Faraj, Daniel A. | Engineering Papers

DOE OSTI2021

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.↗

DOE OSTI2019

Performing optimized collective operations in an 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 D.↗

DOE OSTI2018

Providing full point-to-point communications among compute nodes of an operational group in a global combining network of a parallel computer

Methods, apparatus, and products are disclosed for providing full point-to-point communications among compute nodes of an operational group in a global combining network of a parallel computer, each compute node connected to each adjacent compute node in the global combining network through a link, that include: receiving a network packet in a compute node, the network packet specifying a destination compute node; selecting, in dependence upon the destination compute node, at least one of the links for the compute node along which to forward the network packet toward the destination compute node; and forwarding the network packet along the selected link to the adjacent compute node connected to the compute node through the selected link.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Performance regression manager for large scale systems

System and computer program product to perform an operation comprising generating, based on a first output generated by a first execution instance of a command, a first output file specifying a value of at least one performance metric, wherein the first output file is formatted according to a predefined format, comparing the value of the at least one performance metric in the first output file to a value of the performance metric in a second output file, the second output file having been generated based on a second output generated by a second execution instance of the command, and outputting for display an indication of a result of the comparison of the value of the at least one performance metric of the first output file to the value of the at least one performance metric of the second output file.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Performance regression manager for large scale systems

Methods comprising generating, based on a first output generated by a first execution instance of a command, a first output file specifying a value of at least one performance metric, wherein the first output file is formatted according to a predefined format, comparing the value of the at least one performance metric in the first output file to a value of the performance metric in a second output file, the second output file having been generated based on a second output generated by a second execution instance of the command, and outputting for display an indication of a result of the comparison of the value of the at least one performance metric of the first output file to the value of the at least one performance metric of the second output file.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Performance regression manager for large scale systems

System and computer program product to perform an operation comprising generating, based on a first output generated by a first execution instance of a command, a first output file specifying a value of at least one performance metric, wherein the first output file is formatted according to a predefined format, comparing the value of the at least one performance metric in the first output file to a value of the performance metric in a second output file, the second output file having been generated based on a second output generated by a second execution instance of the command, and outputting for display an indication of a result of the comparison of the value of the at least one performance metric of the first output file to the value of the at least one performance metric of the second output file.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2016

Identifying a largest logical plane from a plurality of logical planes formed of compute nodes of a subcommunicator in a parallel computer

In a parallel computer, a largest logical plane from a plurality of logical planes formed of compute nodes of a subcommunicator may be identified by: identifying, by each compute node of the subcommunicator, all logical planes that include the compute node; calculating, by each compute node for each identified logical plane that includes the compute node, an area of the identified logical plane; initiating, by a root node of the subcommunicator, a gather operation; receiving, by the root node from each compute node of the subcommunicator, each node's calculated areas as contribution data to the gather operation; and identifying, by the root node in dependence upon the received calculated areas, a logical plane of the subcommunicator having the greatest area.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2016

Identifying logical planes formed of compute nodes of a subcommunicator in a parallel computer

In a parallel computer, a plurality of logical planes formed of compute nodes of a subcommunicator may be identified by: for each compute node of the subcommunicator and for a number of dimensions beginning with a first dimension: establishing, by a plane building node, in a positive direction of the first dimension, all logical planes that include the plane building node and compute nodes of the subcommunicator in a positive direction of a second dimension, where the second dimension is orthogonal to the first dimension; and establishing, by the plane building node, in a negative direction of the first dimension, all logical planes that include the plane building node and compute nodes of the subcommunicator in the positive direction of the second dimension.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2015

Determining collective barrier operation skew in a parallel computer

Determining collective barrier operation skew in a parallel computer that includes a number of compute nodes organized into an operational group includes: for each of the nodes until each node has been selected as a delayed node: selecting one of the nodes as a delayed node; entering, by each node other than the delayed node, a collective barrier operation; entering, after a delay by the delayed node, the collective barrier operation; receiving an exit signal from a root of the collective barrier operation; and measuring, for the delayed node, a barrier completion time. The barrier operation skew is calculated by: identifying, from the compute nodes' barrier completion times, a maximum barrier completion time and a minimum barrier completion time and calculating the barrier operation skew as the difference of the maximum and the minimum barrier completion time.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2015

