Handling exceptions in a data parallel system

Patent metadata

Field	Value
Publication number	US-9304778-B2
Application number	US-201313752941-A
Country	US
Kind code	B2
Filing date	Jan 29, 2013
Priority date	Dec 9, 2008
Publication date	Apr 5, 2016
Grant date	Apr 5, 2016

Abstract

A method of handling exceptions in a data parallel system includes forwarding exceptions thrown by concurrent worker tasks to a coordination task. The thrown exceptions are aggregated into an aggregation exception structure. It is determined whether the aggregation exception structure will be handled by an exception handler. The concurrent worker tasks are unwound when it is determined that the aggregation exception structure will be handled.

First claim

What is claimed is:

1. A method of handling exceptions in a data parallel system, comprising:
forwarding exceptions thrown by concurrent worker tasks to a coordination task;
aggregating the thrown exceptions into an aggregation exception structure;
determining whether the aggregation exception structure will be handled by an exception handler;
unwinding the concurrent worker tasks when it is determined that the aggregation exception structure will be handled;
preventing unwinding of the concurrent worker tasks when it is determined that the aggregation exception structure will not be handled; and
wherein the concurrent worker tasks block after throwing an exception until notification is provided to the concurrent worker tasks.

2. The method of claim 1 , and further comprising:
providing the notification to the concurrent worker tasks when it is determined that the aggregation exception structure will not be handled.

3. The method of claim 2 , and further comprising:
searching for an exception handler for the exceptions thrown by the concurrent worker tasks when the concurrent worker tasks are provided the notification.

4. The method of claim 3 , and further comprising:
causing at least one of the concurrent worker tasks to crash if an exception handler is not found for an exception thrown by the at least one concurrent worker task.

5. The method of claim 1 , and further comprising:
terminating the concurrent worker tasks;
throwing the aggregation exception structure with the coordination task after the concurrent worker tasks have terminated.

6. The method of claim 1 , wherein the forwarding, aggregating, and determining occur during a first pass of a two-pass structured exception handling system.

7. The method of claim 6 , wherein the unwinding occurs during a second pass of the two-pass structured exception handling system.

8. A computer-readable storage medium storing computer-executable instructions for performing a method, comprising:
forwarding exceptions thrown by concurrent worker tasks to a coordination task;
aggregating the thrown exceptions into an aggregation exception structure;
determining whether the aggregation exception structure will be handled by an exception handler;
performing unwinding of the concurrent worker tasks based on the determination of whether the aggregation exception structure will be handled;
preventing unwinding of the concurrent worker tasks when it is determined that the aggregation exception structure will not be handled; and
wherein the concurrent worker tasks block after throwing an exception until a notification is provided to the concurrent worker tasks.

9. The computer-readable storage medium of claim 8 , wherein the method further comprises:
providing the notification to the concurrent worker tasks when it is determined that the aggregation exception structure will not be handled.

10. The computer-readable storage medium of claim 9 , wherein the method further comprises:
searching for an exception handler for the exceptions thrown by the concurrent worker tasks when the concurrent worker tasks are provided the notification.

11. The computer-readable storage medium of claim 10 , wherein the method further comprises:
causing at least one of the concurrent worker tasks to crash if an exception handler is not found for an exception thrown by the at least one concurrent worker task.

12. The computer-readable storage medium of claim 8 , wherein the method further comprises:
terminating the concurrent worker tasks;
throwing the aggregation exception structure with the coordination task after the concurrent worker tasks have terminated.

13. The computer-readable storage medium of claim 8 , wherein the forwarding, aggregating, and determining occur during a first pass of a two-pass structured exception handling system.

14. The computer-readable storage medium of claim 13 , wherein the unwinding occurs during a second pass of the two-pass structured exception handling system.

15. A method of handling exceptions in a data parallel system, comprising:
forwarding a plurality of exceptions thrown by a plurality of concurrent threads to a coordination task;
aggregating the thrown exceptions into an aggregation exception structure;
determining whether the aggregation exception structure, comprising the plurality of exceptions thrown by the plurality of concurrent threads, will be handled by an exception handler;
unwinding the concurrent threads when it is determined that the aggregation exception structure will be handled;
preventing unwinding of the concurrent threads when it is determined that the aggregation exception structure will not be handled; and
wherein the concurrent threads block after throwing an exception until a notification is provided to the concurrent threads.

16. The method of claim 15 , wherein the method further comprises:
providing the notification to the concurrent threads when it is determined that the aggregation exception structure will not be handled; and
searching for an exception handler for the exceptions thrown by the concurrent threads when the concurrent threads are provided the notification.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Duffy John

Classifications

• G06F9/4812 Primary

  by interrupt, e.g. masked · CPC title

• G06F2209/481

  Exception handling · CPC title

• G06F9/546

  Message passing systems or structures, e.g. queues · CPC title

• G06F11/0793

  Remedial or corrective actions (recovery from an exception in an instruction pipeline G06F9/3861; by retry G06F11/1402; for recovering from a failure of a protocol instance or entity H04L69/40) · CPC title

• G06F13/24

  using interrupt (G06F13/32 takes precedence) · CPC title