Workgroup handling in pipelined circuits

• US 9,135,087 B1
• Filed: 12/27/2012
• Issued: 09/15/2015
• Est. Priority Date: 12/27/2012
• Status: Active Grant

• Current Assignee: Altera Corporation (Intel Corporation)
• Original Assignee: Altera Corporation (Intel Corporation)
• Inventors: Czajkowski, Tomasz, Freeman, John, Yiannacouras, Peter
• Primary Examiner(s): Bullock, Jr., Lewis A
• Assistant Examiner(s): DASCOMB, JACOB D
• Application Number:
  US13/727,813
• Time in Patent Office: 992 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 9/54 Interprogram communication

Abstract

Systems and methods for limiting resource usage of a kernel of an integrated circuit are provided. For example, in one embodiment a method for limiting a number of workgroups that may simultaneously access a kernel of an integrated circuit (IC) includes determining a threshold number of workgroups that may access the kernel simultaneously. A thread of execution is received. The thread of execution is allowed to access the kernel when the threshold number of workgroups would not be exceeded by the thread of execution accessing the kernel.

Claims

1. A method for limiting a number of workgroups that may simultaneously access a kernel of an integrated circuit (IC), the method comprising:

tracking a plurality of threads that currently access the kernel and workgroup information associated with the plurality of threads;

determining a threshold number of workgroups that may access the kernel simultaneously, wherein the threshold is based on an allotment of additional workgroups that are expected not to cause overuse, depletion or both of hardware resources of the integrated circuit;

receiving a first thread of execution and workgroup information associated with the first thread;

determining, based on the workgroup information associated with the first thread and workgroup information associated with the plurality of threads that currently access the kernel, if allowing the first thread access to the kernel will result in an additional workgroup accessing the kernel;

allowing the first thread to access the kernel when it is determined the first thread would not result in an additional workgroup accessing the kernel;

allowing the first thread to access the kernel when it is determined the first thread would not exceed the threshold; and

disallowing the first thread to access the kernel when doing so would exceed the threshold number of workgroups.

2. The method of claim 1, comprising:

determining whether or not the first thread of execution is valid; and

allowing the first thread of execution to access the kernel only when the thread of execution is valid.

3. The method of claim 1, comprising:

wherein the threshold number of workgroups is determined by;

accessing a workgroup table;

determining the number of empty rows in the workgroup; and

setting the threshold to the number of empty rows in the workgroup table.

4. The method of claim 3, comprising:

tracking the threads that access the kernel and workgroup information associated with the threads that access the kernel;

receiving a second thread of execution and associated workgroup information;

determining whether an additional workgroup may access the kernel by comparing the threshold with the workgroup information associated with the threads that access the kernel;

determining whether an additional workgroup would be accessing the kernel if the second thread was allowed to access the kernel by comparing the workgroup information associated with the second thread of execution with the workgroup information associated with the threads that access the kernel; and

allowing the second thread of execution to access the kernel when an additional workgroup may access the kernel, when the second thread of execution does not result in an additional workgroup accessing the kernel, or both.

5. The method of claim 1, comprising:

receiving workgroup information associated with the first thread of execution;

accessing a workgroup table upon receiving the first thread of execution;

determining whether the workgroup information corresponds to workgroup information already stored in the workgroup table;

when the workgroup information does not correspond to workgroup information already stored in the workgroup table and there is an empty row in the workgroup table, adding the workgroup information to the empty slot in the workgroup table; and

when the workgroup information does not correspond to workgroup information already stored in the workgroup table and there is not an empty row in the workgroup table disallowing the first thread of execution from accessing the kernel.

6. The method of claim 5, comprising disallowing the first thread of execution by providing a stall signal to an upstream component providing the first thread of execution.

7. The method of claim 5, comprising:

determining when the first thread of execution has finished accessing the kernel; and

removing the workgroup information associated to the first thread of execution from the workgroup table when the first thread of execution has finished accessing the kernel.

8. An integrated circuit (IC), comprising:

programmable logic configured to store a programmable design, wherein the programmable design is configured to implement customized functions on the IC device; and

a workgroup limiter configured to limit a number of workgroups accessing a kernel of the programmable design, the workgroup limiter configured to;

track a plurality of threads that currently access the kernel and workgroup information associated with the plurality of threads;

determine a threshold number of workgroups that may access the kernel simultaneously, wherein the threshold is based on an allotment of additional workgroups that are expected not to cause overuse, depletion or both of hardware resources of the integrated circuit;

receive a first thread of execution and workgroup information associated with the first thread;

determine, based on the workgroup information associated with the first thread and workgroup information associated with the plurality of threads that currently access the kernel, if allowing the first thread access to the kernel will result in an additional workgroup accessing the kernel;

allow the first thread to access the kernel when it is determined the first thread would not result in an additional workgroup accessing the kernel;

allow the first thread to access the kernel when it is determined the first thread would not exceed the threshold; and

disallow the first thread to access the kernel when doing so would exceed the threshold number of workgroups.

