorithms valid for use with the endpoints of the geometry. Embodiments also include registering in each endpoint in the geometry a dispatch callback function for a collective operation and executing without blocking, through a single one of the endpoints in the geometry, an instruction for the collective operation.

Abstract: Collective operation protocol selection in a parallel computer that includes compute nodes may be carried out by calling a collective operation with operating parameters; selecting a protocol for executing the operation and executing the operation with the selected protocol. Selecting a protocol includes: iteratively, until a prospective protocol meets predetermined performance criteria: providing, to a protocol performance function for the prospective protocol, the operating parameters; determining whether the prospective protocol meets predefined performance criteria by evaluating a predefined performance fit equation, calculating a measure of performance of the protocol for the operating parameters; determining that the prospective protocol meets predetermined performance criteria and selecting the protocol for executing the operation only if the calculated measure of performance is greater than a predefined minimum performance threshold.

Abstract: Endpoint-based parallel data processing with non-blocking collective instructions in a PAMI of a parallel computer is disclosed. The PAMI is composed of data communications endpoints, each including a specification of data communications parameters for a thread of execution on a compute node, including specifications of a client, a context, and a task. The compute nodes are coupled for data communications through the PAMI. The parallel application establishes a data communications geometry specifying a set of endpoints that are used in collective operations of the PAMI by associating with the geometry a list of collective algorithms valid for use with the endpoints of the geometry; registering in each endpoint in the geometry a dispatch callback function for a collective operation; and executing without blocking, through a single one of the endpoints in the geometry, an instruction for the collective operation.

Abstract: Methods, apparatuses, and computer program products for scheduling synchronization in association with collective operations in a parallel computer that includes a shared memory and a plurality of compute nodes that execute a parallel application utilizing the shared memory are provided. Embodiments include acquiring an available channel of the shared memory; posting to the acquired channel of the shared memory one or more collective operations and a synchronization point; determining that processing within the acquired channel has reached the synchronization point; and posting to the acquired channel, in response to determining that processing within the acquired channel has reached the synchronization point, a background synchronization operation corresponding to the one or more collective operations.

Abstract: Algorithm selection for collective operations in a parallel computer that includes a plurality of compute nodes may include: profiling a plurality of algorithms for each of a set of collective operations, including for each collective operation: executing the operation a plurality times with each execution varying one or more of: geometry, message size, data type, and algorithm to effect the collective operation, thereby generating performance metrics for each execution; storing the performance metrics in a performance profile; at load time of a parallel application including a plurality of parallel processes configured in a particular geometry, filtering the performance profile in dependence upon the particular geometry; during run-time of the parallel application, selecting, for at least one collective operation, an algorithm to effect the operation in dependence upon characteristics of the parallel application and the performance profile; and executing the operation using the selected algorithm.

Abstract: Eager send data communications in a parallel active messaging interface (‘PAMI’) of a parallel computer, the PAMI composed of data communications endpoints that specify a client, a context, and a task, including receiving an eager send data communications instruction with transfer data disposed in a send buffer characterized by a read/write send buffer memory address in a read/write virtual address space of the origin endpoint; determining for the send buffer a read-only send buffer memory address in a read-only virtual address space, the read-only virtual address space shared by both the origin endpoint and the target endpoint, with all frames of physical memory mapped to pages of virtual memory in the read-only virtual address space; and communicating by the origin endpoint to the target endpoint an eager send message header that includes the read-only send buffer memory address.

Abstract: Methods, apparatuses, and computer program products for collective operation management in a parallel computer are provided. Embodiments include a parallel computer having a first compute node operatively coupled for data communications over a tree data communications network with a plurality of child compute nodes. Embodiments also include each child compute node performing a first collective operation. The first compute rode, for each child compute node, receives from the child compute node, a result of the first collective operation performed by the child compute node. In response to receiving at least one result, the first compute node reduces a power consumption level of the child compute node.

Abstract: Methods, apparatuses, and computer program products for collective operation management in a parallel computer are provided. Embodiments include a parallel computer having a first compute node operatively coupled for data communications over a tree data communications network with a plurality of child compute nodes. Embodiments also include each child compute node performing a first collective operation. The first compute rode, for each child compute node, receives from the child compute node, a result of the first collective operation performed by the child compute node. For each result received from a child compute node, the first compute node stores a timestamp indicating a time that the child compute node completed the first collective operation. The first compute node also manages, based on the stored timestamps, execution of a second collective operation over the tree data communications network.

Abstract: Eager send data communications in a parallel active messaging interface (‘PAMI’) of a parallel computer, the PAMI composed of data communications endpoints that specify a client, a context, and a task, including receiving an eager send data communications instruction with transfer data disposed in a send buffer characterized by a read/write send buffer memory address in a read/write virtual address space of the origin endpoint; determining for the send buffer a read-only send buffer memory address in a read-only virtual address space, the read-only virtual address space shared by both the origin endpoint and the target endpoint, with all frames of physical memory mapped to pages of virtual memory in the read-only virtual address space; and communicating by the origin endpoint to the target endpoint an eager send message header that includes the read-only send buffer memory address.

