Directed event signaling for multiprocessor systems

Abstract

The multiprocessor computing device method manages resource access by signaling an event manager to signal a processor element requesting access to a resource to wake up to access the resource when the resource is available, or the process to access the resource when the resource is busy. The device component waits for an event. A processor element may enter a sleep state while waiting to access a requested resource. When multiple elements are waiting for the resource, the processor element with the highest assigned priority is signaled to wake up when the resource is available (608), without waking up other elements. Priority may be assigned to processor elements waiting for the resource based on heuristics or parameters depending on the state of the computing device or processor elements. A sleep duration for a processor element waiting for a resource can be estimated, and the processor element can be removed from the scheduling queue or assigned another thread during the sleep duration.

Description

Background technique [0001] Multicore processors rely on atomicity to ensure the correctness of parallel applications. In order to prevent other processes from interrupting the execution of atomic operations, one implemented solution is for processes to acquire locks on the resources needed to execute critical sections in a mutex, so that until these resources are released from these locks, other processes are prevented from using the same resource. An example of one such lock is a spin lock. A spinlock is an atomic operation that actively waits until the lock is acquired, repeatedly checking whether the lock is available. Because the spinlock process remains active but not performing useful tasks, this active waiting operation consumes energy as the processor continually loads the lock from memory. Other locking or other resource acquisition schemes are similarly associated with some cost to achieve efficiency and performance of the device. To save energy, some processor ...

AI Technical Summary

Problems solved by technology

This can lead to performance degradation and increased resource usage in architectures using the indiscriminate instruction signaling described above when only one core is able to use the resource, where multiple processes are signaled The processor core wakes up to obtain an available lock

Images