Techniques for providing improved affinity scheduling in a multiprocessor computer system

• US 8,051,418 B1
• Filed: 03/21/2005
• Issued: 11/01/2011
• Est. Priority Date: 03/21/2005
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Oracle America, Inc. (Oracle Corporation)
• Inventors: Dice, David
• Primary Examiner(s): An; Meng
• Assistant Examiner(s): LEE, TAMMY EUNHYE
• Application Number:
  US11/084,951
• Time in Patent Office: 2,416 Days
• Field of Search: 718/100, 718/102, 718/103, 718/104, 718/105
• US Class Current: 718/102
• CPC Class Codes: G06F 9/5033 considering data affinity

Abstract

Techniques for controlling a thread on a computerized system having multiple processors involve accessing state information of a blocked thread, and maintaining the state information of the blocked thread at current values when the state information indicates that less than a predetermined amount of time has elapsed since the blocked thread ran on the computerized system. Such techniques further involve setting the state information of the blocked thread to identify affinity for a particular processor of the multiple processors when the state information indicates that at least the predetermined amount of time has elapsed since the blocked thread ran on the computerized system. Such operation enables the system to place a cold blocked thread which shares data with another thread on the same processor of that other thread so that, when the blocked thread awakens and runs, that thread is closer to the shared data.

Claims

1. A method of controlling a thread on a computerized system having multiple processors, the method comprising:

accessing state information of a blocked thread;

maintaining the state information of the blocked thread when the state information indicates that less than a predetermined amount of time has elapsed since the blocked thread ran on the computerized system;

setting the state information of the blocked thread to identify affinity for a particular processor of the multiple processors when the state information indicates that at least the predetermined amount of time has elapsed since the blocked thread ran on the computerized system, wherein the particular processor of the multiple processors is selected based on state information of another thread run on the particular processor, the another thread waking the blocked thread and having shared data with the blocked thread; and

while setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors, setting the state information of the blocked thread to indicate that less than the predetermined amount of time has elapsed since the blocked thread ran on the computerized system.

2. The method of claim 1, further comprising:

after setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors and after setting the state information of the blocked thread to indicate that less than the predetermined amount of time has elapsed, waking the blocked thread.

3. The method of claim 2, further comprising:

in response to waking of the blocked thread, setting the state information of the blocked thread to no longer identify affinity for the particular processor of the multiple processors.

4. The method of claim 2, further comprising:

limiting a total number of blocked threads having affinity for the particular processor of the multiple processors to a preset number.

5. The method of claim 2 wherein the multiple processors have processor identifiers which distinguish the multiple processors from each other;

and wherein setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors includes;

in response to running the another thread on the particular processor, reading a processor identifier of the particular processor from state information of the another thread and modifying the state information of the blocked thread to include the processor identifier to set the state information of the blocked thread to identify affinity for the particular processor.

6. The method of claim 5, further comprising:

sharing data within shared memory between the blocked thread and the another thread.

7. The method of claim 6 wherein the blocked thread and the another thread are configured to coordinate access to the data within the shared memory using JAVA monitors;

and wherein accessing, maintaining and setting are performed in response to a JAVA-based procedure call performed by the another thread.

8. The method of claim 5 wherein the computerized system has a multiprocessor-per-die-based configuration formed, at least in part, by (i) a first die which includes first processing units and a first resident cache and (ii) a second die which includes second processing units and a second resident cache;

wherein the blocked thread ran most-recently on the first die using cache lines of the first resident cache;

wherein the another thread runs on the second die using cache lines of the second resident cache; and

wherein modifying the state information of the blocked thread to include the processor identifier includes;

adjusting the state information of the blocked thread to specify, as the particular processor of the multiple processors, one of the second processing units of the second die.

9. The method of claim 5 wherein the computerized system has a NUMA-based configuration formed, at least in part, by (i) a first node which includes first processing units and a first adjacent cache and (ii) a second node which includes second processing units and a second adjacent cache;

wherein the blocked thread ran most-recently on the first node using cache lines of the first adjacent cache;

wherein the another thread runs on the second node using cache lines of the second adjacent cache; and

wherein modifying the state information of the blocked thread to include the processor identifier includes;

adjusting the state information of the blocked thread to specify, as the particular processor of the multiple processors, one of the second processing units of the second node.

10. A non-transitory computer readable medium having computer executable instructions stored thereon for controlling a thread on a computer system having multiple processors, the instructions being configured to cause the computer system to:

access state information of a blocked thread;

maintain the state information of the blocked thread when the state information indicates that less than a predetermined amount of time has elapsed since the blocked thread ran on the computerized system;

set the state information of the blocked thread to identify affinity for a particular processor of the multiple processors when the state information indicates that at least the predetermined amount of time has elapsed since the blocked thread ran on the computerized system, wherein the particular processor of the multiple processors is selected based on state information of another thread run on the particular processor, the another thread waking the blocked thread and having shared data with the blocked thread; and

while setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors, set the state information of the blocked thread to indicate that less than the predetermined amount of time has elapsed since the blocked thread ran on the computerized system.

11. The computer readable medium of claim 10 wherein the instructions are further configured to cause the computer system to:

after setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors and after setting the state information of the blocked thread to indicate that less than the predetermined amount of time has elapsed, wake the blocked thread.

12. The computer readable medium of claim 11 wherein the instructions are further configured to cause the computer system to:

in response to waking of the blocked thread, set the state information of the blocked thread to no longer identify affinity for the particular processor of the multiple processors.

13. The computer readable medium of claim 11 wherein the multiple processors have processor identifiers which distinguish the multiple processors from each other;

and wherein the instructions, when causing the computer system to set the state information of the blocked thread to identify affinity for the particular processor of the multiple processors, are configured to cause the computer system to;

in response to running the another thread on the particular processor, read a processor identifier of the particular processor from state information of the another thread and modify the state information of the blocked thread to include the processor identifier to set the state information of the blocked thread to identify affinity for the particular processor.

14. A computerized system, comprising:

multiple processors configured to run threads, the multiple processors being configured to;

access state information of a blocked thread;

maintain the state information of the blocked thread when the state information indicates that less than a predetermined amount of time has elapsed since the blocked thread ran on the computerized system;

set the state information of the blocked thread to identify affinity for a particular processor of the multiple processors when the state information indicates that at least the predetermined amount of time has elapsed since the blocked thread ran on the computerized system, wherein the particular processor of the multiple processors is selected based on state information of another thread run on the particular processor, the another thread waking the blocked thread and having shared data with the blocked thread; and

while setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors, set the state information of the blocked thread to indicate that less than the predetermined amount of time has elapsed since the blocked thread ran on the computerized system.

15. The computerized system of claim 14 wherein the multiple processors are further configured to:

after setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors and after setting the state information of the blocked thread to indicate that less than the predetermined amount of time has elapsed, wake the blocked thread.

16. The computerized system of claim 15 wherein the multiple processors are further configured to:

in response to waking of the blocked thread, set the state information of the blocked thread to no longer identify affinity for the particular processor of the multiple processors.

17. The computerized system of claim 15 wherein the multiple processors have processor identifiers which distinguish the multiple processors from each other;

and wherein the multiple processors, when setting the state information of the blocked thread to identify affinity for the particular processor of the multiple processors, are configured to;

in response to running the another thread on the particular processor, read a processor identifier of the particular processor from state information of the another thread and modify the state information of the blocked thread to include the processor identifier to set the state information of the blocked thread to identify affinity for the particular processor.