Abstract: An online file transfer tasks scheduling method for processing multiple file path access requests in a network, the method including: sending at least one file path access request to a corresponding one of at least one file access service module, executing a processing procedure in each of the at least one file access service module to generate a task package according to each received file path access request, and using one or more task execution units to process one or more of the task packages; and using a task execution unit balance module to periodically evaluate a load ratio between a plurality of the file access service modules, and determining the number of the task execution units for each file access service module according to the load ratio.

Abstract: An electronic device includes a plurality of cores, and a clock generator configured to provide a plurality of clock signals to the plurality of cores, respectively, wherein the plurality of cores includes a system core that controls the clock generator to generate the clock signals having frequencies of the respective cores, wherein the frequencies are optimized and determined based on a type of an event of the electronic device, and wherein to clock signals with optimized frequencies are applied to the respective cores in order to perform the event.

Abstract: An acquisition unit acquires a state of a server including pieces of virtual CPE, and an instructing unit instructs CPE connected to the virtual CPE to control a communication band based on the state of the server. The instructing unit instructs the CPE to limit a communication band to a predetermined value or less based on a CPU usage rate acquired by the acquisition unit exceeding a predetermined threshold value. The instructing unit instructs the CPE to limit a communication band to a predetermined value or less in based on the number of connections of the CPE acquired by the acquisition unit being less than a predetermined threshold value. The acquisition unit acquires presence or absence of CPE not connected to the virtual CPE, and the instructing unit instructs the CPE to block communication based on the presence of CPE not connected to the virtual CPE.

Abstract: A telecommunications edge cloud (TEC) element comprising a plurality of TEC hardware components comprising a computing component that includes one or more processors, a storage component that includes a memory, a networking component, with the TEC device being deployed between a client and a packet network, the computing component being configured to manage a computing resource in the computing component, a networking resource in the networking component, and a storage resource in the storage component, allocate the computing resource for a first application of the plurality of different applications, execute the first application using the computing resource, allocate the networking resource for a second application of the plurality of different applications, execute the second application using the networking resource, allocate the storage resource for a third application of the plurality of different applications, execute the third application using the storage resource.

Abstract: An interactive voice response system for interpreting sentiment of a conversation in real time is provided. The system may include a receiver. The receiver may be operable to receive a plurality of sentiment signals and a first portion of a conversation. The first portion of the conversation may include a plurality of interaction pairs. Each interaction pair may include an utterance and a system response. The system may include an utterance sentiment category determination module that may be operable to determine a sentiment category for each utterance. The system may include a sequential module. The sequential module may determine a conversation sentiment category based on the utterance sentiment categories and the order in which the utterance sentiment categories are received. The system may include an output module. The output module may output the conversation sentiment category. The conversation sentiment category may be used to determine a future system response.

Abstract: A system for reducing energy consumption of datalogger devices while maintaining high sampling rate and real time alarm function is provided. The system comprises a datalogger device, an algorithm stored in the datalogger device, and an application stored in the datalogger device that when executed implements a regular high sampling interval and extended server update interval for non-emergency situations and based on detection of an alarm, implements an emergency state. Based on the implementation, the application sends an immediate alert to the server, maintains the high sampling rate, shortens update interval in accordance with at least algorithm processing of inputs, continues updating on the shortened interval based at least on continued algorithm processing, determines, based on at least one change in alarm activity, to exit the emergency state, and returns to the extended update interval.

Abstract: The present invention relates to a method for distributing resources to different data sources based on their knowledge contribution. In addition, the invention relates to a resource distribution data system for distributing resources, a computer program product for distributing resources and a computer readable medium. It may comprise the steps of receiving values from a plurality of different data sources, blending the received values for the attributes into a dataset, assigning data lineages to the values for the attributes, receiving a query for providing a data subset, providing the data subset based on the query, determining a knowledge contribution of each of the data sources to the data subset based on the data lineages of the values and instructing a distribution of shares of resources to the different data sources based on the knowledge contribution of each of the data sources to the data subset.

Abstract: The systems and methods manage thermal states of a device through user configuration of a client application on the device. The systems and methods set thermal thresholds associated with the device. The systems and methods infer the thermal thresholds from information gathered by a client application running on the device. The systems and methods implement a stored policy associated with a violation of one of the thermal thresholds by one of the monitored thermal states.