Determining collective barrier operation skew in a parallel computer

Determining collective barrier operation skew in a parallel computer that includes a number of compute nodes organized into an operational group includes: for each of the nodes until each node has been selected as a delayed node: selecting one of the nodes as a delayed node; entering, by each node other than the delayed node, a collective barrier operation; entering, after a delay by the delayed node, the collective barrier operation; receiving an exit signal from a root of the collective barrier operation; and measuring, for the delayed node, a barrier completion time. The barrier operation skew is calculated by: identifying, from the compute nodes' barrier completion times, a maximum barrier completion time and a minimum barrier completion time and calculating the barrier operation skew as the difference of the maximum and the minimum barrier completion time.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2015

Data communications for a collective operation in a parallel active messaging interface of a parallel computer

Algorithm selection for data communications in a parallel active messaging interface (`PAMI`) of a parallel computer, the PAMI composed of data communications endpoints, each endpoint including specifications of a client, a context, and a task, endpoints coupled for data communications through the PAMI, including associating in the PAMI data communications algorithms and bit masks; receiving in an origin endpoint of the PAMI a collective instruction, the instruction specifying transmission of a data communications message from the origin endpoint to a target endpoint; constructing a bit mask for the received collective instruction; selecting, from among the associated algorithms and bit masks, a data communications algorithm in dependence upon the constructed bit mask; and executing the collective instruction, transmitting, according to the selected data communications algorithm from the origin endpoint to the target endpoint, the data communications message.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2014

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↗

DOE OSTI2014

Data communications in a parallel active messaging interface of a parallel computer

Algorithm selection for data communications in a parallel active messaging interface (`PAMI`) of a parallel computer, the PAMI composed of data communications endpoints, each endpoint including specifications of a client, a context, and a task, endpoints coupled for data communications through the PAMI, including associating in the PAMI data communications algorithms and ranges of message sizes so that each algorithm is associated with a separate range of message sizes; receiving in an origin endpoint of the PAMI a data communications instruction, the instruction specifying transmission of a data communications message from the origin endpoint to a target endpoint, the data communications message characterized by a message size; selecting, from among the associated algorithms and ranges, a data communications algorithm in dependence upon the message size; and transmitting, according to the selected data communications algorithm from the origin endpoint to the target endpoint, the data communications message.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2014

Constructing a logical, regular axis topology from an irregular topology

Constructing a logical regular topology from an irregular topology including, for each axial dimension and recursively, for each compute node in a subcommunicator until returning to a first node: adding to a logical line of the axial dimension a neighbor specified in a nearest neighbor list; calling the added compute node; determining, by the called node, whether any neighbor in the node's nearest neighbor list is available to add to the logical line; if a neighbor in the called compute node's nearest neighbor list is available to add to the logical line, adding, by the called compute node to the logical line, any neighbor in the called compute node's nearest neighbor list for the axial dimension not already added to the logical line; and, if no neighbor in the called compute node's nearest neighbor list is available to add to the logical line, returning to the calling compute node.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2014

Constructing a logical, regular axis topology from an irregular topology

Constructing a logical regular topology from an irregular topology including, for each axial dimension and recursively, for each compute node in a subcommunicator until returning to a first node: adding to a logical line of the axial dimension a neighbor specified in a nearest neighbor list; calling the added compute node; determining, by the called node, whether any neighbor in the node's nearest neighbor list is available to add to the logical line; if a neighbor in the called compute node's nearest neighbor list is available to add to the logical line, adding, by the called compute node to the logical line, any neighbor in the called compute node's nearest neighbor list for the axial dimension not already added to the logical line; and, if no neighbor in the called compute node's nearest neighbor list is available to add to the logical line, returning to the calling compute node.

97 MATHEMATICS AND COMPUTING↗