Difference: ACC (8 vs. 9)

Revision 92010-03-04 - MichaelScott

Line: 1 to 1
 
META TOPICPARENT name="KonstantinosMenychtas"

Accelerator Centric Computing

Line: 61 to 61
 

Task queues

  1. There exists one accelerator task queue for every type of accelerator. Every task queue is unique and can be identified by its id.
Changed:
<
<
  1. Whether the task-queue appears as a blocking LIFO queue, or some other data structure, is a matter of policy implementation. Assume a LIFO queue for now. But later software running on TQ Controller can dynamically schedule tasks based on various constraints: power/performance/priority, note that the scheduling algorithm runs in software.
>
>
  1. Whether the task-queue appears as a blocking FIFO queue, or some other data structure, is a matter of policy implementation. Assume a FIFO queue for now. But later software running on TQ Controller can dynamically schedule tasks based on various constraints: power/performance/priority. Note that the scheduling algorithm runs in software.
 
  1. The contents of the task queue lie in main memory, in the kernel address space. This allows it to expand without limitations and be protected from arbitrary user writes.
  2. The TQ can only be read or written by the task queue manager. The only way it is visible to the system, is by expand/shrink requests in (hopefully) rare occasions.
  3. The TQ contains task descriptors. The task descriptors are described through a software/hardware contract. Only pointers of task descriptor are stored in TQ, which makes all TQ the same.
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2017 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding URCS? Send feedback