Abstract: Background collective operation management in a parallel computer, the parallel computer including one or more compute nodes operatively coupled for data communications over one or more data communications networks, including: determining, by a management availability module, whether a compute node in the parallel computer is available to perform a background collective operation management task; responsive to determining that the compute node is available to perform the background collective operation management task, determining, by the management availability module, whether the compute node has access to sufficient resources to perform the background collective operation management task; and responsive to determining that the compute node has access to sufficient resources to perform the background collective operation management task, initiating, by the management availability module, execution of the background collective operation management task.

Abstract: Endpoint-based parallel data processing in a parallel active messaging interface (‘PAMI’) of a parallel computer, the PAMI composed of data communications endpoints, each endpoint including a specification of data communications parameters for a thread of execution on a compute node, including specifications of a client, a context, and a task, the compute nodes coupled for data communications through the PAMI, including establishing a data communications geometry, the geometry specifying, for tasks representing processes of execution of the parallel application, a set of endpoints that are used in collective operations of the PAMI including a plurality of endpoints for one of the tasks; receiving in endpoints of the geometry an instruction for a collective operation; and executing the instruction for a collective operation through the endpoints in dependence upon the geometry, including dividing data communications operations among the plurality of endpoints for one of the tasks.

Abstract: Methods, apparatus, and products are disclosed for balancing a data processing load among a plurality of compute nodes in a parallel computer that include: partitioning application data for processing on the plurality of compute nodes into data chunks; receiving, by each compute node, at least one of the data chunks for processing; estimating, by each compute node, processing time involved in processing the data chunks received by that compute node for processing; and redistributing, by at least one of the compute nodes to at least one of the other compute nodes, a portion of the data chunks received by that compute node in dependence upon the processing time estimated by that compute node.

Abstract: Token-based flow control of messages in a parallel computer, the parallel computer including a plurality of compute nodes, each compute node including one or more computer processors, including: allocating, by a token administration module to a plurality of the computer processors in the parallel computer, a number of data communications tokens; identifying all communicators executing on each computer processor, where each communicator is participating in a distinct parallel operation executing on the parallel computer; allocating, to the communicators, the data communications tokens; determining, by a communicator attempting to send data to the destination, whether the communicator has enough available data communications tokens to send the data to the destination; and responsive to determining that the communicator has enough available data communications tokens to send the data, sending, by the communicator, the data to the destination.

Abstract: Methods, apparatus, and products are disclosed for performing an all-to-all exchange on n number of data buffers using XOR swap operations. Each data buffer has n number of data elements. Performing an all-to-all exchange on n number of data buffers using XOR swap operations includes for each rank value of i and j where i is greater than j and where i is less than or equal to n: selecting data element i in data buffer j; selecting data element j in data buffer i; and exchanging contents of data element i in data buffer j with contents of data element j in data buffer i using an XOR swap operation.

Abstract: Send-side matching of data communications messages includes a plurality of compute nodes organized for collective operations, including: issuing by a receiving node to source nodes a receive message that specifies receipt of a single message to be sent from any source node, the receive message including message matching information, a specification of a hardware-level mutual exclusion device, and an identification of a receive buffer; matching by two or more of the source nodes the receive message with pending send messages in the two or more source nodes; operating by one of the source nodes having a matching send message the mutual exclusion device, excluding messages from other source nodes with matching send messages and identifying to the receiving node the source node operating the mutual exclusion device; and sending to the receiving node from the source node operating the mutual exclusion device a matched pending message.

Abstract: Embodiments of the invention provide a method for querying performance counter data on a massively parallel computing system, while minimizing the costs associated with interrupting computer processors and limited memory resources. DMA descriptors may be inserted into an injection FIFO of a remote compute node in the massively parallel computing system. Upon executing the DMA operations described by the DMA descriptors, performance counter data may be transferred from the remote compute node to a destination node.

Abstract: Methods, parallel computers, and computer program products for analyzing update conditions for shared variable directory (SVD) information in a parallel computer are provided. Embodiments include a runtime optimizer receiving a compare-and-swap operation header. The compare-and-swap operation header includes an SVD key, a first SVD address, and an updated first SVD address. The first SVD address is associated with the SVD key in a first SVD associated with a first task. Embodiments also include the runtime optimizer retrieving from a remote address cache associated with the second task, a second SVD address indicating a location within a memory partition associated with the first SVD in response to receiving the compare-and-swap operation header. Embodiments also include the runtime optimizer determining whether the second SVD address matches the first SVD address and transmitting a result indicating whether the second SVD address matches the first SVD address.

Abstract: Methods, parallel computers, and computer program products for acquiring remote shared variable directory (SVD) information in a parallel computer are provided. Embodiments include a runtime optimizer determining that a first thread of a first task requires shared resource data stored in a memory partition corresponding to a second thread of a second task. Embodiments also include the runtime optimizer requesting from the second thread, in response to determining that the first thread of the first task requires the shared resource data, SVD information associated with the shared resource data. Embodiments also include the runtime optimizer receiving from the second thread, the SVD information associated with the shared resource data.