Abstract: Implementations described and claimed herein provide systems and methods for tuning and sizing one or more storage appliances in a storage system with respect to an application load and for optimizing a storage system based on a configuration of a client network and/or a storage appliance in a storage network. In one implementation, data corresponding to an application load configured to be applied to a storage appliance in the storage system is obtained. The application load is characterized in the context of a configuration of the storage system. One or more recommendations for optimizing performance of the storage system based on the characterized application load are generated.

Abstract: An interrogation data management system operates to manage implantable cardiac device interrogation data and reports. In one example the system generates a report including a reading physician note, a summary report, and an implantable cardiac device manufacturer's report. Some embodiments include smart routing, to automatically route interrogation data reports to medical professionals based on predetermined rules. The interrogation data management system interacts with the reading physician to formally review the interrogation data reports and to generate a reading physician note. The reading physician note includes the reading physician's findings/interpretations and conclusions. Other aspects are also described in this disclosure.

Abstract: A method of determining trigger conditions for scaling a scalable unit of network function comprising identifying a primary set of metrics associated with usage of an instance of the unit of network function as a primary indicator of occurrence of a load state thereof, and determining usage points when the primary indicator indicates that the load state occurs. Deriving a secondary set of the metrics, different to the primary set, as a secondary indicator of occurrence of the load state of the instance at each of a group of one or more of the usage points when the primary indicator indicates that the load state occurs, and measured data corresponding to values of the metrics in the secondary set of metrics at each of the group of usage points. Storing a trigger condition for scaling the unit of network function based on the secondary set and the measured data.

Abstract: A method for monitoring operation of a network service for the purpose of licensing control in a virtual network environment managed by a Network Functions Virtualisation Management and Orchestration, NFV-MANO, system, is disclosed. The method comprises: defining (202) the network service in a Network Service Descriptor and deploying the network service using at least one Virtual Network Function instance identified in said Network Service Descriptor by instantiating (204) at least one Virtual Network Function Component defined by a Virtualisation Deployment Unit, wherein said Virtualisation Deployment Unit is associated with said Virtual Network Function.

Abstract: Methods, systems, and computer-readable mediums for an alternate control channel for a network protocol stack are disclosed. In some embodiments a controller device provides instructions to one or more source devices. The controller device instructs the one or more source devices to override network parameters associated with network communication performed by the source devices. The network parameters to be overridden may include transport level source ports, source network addresses, or source link level addresses. In some embodiments, a range of override values are specified. In some of these aspects, a source device may perform time division multiplexing via the multiple override values, such that data generated by a single device may appear to be transmitted by multiple devices.

Abstract: Some embodiments are directed to a monitoring system for monitoring a supercomputer architecture including a plurality of devices, including an events analysis module adapted for the reception of an event relating to probes associated with these devices and for the determination of items of information on at least one device as a function of data contained in the event; an information aggregation module adapted to determine second items of information on the basis of these items of information, and as a function of data on the topology of the architecture of the supercomputer; and an information transmission module, adapted for the transmission of the second items of information to at least one supervision data viewing tool.

Abstract: An opportunistic hypervisor determines that a guest virtual machine of a virtualization host has voluntarily released control of a physical processor. The hypervisor uses the released processor to identify and initiate a virtualization management task which has not been completed. In response to determining that at least a portion of the task has been performed, the hypervisor enters a quiescent state, releasing the physical processor to enable resumption of the guest virtual machine.

Abstract: A maintenance contract fee calculation system calculates a maintenance contract fee for an air conditioner including an indoor unit and an outdoor unit. The maintenance contract fee calculation system includes: an acquisition unit that obtains first environmental information related to an installation environment of the indoor unit and second environmental information related to an installation environment of the outdoor unit; and a calculation unit that calculates the maintenance contract fee in accordance with the first environmental information and the second environmental information.

Abstract: According to an embodiment of the disclosure, an electronic device includes a processor and a memory operationally connected to the processor and configured to store instructions that, when executed by the processor, cause the processor to configure a time period comprising multiple unit durations, check for utilization of the processor for each of the multiple unit durations of the time period, collect at least one variation of the utilization of the processor based on the utilization of the processor for each of the multiple unit durations, acquire a temporal probability density function based on the at least one collected variation, determine a probability density function corresponding to the temporal probability density function based on a previously stored probability density function table, and determine an operating frequency for a next unit duration based on at least part of the identified probability density function. Various other embodiments are possible.