9. The integrated circuit of claim 8, wherein the workgroup limiter comprises dedicated circuitry apart from the programmable logic.

10. The integrated circuit of claim 8, wherein the workgroup limiter comprises:

thread inputs configured to provide thread information relating to threads attempting to access the kernel of the programmable design;

a workgroup table configured to store data relating to threads that are accessing the kernel; and

a stall output configured to provide a stall signal to upstream components indicating that the thread is not permitted to access the kernel.

11. The integrated circuit of claim 8, wherein the workgroup limiter comprises an encoder configured to:

obtain a workgroup identifier representative of a workgroup associated with threads attempting to access the kernel;

obtain a thread identifier representative of the threads attempting to access the kernel;

determine whether the workgroup identifier is in a workgroup table;

determine whether there is an empty row in the workgroup table;

insert the workgroup identifier and the thread identifier in the workgroup table if the workgroup identifier is not in the workgroup table and there is an empty row in the workgroup table; and

provide the stall signal when the workgroup identifier is not in the workgroup table and there is not an empty row in the workgroup table.

12. The integrated circuit of claim 11, wherein the encoder is configured to store the thread identifier in a row containing the workgroup identifier when the workgroup identifier is in the workgroup table.

13. The integrated circuit of claim 11, wherein the workgroup limiter comprises thread outputs configured to provide permitted thread data comprising the thread identifier and the workgroup identifier to the kernel after the workgroup identifier and the thread identifier is inserted into the workgroup table or when the workgroup identifier is in the workgroup table.

14. The integrated circuit of claim 11, wherein the encoder is configured to remove the workgroup identifier from the workgroup table when all threads associated with the workgroup identifier are no longer accessing the kernel.

15. The integrated circuit of claim 11, wherein the workgroup table is configured to store a size of workgroups accessing the kernel, threads associated with a workgroup that are accessing the kernel, an amount of time that each workgroup has been accessing the kernel, an amount of time that each thread has been accessing the kernel, or a combination thereof.

16. The integrated circuit of claim 11, wherein the workgroup limiter comprises a stall input configured to receive a stall signal from downstream components, wherein the encoder is configured to provide a stall signal to upstream components when the stall signal is received from the stall input.

17. A tangible, non-transitory, computer-readable medium, comprising instructions to:

provide a programmable logic interface for an integrated circuit, the programmable logic interface being configured to enable a designer to implement an integrated circuit design in programmable logic of the integrated circuit; and

generate a workgroup limiter in the programmable logic, the workgroup limiter configured to;

track a plurality of threads that currently access a kernel and workgroup information associated with the plurality of threads;

determine a threshold number of workgroups that may access the kernel simultaneously, wherein the threshold is based on an allotment of additional workgroups that are expected not to cause overuse, depletion or both of hardware resources of the integrated circuit;

receive a first thread of execution and workgroup information associated with the first thread;

determine, based on the workgroup information associated with the first thread and workgroup information associated with the plurality of threads that currently access the kernel, if allowing the first thread access to the kernel will result in an additional workgroup accessing the kernel;

allow the first thread to access the kernel when it is determined the first thread would not result in an additional workgroup accessing the kernel;

allow the first thread to access the kernel when it is determined the first thread would not exceed the threshold; and

disallow the first thread to access the kernel when doing so would exceed the threshold number of workgroups.

18. The tangible, non-transitory, computer-readable medium of claim 17, comprising instructions to analyze the workgroup limiter to obtain operational parameters of the integrated circuit design, the operational parameters comprising:

an access time of threads, workgroups, or both accessing the kernel;

a number of workgroups accessing the kernel;

a number of threads accessing the kernel;

thread identities of threads accessing the kernel;

a number of threads of a particular workgroup that are accessing the kernel;

the size of a workgroup accessing the kernel;

or

a combination thereof.

19. The tangible, non-transitory, computer-readable medium of claim 17, comprising instructions to dynamically adjust the threshold number of workgroups that may access the kernel of the integrated circuit design based upon an analysis of the implementation, a user interaction, or both.