Systems and methods for automatic data management for an asynchronous task-based runtime
A compilation system can define, at compile time, the data blocks to be managed by an Even Driven Task (EDT) based runtime/platform, and can also guide the runtime/platform on when to create and/or destroy the data blocks, so as to improve the performance of the runtime/platform. The compilation system can also guide, at compile time, how different tasks may access the data blocks they need in a manner that can improve performance of the tasks.
1. A method of creating at runtime, in a memory efficient manner, one or more data blocks to be used by one or more tasks to be executed by an event-driven task (EDT) platform comprising a plurality of processing nodes, the method comprising:
performing, by a processor configured as a compiler, compile-time operations comprising:
identifying a data block to be used by a first task to be executed by the event-driven task (EDT) platform;
identifying a first set of predecessor tasks comprising each task that is: (i) predecessor to the first task, and (ii) independent of the data block;
generating a data block creation task directing creation of the data block; and
specifying: (i) the data block creation task as a predecessor of the first task; and (ii) at least one task in the first set of predecessor tasks as a predecessor of the data block creation task, thereby delaying runtime creation and corresponding memory usage of the data block.
2. The method of claim 1 , wherein the first set of predecessor tasks consists essentially of each task that is: (i) an immediate predecessor to the first task, and (ii) independent of the data block.
3. The method of claim 2 , wherein the specifying step comprises specifying each task in the second set of predecessor tasks as a predecessor of the data block creation task.
4. The method of claim 1 , wherein the specifying step comprises specifying each task in the first set of predecessor tasks as a predecessor of the data block creation task.
5. The method of claim 1 , further comprising:
identifying a second task to be executed by the event-driven task (EDT) platform, the second task also using the data block; and
identifying a second set of predecessor tasks comprising each task that is: (i) predecessor to the second task, and (ii) independent of the data block,
wherein the specifying step comprises specifying: (i) the data block creation task as a predecessor of the second task; and (ii) at least one task in the second set of predecessor tasks as a predecessor of the data block creation task.
6. The method of claim 1 , further comprising:
identifying a first set of successor tasks comprising each task that is: (i) successor of the first task, and (ii) independent of the data block;
generating a data block deletion task directing deletion of the data block; and
specifying: (i) the data block deletion task as a successor of the first task; and (ii) at least one task in the first set of successor tasks as a successor of the data block deletion task, thereby advancing runtime deletion and corresponding freeing of memory allocated to the data block.
7. The method of claim 6 , wherein the first set of successor tasks consists essentially of each task that is: (i) an immediate successor of the first task, and (ii) independent of the data block.
8. The method of claim 6 , wherein the specifying step comprises specifying each task in the first set of successor tasks as a successor of the data block deletion task.
9. The method of claim 6 , further comprising:
identifying a second task to be executed by the event-driven task (EDT) platform, the second task also using the data block; and
identifying a second set of successor tasks comprising each task that is: (i) successor of the second task, and (ii) independent of the data block,
wherein the specifying step comprises specifying: (i) the data block deletion task as a successor of the second task; and (ii) at least one task in the second set of successor tasks as a successor of the data block deletion task.
10. The method of claim 9 , wherein the specifying step comprises specifying each task in the second set of successor tasks as a successor of the data block deletion task.