Abstract: Disclosed are various embodiments for assigning compute kernels to compute accelerators that form an aggregated virtualized compute accelerator. A directed, acyclic graph (DAG) representing a workload assigned to a virtualized compute accelerator is generated. The workload can include a plurality of compute kernels and the DAG comprising a plurality of nodes and a plurality of edges, each of the nodes representing a respective compute kernel, each edge representing a dependency between a respective pair of the compute kernels, and the virtualized compute accelerator representing a logical interface for a plurality of compute accelerators. The DAG can be analyzed to identify sets of dependent compute kernels, each set of dependent compute kernels being independent of the other sets of dependent compute kernels and execution of at least one compute kernel in a set of dependent compute kernels depending on a previous execution of another computer kernel in the set of dependent compute kernels.

Abstract: Provided is a method of coordinating lock requests in a network key-value (NKV) store including one or more NKV clients and one or more NKV targets, the method including receiving, by a first NKV client, a lock request from a client device over a communication interface to lock an object key stored in one or more of the NKV targets, determining, by the first NKV client, one of the NKV targets corresponding to the lock request, transmitting, by the first NKV client, the lock request to the one of the NKV targets over a storage interface, determining, by the one of the NKV targets, a priority of the lock request, and locking, by the one of the NKV targets, the lock request based on the priority.

Abstract: Providing application-specific storage by a cloud-based storage system, including: identifying, for an application that utilizes resources within the cloud-based storage system, one or more characteristics associated with the application; and selecting, in dependence upon the one or more characteristics associated with the application and characteristics of resources within the cloud-based storage system, one or more resources within the cloud-based storage system to support the execution of the application, wherein at least a portion of a dataset associated with the application is stored as blocks within block storage resources in the cloud-based storage system and also stored as objects within object storage resources in the cloud-based storage system.

Abstract: An apparatus comprises at least one processing device configured to control delivery of input-output (IO) operations from a host device to a storage system over selected paths through a network. The at least one processing device is further configured to determine link availability information for each of a plurality of initiator-target pairs, with the initiators being implemented on the host device and the targets being implemented on the storage system, and each initiator-target pair being associated with a corresponding subset of the plurality of paths. The at least one processing device is still further configured to communicate the link availability information to the storage system, which comprises a distributed storage system that includes a plurality of interconnected storage nodes. The link availability information is utilized in the storage system to rebalance IO operation processing load across at least a subset of the storage nodes of the storage system.

Abstract: Apparatuses, methods and storage medium for computing including determination of work placement on processor cores are disclosed herein. In embodiments, an apparatus may include one or more processors, devices, and/or circuitry to identify a favored core of the processor cores. The one or more processors, devices, and/or circuitry may be configured to determine whether to migrate a thread to or from the favored core. In some embodiments, the determination may be by a process executed by a driver and/or by an algorithm executed by a power control unit of the processor.

Abstract: Disclosed is an improved system, method, and computer program product to allocate backup devices for a database system, where an appropriate appliance can be automatically allocated to accommodate backups. The allocation selects an appliance that effectively balances space allocation needs of the database with overall performance effects on a multi-tenant backup system that handles multiple databases of varying sizes.

Abstract: Methods, systems, and computer-readable mediums for an alternate control channel for a network protocol stack are disclosed. In some embodiments a controller device provides instructions to one or more source devices via a shared memory. The controller device instructs the one or more source devices to override network parameters associated with network communication performed by the source devices. The network parameters to be overridden may include transport level source ports, source network addresses, or source link level addresses. In some embodiments, a range of override values are specified. In some of these aspects, a source device may perform time division multiplexing via the multiple override values, such that data generated by a single device may appear to be transmitted by multiple devices.

Abstract: A computer-implemented optimization problem arithmetic method includes determining, based on management information indicating a partition mode that defines a logically divided state of each of a plurality of arithmetic circuits and utilization information relating to each of the plurality of arithmetic circuits, a partition mode of each of the plurality of arithmetic circuits, receiving a combinatorial optimization problem, selecting, based on information relating to scale or requested accuracy of the combinatorial optimization problem and the determined partition mode of each of the plurality of arithmetic units, a first arithmetic circuit from among the plurality of arithmetic circuits, and causing the selected first arithmetic circuit to execute arithmetic operation of the combinatorial optimization problem based on a first partition mode determined as the partition mode of the first arithmetic circuit.

