US 9,542,238 B2Granted

Systems and methods for direct memory access coherency among multiple processing cores

AssigneeNXP USA, Inc.

InventorsKun Xu, Tommi M. Jokinen, Zheng Xu

Life of this patent

Estimating the term…

Loading patent…

Abstract

A multi-core system configured to execute a plurality of tasks and having a semaphore engine and a direct memory access (DMA) engine capable of selecting, by a task scheduler of a first core, a first task for execution by the first core. In response to a semaphore lock request, the task scheduler of the first core switches the first task to an inactive state and selects a next task for execution by the first core. After the semaphore engine acquires the semaphore lock of the first semaphore, a data transfer request is provided to the DMA engine. In response to the data transfer request, the DMA engine transfers data associated with the locked first semaphore to the entry of the workspace of the first core.

Claims18

Dependent claims

In a multi-core system configured to execute a plurality of tasks and having a semaphore engine and a direct memory access (DMA) engine, a method comprises:

selecting, by a task scheduler of a first core, a first task for execution by the first core, wherein executing the first task comprises allocating an entry in a workspace memory of the first core, and requesting a semaphore lock of a first semaphore identified by a first semaphore identifier;
in response to the semaphore lock request, the task scheduler of the first core switching the first task to an inactive state and selecting a next task for execution by the first core;
in response to the semaphore lock request, the semaphore engine, acquiring the semaphore lock of the first semaphore for the first task;
in response to acquiring the semaphore lock of the first semaphore for the first task, the semaphore engine providing a data transfer request to the DMA engine;
in response to the data transfer request, transferring, by the DMA engine, data associated with the locked first semaphore to the entry of the workspace memory of the first core;
after the transferring the data, the task scheduler switching the first task from the inactive state to a ready for scheduling state so that the first task can be selected for execution by the task scheduler;
when there is no other task waiting to acquire lock of the first semaphore, the DMA engine transferring modified data from the entry of the workspace memory of the first core to external memory;
and
when there is at least one other task waiting to acquire lock of the first semaphore, the method further comprises:
selecting, by the semaphore engine, a task of the at least one other task to acquire lock of the first semaphore;
and
transferring, by the DMA engine, the modified data from the entry of the workspace memory of the first core to an entry of a workspace allocated by the selected task of the at least one other task.

9 more elements

A multi-core system configured to execute a plurality of tasks, comprising:

a plurality of cores, each core having a workspace memory and a task scheduler which selects tasks for execution by the core, wherein when a selected task of a core requests a semaphore lock of a semaphore identified by a semaphore identifier, the task scheduler of the core executing the selected task is configured to, in response to the semaphore lock request, switch the selected task to an inactive state such that a next task for execution can be selected for execution by the core;
a semaphore engine coupled to the plurality of cores, wherein the semaphore engine is configured to,
in response to the semaphore lock request by the selected task, acquire the semaphore lock of the semaphore for the selected task and provide a data transfer request when the semaphore lock is acquired;
and
when there is at least one other task waiting to acquire lock of the first semaphore, select a task of the at least one other task to acquire lock of the first semaphore;
a direct memory access (DMA) engine coupled to the semaphore engine and the plurality of cores, wherein the DMA engine is configured to,
in response to the data transfer request from the semaphore engine, transfer data associated with the locked semaphore to the workspace memory of the core which executed the selected task, wherein after the data transfer is complete, the task scheduler which selected the selected task is configured to switch the selected task from the inactive state to a ready for scheduling state so that the selected task can again be selected for execution;
when there is no other task waiting to acquire lock of the first semaphore, transfer modified data from the entry of the workspace memory of the first core to external memory;
and
transfer the modified data from the entry of the workspace memory of the first core to an entry of a workspace allocated by the selected task of the at least one other task.

7 more elements

Dependencies are inferred from the claim text.

Facts

Application: 13/900,761Filed 23 May 2013
Published: US 2014/0351825 A127 Nov 2014
Granted: 10 Jan 2017
Term ends: Not estimatedThe source records don’t support an estimate.

Classification7 codes

G06F 9/526
Program synchronisation; Mutual exclusion, e.g. by means of semaphores: mutual exclusion algorithms
G06F 9/52
Multiprogramming arrangements: program synchronisation; Mutual exclusion, e.g. by means of semaphores
G06F 9/46
Using stored programs, i.e. using an internal store of processing equipment to receive or retain programs: multiprogramming arrangements
G06F 9/4843
Task transfer initiation or dispatching: by program, e.g. task dispatcher, supervisor, operating system