Hardware counters to track utilization in a multithreading computer system

• US 10,102,004 B2
• Filed: 03/27/2014
• Issued: 10/16/2018
• Est. Priority Date: 03/27/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bartik, Jane H., Bradbury, Jonathan D., King, Gary M., Rosa, Daniel V., Schmidt, Donald W.
• Primary Examiner(s): Petranek, Jacob
• Application Number:
  US14/226,980
• Publication Number:
  US 20150277922A1
• Time in Patent Office: 1,664 Days
• CPC Class Codes: G06F 1/10 Distribution of clock signals , e.g. skew
• CPC Class Codes: G06F 9/30145 Instruction analysis, e.g. decoding, instruction word fields
• CPC Class Codes: G06F 9/3851 from multiple instruction streams, e.g. multistreaming
• CPC Class Codes: G06F 9/3861 Recovery, e.g. branch miss-prediction, exception handling error detection or correction G06F11/00
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system

Abstract

Embodiments relate tracking utilization in a multithreading (MT) computer system. According to one aspect, a computer system includes a configuration with a core configured to operate in a MT that supports multiple threads on shared resources of the core. The core is configured to perform a method that includes resetting a plurality of utilization counters. The utilization counters include a plurality of sets of counters. During each clock cycle on the core, a set of counters is selected from the plurality of sets of counters. The selecting is based on a number of currently active threads on the core. In addition, during each clock cycle a counter in the selected set of counters is incremented based on an aggregation of one or more execution events at the multiple threads of the core. Values of the utilization counters are provided to a software program.

Claims

1. A computer system, comprising:

a configuration comprising a core configured to operate in a multithreading (MT) mode, the MT mode supporting multiple threads on shared resources of the core;

the core configured to perform a method comprising;

resetting a plurality of utilization counters, the utilization counters comprising a plurality of sets of counters including a first set of counters and a second set of counters, each set of counters corresponding to a different number of currently active threads and each set of counters non-overlapping with each other set of counters;

performing for each clock cycle on the core;

selecting a set of counters from the plurality of sets of counters, the selecting based on a number of currently active threads on the core and not based on which specific threads of the multiple threads are currently active, wherein the first set of counters is selected based on exactly one of the multiple threads being currently active, and the second set of counters is selected based on exactly two of the multiple threads being currently active; and

incrementing a counter in the selected set of counters, the incrementing based on an aggregation of one or more execution events at the multiple threads of the core; and

providing values of the utilization counters to a software program.

2. The computer system of claim 1, wherein the execution event includes a clock cycle and the counter in the selected set of counters is incremented by one.

3. The computer system of claim 2, wherein the execution event further includes an instruction completion and an other counter in the selected set of counters is incremented based on a number of instruction completions on all of the currently active threads during the clock cycle.

4. The computer system of claim 2, wherein the execution event further includes a cache miss and an other counter in the selected set of counters is incremented based on a number of cache misses on all of the currently active threads during the clock cycle.

5. The computer system of claim 2, wherein the execution event further includes a branch misprediction and an other counter in the selected set of counters is incremented based on a number of branch mispredictions on all of the currently active threads during the clock cycle.

6. The computer system of claim 1, wherein a thread is currently active when the thread is valid and not in a wait state.

7. The computer system of claim 1, wherein the software program is an operating system or a hypervisor.

8. A computer program product for tracking utilization in a configuration comprising a core configured to operate in a multithreading (MT) mode, the MT mode supporting multiple threads on shared resources of the core, the computer program product comprising:

a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a signal, the program instructions readable by a processing circuit to cause the processing circuit to perform a method comprising;

resetting a plurality of utilization counters, the utilization counters comprising a plurality of sets of counters including a first set of counters and a second set of counters, each set of counters corresponding to a different number of currently active threads and each set of counters non-overlapping with each other set of counters;

performing for each clock cycle on the core;

selecting a set of counters from the plurality of sets of counters, the selecting based on a number of currently active threads on the core and not based on which specific threads of the multiple threads are currently active, wherein the first set of counters is selected based on exactly one of the multiple threads being currently active, and the second set of counters is selected based on exactly two of the multiple threads being currently active; and

incrementing a counter in the selected set of counters, the incrementing based on an aggregation of one or more execution events at the multiple threads of the core; and

providing values of the utilization counters to a software program.

9. The computer program product of claim 8, wherein the execution event includes a clock cycle and the counter in the selected set of counters is incremented by one.

10. The computer program product of claim 9, wherein the execution event further includes an instruction completion and an other counter in the selected set of counters is incremented based on a number of instruction completions on all of the currently active threads during the clock cycle.

11. The computer program product of claim 9, wherein the execution event further includes a cache miss and an other counter in the selected set of counters is incremented based on a number of cache misses on all of the currently active threads during the clock cycle.

12. The computer program product of claim 9, wherein the execution event further includes a branch misprediction and an other counter in the selected set of counters is incremented based on a number of branch mispredictions on all of the currently active threads during the clock cycle.

13. The computer program product of claim 8, wherein a thread is currently active when the thread is valid and not in a wait state.