Abstract: The invention relates to a method and device for determining an anomaly prediction performance index value, said method comprising: a step of receiving (120) performance indicator values, a step of identifying (140) anomalous performance indicators, a step of identifying (150) first at-risk indicators, and a step of determining (170) an anomaly prediction performance index value comprising receiving (171) new performance indicator values; identifying (172) new confirmed anomaly indicators; and comparing (174) identified at-risk indicators to the new confirmed anomalous indicators in order to generate a performance index value.

Abstract: An information processing device including a receiver that receives a plurality of pieces of control information from another device, and a controller that adds, in order of reception, a plurality of pieces of control information received from another device to a predetermined storage area, and adds, to a first buffer area related to first control information stored in the predetermined storage area, partial data of first input data corresponding to the first control information, in accordance with order of reception of the partial data.

Abstract: Process core isolation for execution of multiple operating systems on a multicore computer system includes booting first and second operating systems (OSs) on a computer system that includes a plurality of processor cores across physical central processing unit(s) (CPUs). Booting the first OS boots the first OS on a first subset of the processor cores. Booting the second OS, different from the first OS, boots the second OS on a second subset of the processor cores. The first and second subsets are mutually exclusive, where a first processor core of a physical CPU of the physical CPUs is included in the first subset and a second processor core of the physical CPU is included in the second subset, and where the first and second OSs execute concurrently on the computer system, and each of the first and second OS executes on only its respective subset of processor core(s).

Abstract: There is provided a control device, a control method, and a control system that implement a robot that flexibly executes a task in cooperation with another robot, the control device including: an ability management unit that determines capability indicating ability that can be executed by a first robot at predetermined timing as of that timing; a help management unit that compares ability required for a task to be executed by the first robot with the capability of the first robot and generates a help list indicating ability required for execution of the task; and a cooperation management unit that instructs a second robot having the capability that satisfies the ability indicated in the help list to execute the task in cooperation with the first robot.

Abstract: A method and system for allocating tasks among processing devices in a data center. The method may include receiving a request to allocate a task to one or more processing devices, the request indicating a required bandwidth for performing the task, a list of predefined processing device groups connected to a host server and indicating availability of the processing device groups included therein for allocation of tasks and available bandwidth for each available processing device group, assigning the task to a processing device group having an available bandwidth greater than or equal to the required bandwidth for performing the task, and updating the list to indicate that each of the processing device group to which the task is assigned and other processing device group sharing at least one processing device is unavailable. The task may be assigned to an available processing device group having a lowest amount of power needed.

Abstract: Technologies for utilizing a split memory pool include a compute sled. The compute sled includes multiple processors communicatively coupled together through a processor communication link. Each processor is to communicate with a different memory sled through a respective memory network dedicated to the corresponding processor and memory sled. The compute sled includes a compute engine to generate a memory access request to access a memory address in far memory. The far memory includes memory located on one of the memory sleds. The compute engine is also to determine, as a function of the memory address and a map of memory address ranges to the memory sleds, the memory sled on which to access the far memory, and send the memory access request to the determined memory sled to access the far memory associated with the memory address.

Abstract: Circuitry comprises an instruction decoder to decode a gather load instruction having a vector operand comprising a plurality of vector entries, in which each vector entry defines, at least in part, a respective address from which data is to be loaded; the instruction decoder being configured to generate a set of load operations relating to respective individual addresses in dependence upon the vector operand, each of the set of load operations having a respective identifier which is unique with respect to other load operations in the set, and control circuitry to maintain a data item for the gather load instruction, the data item including a count value representing a number of load operations in the set of load operations awaiting issue for execution; and execution circuitry to execute the set of load operations; the control circuitry being configured, in response to a detection from the count value of the data item associated with a given gather load instruction that the set of load operations generated fo

Abstract: A distributed pessimistic lock based on HBase storage and a method for implementing a database pessimistic lock; the distributed pessimistic lock including a lock manager configured to be installed on a Region of a RegionServer node of a HBase system, the lock manager has a lock and unlock interface; and the distributed pessimistic lock, an operation transaction, and a lock holder form a cross linked list format; a horizontal dimension is an information of a current data row, a vertical dimension is an information of the operation transaction, and an intersection point between the horizontal dimension and the vertical dimension is the lock holder. By installing the lock manager on a node Region of a HBase storage system, the lock manager locks and unlocks a data operation of the HBase system with the distributed pessimistic lock.

