Run-time code parallelization with continuous monitoring of repetitive instruction sequences

• US 9,348,595 B1
• Filed: 12/22/2014
• Issued: 05/24/2016
• Est. Priority Date: 12/22/2014
• Status: Expired due to Fees

• Current Assignee: Centipede Semi ltd.
• Original Assignee: Centipede Semi ltd.
• Inventors: Mizrahi, Noam, Mandler, Alberto, Koren, Shay, Friedmann, Jonathan
• Primary Examiner(s): Partridge, William B.
• Application Number:
  US14/578,516
• Time in Patent Office: 519 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 11/30 Monitoring
• CPC Class Codes: G06F 11/3466 Performance evaluation by tracing or monitoring
• CPC Class Codes: G06F 2201/865 Monitoring of software
• CPC Class Codes: G06F 9/3838 Dependency mechanisms, e.g. register scoreboarding
• CPC Class Codes: G06F 9/3851 from multiple instruction streams, e.g. multistreaming

Abstract

A method includes, in a processor that executes instructions of program code, monitoring instructions of a repetitive sequence of the instructions that traverses a flow-control trace so as to construct a specification of register access by the monitored instructions. Based on the specification, multiple hardware threads are invoked to execute respective segments of the repetitive instruction sequence at least partially in parallel. Monitoring of the instructions continues in at least one of the segments during execution.

Claims

1. A processor, comprising:

a processing pipeline comprising multiple hardware threads and configured to execute pre-compiled software code instructions that are stored in a memory;

multiple registers, configured to be read and written to by the processing pipeline during execution of the instructions; and

a monitoring unit, which is configured to monitor execution of the instructions by the processing pipeline and to record, while the processing pipeline executes a sequence of the instructions, a first flow control trace traversed by the sequence and a first monitoring table indicating the registers accessed by the processing pipeline in executing the instructions in the sequence according to the first flow control trace, and which is configured to parallelize among the hardware threads of the processor, using the first monitoring table, execution of first repetitions of the sequence in accordance with the first flow control trace,

wherein the monitoring unit is configured to detect, while the processing pipeline executes the sequence of the instructions, a second flow control trace traversed by the sequence, different from the first flow control trace, and to record a second monitoring table indicating the registers accessed by the processing pipeline in executing the instructions in the sequence according to the second flow control trace, and is configured to parallelize among the hardware threads of the processor, using the second monitoring table, the execution of second repetitions of the sequence in accordance with the second flow control trace,

wherein the first and second monitoring tables comprise, for each register written to by the processing pipeline in executing the instructions in the sequence, a record of the last instruction in the sequence at which the processing pipeline wrote to the register.

2. The processor according to claim 1, wherein the monitoring unit is configured to decide, during the execution of each of the repetitions, when to issue the instructions with regard to a given register in subsequent repetitions based on the record of the last instruction at which the processing pipeline wrote to the given register.

3. The processor according to claim 2, wherein the record of the last instruction comprises a count of a number of times that each register was written to by the processing pipeline in executing the instructions in the sequence, and wherein the monitoring unit is configured to determine, in each of the repetitions, that the last instruction has been written to the given register by counting writes in each of the repetitions up to the count in the record.

4. The processor according to claim 1, wherein the monitoring unit is configured to alternate between the execution of the first repetitions, using the first monitoring table, and the second repetitions, using the second monitoring table.

5. The processor according to claim 4, wherein the execution alternates between the first and second repetitions depending upon a conditional branch instruction in the sequence.

6. The processor according to claim 1, wherein the sequence comprises a loop or a function.

7. The processor according to claim 1, wherein the monitoring unit is configured to continue to monitor the execution of the instructions during all the repetitions of the sequence.

8. The processor according to claim 1, wherein the monitoring unit is configured to select a subset of the repetitions of the sequence, and to continue to monitor the execution of the instructions only within the selected subset.

9. The processor according to claim 8, wherein the monitoring unit is configured to select the subset by performing at least one of:

selecting for continued monitoring every N^th repetition;

selecting the repetitions for continued monitoring in accordance with a predefined periodic pattern; and

selecting the repetitions for continued monitoring at random.

10. The processor according to claim 1, wherein the monitoring unit is configured to terminate monitoring of the instructions in the sequence in accordance with a predefined termination criterion before the sequence ends.

11. A method for operating a processing pipeline that includes multiple hardware threads and is configured to execute pre-compiled software code instructions that are stored in a memory while reading and writing to multiple registers during execution of the instructions, the method including:

monitoring execution of the instructions by the processing pipeline and recording, while the processing pipeline executes a sequence of the instructions, a first flow control trace traversed by the sequence and a first monitoring table indicating the registers accessed by the processing pipeline in executing the instructions in the sequence according to the first flow control trace;

parallelizing among the hardware threads of the processor, using the first monitoring table, execution of first repetitions of the sequence in accordance with the first flow control trace;

while the processing pipeline executes the sequence of the instructions, detecting a second flow control trace traversed by the sequence, different from the first flow control trace;

recording a second monitoring table indicating the registers accessed by the processing pipeline in executing the instructions in the sequence according to the second flow control trace; and

parallelizing among the hardware threads of the processor, using the second monitoring table, the execution of second repetitions of the sequence in accordance with the second flow control trace,

wherein the first and second monitoring tables comprise, for each register written to by the processing pipeline in executing the instructions in the sequence, a record of the last instruction in the sequence at which the processing pipeline wrote to the register.

12. The method according to claim 11, wherein parallelizing the execution comprises deciding, during the execution of each of the repetitions, when to issue the instructions with regard to a given register in subsequent repetitions based on the record of the last instruction at which the processing pipeline wrote to the given register.

13. The method according to claim 12, wherein the record of the last instruction comprises a count of a number of times that each register was written to by the processing pipeline in executing the instructions in the sequence, and wherein deciding when to issue the instructions comprises determining, in each of the repetitions, that the last instruction has been written to the given register by counting writes in each of the repetitions up to the count in the record.

14. The method according to claim 11, wherein parallelizing the execution comprises alternating between the execution of the first repetitions, using the first monitoring table, and the second repetitions, using the second monitoring table.

15. The method according to claim 14, wherein the execution alternates between the first and second repetitions depending upon a conditional branch instruction in the sequence.

16. The method according to claim 11, wherein the sequence comprises a loop or a function.

17. The method according to claim 11, wherein monitoring the execution comprises continuing to monitor the execution of the instructions during all the repetitions of the sequence.

18. The method according to claim 11, wherein monitoring the execution comprises selecting a subset of the repetitions of the sequence, and continuing to monitor the execution of the instructions only within the selected subset.

19. The method according to claim 18, wherein the selecting the subset comprises performing at least one of:

selecting for continued monitoring every N^th repetition;

selecting the repetitions for continued monitoring in accordance with a predefined periodic pattern; and

selecting the repetitions for continued monitoring at random.

20. The method according to claim 11, and comprising terminating monitoring of the instructions in the sequence in accordance with a predefined termination criterion before the sequence ends.