Abstract: The performance of an application is improved by identifying and eliminating items with dead time-stamps and eliminating work on items with irrelevant time-stamps. An algorithm executing in each node of a task graph computes and propagates guarantees which are used to eliminate both items with dead time-stamps and irrelevant computation on dead time-stamps. A continuous garbage collector eliminates items with dead time-stamps while the node continues to process received items. Unnecessary computations are reduced by automatically discerning the interest set of downstream modules for time-stamps and feeding the interest set back to upstream modules.

Abstract: A static schedule is selected from a set of static schedules for an application dependent on the state of the application. A scheduling system stores a set of pre-defined static schedules for each state of the application. A scheduling system learns the costs of predefined schedules for each state of the application on-line as the application executes. Upon the detection of a state change in the application during run-time, the scheduling system selects a new static schedule for the application. The new static schedule is determined based on schedule costs and exploration criteria.

Abstract: A static schedule is selected from a set of static schedules for an application dependent on the state of the application. A scheduling system stores a set of pre-defined static schedules for each state of the application. A scheduling system learns the costs of predefined schedules for each state of the application on-line as the application executes. Upon the detection of a state change in the application during run-time, the scheduling system selects a new static schedule for the application. The new static schedule is determined based on schedule costs and exploration criteria.

Abstract: A static schedule is selected from a set of static schedules for an application dependent on the state of the application. A scheduling system stores a set of pre-defined static schedules for each state of the application. A scheduling system learns the costs of predefined schedules for each state of the application on-line as the application executes. Upon the detection of a state change in the application during run-time, the scheduling system selects a new static schedule for the application. The new static schedule is determined based on schedule costs and exploration criteria.

Abstract: The performance of an application is improved by identifying and eliminating items with dead time-stamps and eliminating work on items with irrelevant time-stamps. An algorithm executing in each node of a task graph computes and propagates guarantees which are used to eliminate both items with dead time-stamps and irrelevant computation on dead time-stamps. A continuous garbage collector eliminates items with dead time-stamps while the node continues to process received items. Unnecessary computations are reduced by automatically discerning the interest set of downstream modules for time-stamps and feeding the interest set back to upstream modules.

Abstract: A system for learning and applying a task and data parallel strategy to an application that includes at least one task for processing an input data stream to produce an output data stream includes the following components. A controller measuring an execution of the application to generate an action space representing a task and data parallel strategy. A run-time system applying the action space to implement the task and data parallel strategy.

Abstract: A system for integrating task and data parallelism in a dynamic application that includes at least one task for processing an input data stream to produce an output data stream replaces the at least one task with the following components. A splitter task for partitioning the input data stream into a plurality of data chunks. A plurality of worker tasks for processing subsets of the data chunks, each worker task being an instance of the at least one task, and a joiner task combining the processed data chunks to produce the output data stream.

Abstract: A system for learning and applying a task and data parallel strategy to an application that includes at least one task for processing an input data stream to produce an output data stream includes the following components. A controller measuring an execution of the application to generate an action space representing a task and data paralell strategy. A run-time system applying the action space to implement the task and data prallel strategy.

Abstract: A system for integrating task and data parallelism in a dynamic application that includes at least one task for processing an input data stream to produce an output data stream replaces the at least one task with the following components. A splitter task for partitioning the input data stream into a plurality of data chunks. A plurality of worker tasks for processing subsets of the data chunks, each worker task being an instance of the at least one task, and a joiner task combining the processed data chunks to produce the output data stream.