International Business Machines Corporation

Globalfoundries Inc.

Robert Matthew Senger

Dong Chen

Philip Heidelberger

Yutaka Sugawara

Burkhard Steinmacher-Burow

Noel A Eisley

Paul William Coteus

Valentina Salapura

Alan G Gara

Todd Edward Takken

Martin Ohmacht

Alan J Drake

Sameer Kumar

Rudolf Adriaan Haring

Thomas M Gooding

Gerard Vincent Kopcsay

Michael K Gschwind

Alper Buyuktosunoglu

Pradip Bose

Michael A Blocksome

Hubertus Franke

James Allan Kahle

Joseph D Ratterman

Adam James Muff

Michael Stephen Floyd

Gary Dale Carpenter

Bruce Martin Fleischer

Alexandre E Eichenberger

Robert William Wisniewski

John A Gunnels

Mark Edwin Giampapa

Kathryn Mary O'Brien

Mark Gregory Megerian

Matthias Augustin Blumrich

Sameh W Asaad

Shawn Anthony Hall

Thomas Winters Fox

John Kevin O'Brien

Fadi Hikmat Gebara

Chen-Yong Cher

Barry Marshall Trager

Peng Wu

Charles David Wait

Jeremy Daniel Schaub

John-David Wellman

James Anthony Marcella

Todd Alan Inglett

George L Chiu

Krishnan Sugavanam

Augusto Javier Vega

Jeffrey John Parker

David L Satterfield

Douglas R Miller

Craig Brian Stunkel

Amith R Mamidala

Kristan D Davis

Jose R Brunheroto

Fabrizio Petrini

Michael B Mundy

Matthew Roger Ellavsky

Robert E Walkup

James Lee Van Oosten

Samuel J Miller

Gabor J Dozsa

Ruth Janine Poole

John Patrick Keane

Peter Boyle

Brant L Knudson

Ralph E Bellofatto

Harmander Singh

Norman Christ

Kahn C Evans

Harmander Singh Deogun

William Maupin Stockdell

Alfred Thomas Watson, Iii

Aporva Amarnath

Brian E Smith

Kevin John Nowka

Xin Zhang

Norman Karl James

Rahul M Rao

Don Darrell Reed

Thomas A Liebsch

Ralph A Bellofatto

Chung-shiang Wu

Described are techniques for scheduling tasks on a heterogeneous system on a chip (SoC). The techniques including receiving a directed acyclic graph at a meta pre-processor associated with a heterogeneous system-on-chip and communicatively coupled to a scheduler, where the directed acyclic graph corresponds to a control flow graph of respective tasks associated with an application executed by the heterogeneous system-on-chip. The techniques further including determining, using a learning agent implementing machine learning algorithms, a rank for a respective task in the directed acyclic graph, wherein the learning agent receives as input the directed acyclic graph, constraints associated with the directed acyclic graph, and heuristics regarding previously completed tasks. The techniques further including providing the respective task to the scheduler for execution on the heterogeneous system-on-chip according to the rank.

Heterogeneous system on a chip scheduler with learning agent

Described are techniques for scheduling tasks on a heterogeneous system on a chip (SoC). The techniques including receiving a directed acyclic graph at a meta pre-processor associated with a heterogeneous SoC and communicatively coupled to a scheduler, wherein the directed acyclic graph corresponds to a control flow graph of tasks associated with an application executed by the heterogeneous SoC. The techniques further including determining a rank for a respective task in the directed acyclic graph, wherein the rank is based on a priority of the respective task and a slack in the directed acyclic graph. The techniques further including providing the respective task to the scheduler for execution on the heterogeneous SoC according to the rank.

Heterogeneous system on a chip scheduler

Methods and systems for monitoring remote transmissions of messages among a plurality of nodes are described. A processing element in a first node may allocate a sequence number to a request to read and/or update data in a second node. The processing element may be different from main processors of the first node. The processing element may send the message and the sequence number to the second node. The processing element may modify a status of the sequence number to an active state, indicating a transmission of the message is pending. The processing element may, in response to a response from the second node, modify the status of the sequence number to an inactive state, indicating a completed transmission of the message. The processing element may, in response to no response from the second node within a time period, resend the message and the sequence number to the second node.

Reliability processing of remote direct memory access

A first method includes determining a total length of pending packets for a network link, determining a currently preferred power mode for the network link based on the total length of pending packets for the network link, and changing a current power mode for the network link to the currently preferred power mode. A corresponding apparatus is also disclosed herein. A second method includes determining a utilization for a network link, determining a currently preferred power mode for the network link based on the utilization for the network link, and changing a current power mode for the network link to the currently preferred power mode. A corresponding apparatus is also disclosed herein.

Power management of network links

A shared memory maintained by sender processes stores a sequence number counter per destination process. A sender process increments the sequence number counter in the shared memory in sending a message to a destination process. The sender process sends a data packet comprising the message and at least a sequence number specified by the sequence number counter. All of the sender processes share a sequence number counter per destination process, each of the sender processes incrementing the sequence number counter in sending a respective message. Receiver processes run on the hardware processor, each of the receiver processes maintaining a local memory counter on the memory, the local memory counter associated with a sending node. The local memory counter stores a sequence number of a message received from the sending node. The receiver process delivers incoming data packets ordered by sequence numbers of the data packets.