Abstract: Achieving search and ingest isolation via resource management in a search and indexing system includes receiving a search query associated with at least one data store, assigning, in response to the search query being associated with the at least one data store, the search query to a first workload pool in a set of query workload pools, and processing the search query using a first hardware resource in the first workload pool. Achieving search and ingest isolation further includes receiving an ingest request comprising data associated with the at least one data store. The ingest request is assigned to a second workload pool in a set of ingest workload pools. The set of query workload pools and the set of ingest workload pools are disjoint. Achieving search and ingest isolation further includes processing the ingest request using a second hardware resource in the second workload pool.

Abstract: A wireless network device can receive, at a traffic director in a kernel space, a data packet from a client device and determine whether the data packet is intended for an application cloud server operating in a cloud environment. The wireless network device can provide, based on determining that the data packet is intended for the application cloud server, the data packet to an application server instance executing on the wireless network device. The application server instance can be implemented in a virtualized software container in a user space, and can be configured to perform one or more operations associated with the application cloud server. The wireless network device can receive, at the traffic director and from the application server instance, a result of the application server instance performing the one or more operations on the data packet, and transmit the result to the application cloud server.

Abstract: In one example, a processing system may support capacity management in a virtualization environment based on hierarchical capacity management. The processing system may maintain a policy for a first capacity agent at a first hierarchical layer. The policy may include a set of key capacity indicators, a capacity limit, and an algorithm. The processing system may obtain, based on the set of key capacity indicators, a set of key capacity indicator information. The processing system may monitor, based on the capacity limit, for a detection of a capacity limit event. The processing system may determine, based on the algorithm, a predicted capacity exhaustion point. The processing system may send, toward a second capacity agent at a second hierarchical layer that is above the first hierarchical layer, the set of key capacity indicator information and the predicted capacity exhaustion point.

Abstract: An application manager receives or defines a service specification for a first application that defines a set of required computing resources that are necessary to run each application component of the first application. A resource supply manager in communication with the application manager manages a plurality of computing resources in a shared computing environment.

Abstract: A query is received at a database execution engine. The query is executed by the database execution engine, the executing including executing one or more threads that operate on versions of data of the delta fragment; acquiring and releasing, by a first thread, a latch on a first version of the data of the delta fragment, the releasing is performed prior to the first thread entering a wait state; and reclaiming the first version of the data of the delta fragment from memory and during execution of the query. Related systems, methods, and articles of manufacture are also described.

Abstract: A method for obtaining a storage system capacity is provided. An available capacity that is of a storage system and that is associated with each stripe length is obtained based on an obtained stripe length that can be effectively configured. Therefore, an available capacity of a system is optimally selected.

Abstract: A system and method schedule work within a workflow with defined process goals. A plurality of work queues are defined that comprise work items. The plurality of work queues are associated with one or more links between a parent work queue and at least one child work queue to form at least one work process. At least one work process goal is defined for each work process. A work schedule to achieve the work process goals is generated.

Abstract: Systems and methods utilize a data model database which includes a plurality of symbol data types. Each of the plurality of symbol data types have one or more symbol data fields. The data model database further includes a plurality of concrete data types. Each of the concrete data types have one or more language-agnostic concrete fields associated with each of the one or more symbol data fields. Each of the one or more language-agnostic concrete fields apply one or more concrete constraints to each of the corresponding symbol data fields. The data model database further includes a plurality of carrier data types. The plurality of carrier data types having one or more language-specific carrier fields associated with each of the one or more language-agnostic concrete fields. Each of the one or more language-specific carrier fields apply one or more carrier constraints to each of the corresponding language-agnostic concrete fields.

Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.