Reduced number of counters for reliable messaging

Embodiments of the invention provide a method, system and computer program product for embedding a global barrier and global interrupt network in a parallel computer system organized as a torus network. The computer system includes a multitude of nodes. In one embodiment, the method comprises taking inputs from a set of receivers of the nodes, dividing the inputs from the receivers into a plurality of classes, combining the inputs of each of the classes to obtain a result, and sending said result to a set of senders of the nodes. Embodiments of the invention provide a method, system and computer program product for embedding a collective network in a parallel computer system organized as a torus network. In one embodiment, the method comprises adding to a torus network a central collective logic to route messages among at least a group of nodes in a tree structure.

Embedding global barrier and collective in a torus network

An apparatus is disclosed that includes a semiconductor device to control a power converter having two or more power FETs. The semiconductor device includes a gate signal generator configured to produce two or more gate signals able to drive the two or more FETs. The gate signal generator has programmable timing configuration parameters to control operation of the two or more gate signals, wherein the timing configuration parameters are digitally programmed via a memory in the semiconductor device. The memory may be implemented with fuses, other non-volatile memory or volatile memory. The parameters may be fixed or updated during a lifetime of the apparatus. A serial-to-parallel conversion may be used to input the parameters. Optimization methods may be performed to determine parameters considered to be optimal. The apparatus may also include the power converter.

Integrated circuits with programmable gate timing signal generation for power converters and apparatus comprising the same

According to one embodiment of the present invention, a system for operating memory includes a first node coupled to a second node by a network, the system configured to perform a method including receiving the remote transaction message from the second node in a processing element in the first node via the network, wherein the remote transaction message bypasses a main processor in the first node as it is transmitted to the processing element. In addition, the method includes accessing, by the processing element, data from a location in a memory in the first node based on the remote transaction message, and performing, by the processing element, computations based on the data and the remote transaction message.

Remote processing and memory utilization

A Multi-Petascale Highly Efficient Parallel Supercomputer of 100 petaflop-scale includes node architectures based upon System-On-a-Chip technology, where each processing node comprises a single Application Specific Integrated Circuit (ASIC). The ASIC nodes are interconnected by a five dimensional torus network that optimally maximize the throughput of packet communications between nodes and minimize latency. The network implements collective network and a global asynchronous network that provides global barrier and notification functions. Integrated in the node design include a list-based prefetcher. The memory system implements transaction memory, thread level speculation, and multiversioning cache that improves soft error rate at the same time and supports DMA functionality allowing for parallel processing message-passing.

Multi-petascale highly efficient parallel supercomputer

Embedding global and collective in a torus network with message class map based tree path selection

According to one embodiment of the present invention, a method of constructing network communication for a grid of node groups is provided, the grid including an M dimensional grid, each node group including N nodes, wherein M is greater than or equal to one and N is greater than one, wherein each node includes a router. The method includes directly connecting each node in each node group to every other node in the node group via intra-group links and directly connecting each node in each node group of the M dimensional grid to a node in each neighboring node group in the M dimensional grid via inter-group links.

T-Star interconnection network topology

A Multi-Petascale Highly Efficient Parallel Supercomputer of 100 petaOPS-scale computing, at decreased cost, power and footprint, and that allows for a maximum packaging density of processing nodes from an interconnect point of view. The Supercomputer exploits technological advances in VLSI that enables a computing model where many processors can be integrated into a single Application Specific Integrated Circuit (ASIC). Each ASIC computing node comprises a system-on-chip ASIC utilizing four or more processors integrated into one die, with each having full access to all system resources and enabling adaptive partitioning of the processors to functions such as compute or messaging I/O on an application by application basis, and preferably, enable adaptive partitioning of functions in accordance with various algorithmic phases within an application, or if I/O or other processors are underutilized, then can participate in computation or communication nodes are interconnected by a five dimensional torus network with DMA that optimally maximize the throughput of packet communications between nodes and minimize latency.

According to one embodiment of the present invention, a system for network communication includes an M dimensional grid of node groups, each node group including N nodes, wherein M is greater than or equal to one and N is greater than one and each node comprises a router and intra-group links directly connecting each node in each node group to every other node in the node group. In addition, the system includes inter-group links directly connecting each node in each node group to a node in each neighboring node group in the M dimensional grid.

T-star interconnection network topology

Fixing a problem is usually greatly aided if the problem is reproducible. To ensure reproducibility of a multiprocessor system, the following aspects are proposed: a deterministic system start state, a single system clock, phase alignment of clocks in the system, system-wide synchronization events, reproducible execution of system components, deterministic chip interfaces, zero-impact communication with the system, precise stop of the system and a scan of the system state.

Reproducibility in a multiprocessor system

Growing community of inventors

Tarrytown, NY, United States of America

Robert Matthew Senger