Abstract: Disclosed are aspects of memory-aware placement in systems that include graphics processing units (GPUs) that are virtual GPU (vGPU) enabled. Virtual graphics processing unit (vGPU) data is identified for graphics processing units (GPUs). A configured GPU list and an unconfigured GPU list are generated using the GPU data. The configured GPU list specifies configured vGPU profiles for configured GPUs. The unconfigured GPU list specifies a total GPU memory for unconfigured GPUs. A vGPU request is assigned to a vGPU of a GPU. The GPU is a first fit, from the configured GPU list or the unconfigured GPU list that satisfies a GPU memory requirement of the vGPU request.

Abstract: This disclosure relates to deployment of additional workload in the NFV-MANO to efficiently utilize resources during a lean workload period of Virtual Network Functions (VNFs) associated with an intelligent cloud platform. The method comprises measuring, over a time period, a current resource utilization level of one or more rendered VNFs. Thereafter, forecasting future resource utilization for the time period based on predictive analysis criteria and the current resource utilization level of the one or more VNFs. The method further comprises a machine learning based inference for determining whether the forecast future resource utilization for said time period is less than a determined optimal resource utilization threshold value of the one or more VNFs. Thereafter, resources of the one or more VNFs to one or more additional workloads are allocated based on the determination of the future resource utilization.

Abstract: An apparatus can include a control board operatively coupled to a modular compute boards and to a resource boards by (1) a first connection associated with control information and not data, and (2) a second connection associated with data and not control information. The control board can determine a computation load and a physical resource requirement for a time period. The control board can send, to the modular compute board and via the first connection, a signal indicating an allocation of that modular compute board during the time period. The control board can send, from the control board to the resource board, a signal indicating an allocation of that resource board to the modular compute board such that that resource board allocates at least a portion of its resources during the time period based on at least one of the computation load or the physical resource requirement.

Abstract: Management of communication systems is facilitated. A method comprises: processing, by a device comprising a processor, first information representative of human-readable language and indicative of terms of agreements associated with respective software for a communication system, wherein the communication system is arranged according to a first configuration of virtual machines and server devices, and wherein the agreements comprise license agreements for the respective software; converting, by the device, the first information into machine-readable rules indicative of the terms of the agreements associated with the respective software for the communication system; and based on a result of comparing the machine-readable rules and first information indicative of the first configuration, determining, by the device, whether the first configuration of the communication system satisfies a defined criterion.

Abstract: Techniques are disclosed for implementing direct memory access in a virtualized computing environment. A memory access policy of the virtualized computing environment is applied to a direct memory access connection request received from a first virtual machine via an exception path. The request is flagged to indicate that the request has been processed and the request is forwarded to a network interface device configured to execute offloaded network functions for one or more virtual machines. A memory access policy of the virtualized computing environment is applied to a direct memory access connection reply received from a second virtual machine on the exception path. The reply is flagged to indicate that the reply has been processed and the reply is forwarded to the network interface device. A direct memory access connection is established between first and second virtual machines in accordance with the request.

Abstract: Leveraging the competence of a computing cluster and a storage cluster by dynamically adjusting an SQL (structured query language) execution path to optimize system performance with fast response time under a distributed relational database on a data storage platform. Also, endpoint devices are divided into groups for the purpose of responding to queries and/or query blocks. The most suitable group of endpoint devices is chosen to respond to a given query or query block.

Abstract: Various embodiments of the present technology can include systems, methods, and non-transitory computer readable media configured to receive request information associated with a first request. The request information comprises user information associated with a first user. A first usage threshold associated with the first user is determined based on the user information. A determination is made that the first request has exceeded the first usage threshold. The first request is automatically terminated based on the determination that the first request has exceeded the first usage threshold.

Abstract: Techniques are disclosed relating to context switching in a SIMD processor. In some embodiments, an apparatus includes pipeline circuitry configured to execute graphics instructions included in threads of a group of single-instruction multiple-data (SIMD) threads in a thread group. In some embodiments, context switch circuitry is configured to atomically: save, for the SIMD group, a program counter and information that indicates whether threads in the SIMD group are active using one or more context switch registers, set all threads to an active state for the SIMD group, and branch to handler code for the SIMD group. In some embodiments, the pipeline circuitry is configured to execute the handler code to save context information for the SIMD group and subsequently execute threads of another thread group. Disclosed techniques may allow instruction-level context switching even when some SIMD threads are non-active.