Analyzing update conditions for shared variable directory information in a parallel computer

• US 9,262,243 B2
• Filed: 02/13/2013
• Issued: 02/16/2016
• Est. Priority Date: 12/18/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Archer, Charles J., Carey, James E., Sanders, Philip J., Smith, Brian E.
• Primary Examiner(s): Rones, Charles
• Assistant Examiner(s): LI, HEWY H
• Application Number:
  US13/766,420
• Publication Number:
  US 20140173205A1
• Time in Patent Office: 1,098 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 2209/547 Messaging middleware
• CPC Class Codes: G06F 8/453 Data distribution
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes
• CPC Class Codes: G06F 9/546 Message passing systems or structures, e.g. queues
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

Methods, parallel computers, and computer program products for analyzing update conditions for shared variable directory (SVD) information in a parallel computer are provided. Embodiments include a runtime optimizer receiving a compare-and-swap operation header. The compare-and-swap operation header includes an SVD key, a first SVD address, and an updated first SVD address. The first SVD address is associated with the SVD key in a first SVD associated with a first task. Embodiments also include the runtime optimizer retrieving from a remote address cache associated with the second task, a second SVD address indicating a location within a memory partition associated with the first SVD in response to receiving the compare-and-swap operation header. Embodiments also include the runtime optimizer determining whether the second SVD address matches the first SVD address and transmitting a result indicating whether the second SVD address matches the first SVD address.

Claims

1. A method of analyzing update conditions for shared variable directory (SVD) information in a parallel computer, the parallel computer comprising a plurality of threads of execution, the threads organized into tasks, the parallel computer further comprising memory partitioned to provide each thread with a private memory and a shared memory, the method comprising:

receiving, by a runtime optimizer of a second task, a compare-and-swap operation header, the compare-and-swap operation header including an SVD key, a first SVD address of resource data, and an updated first SVD address for the resource data, wherein the first SVD address is associated with the SVD key in a first SVD associated with a first task, wherein the first SVD address indicates a first location within a shared memory partition associated with the first task, wherein the resource data is stored at the first location within the shared memory partition;

in response to receiving the compare-and-swap operation header, retrieving from a remote address cache associated with the second task, by the runtime optimizer, a second SVD address indicating a second location within the shared memory partition associated with the first task;

determining, by the runtime optimizer, whether the second SVD address matches the first SVD address;

transmitting, by the runtime optimizer, a result indicating whether the second SVD address matches the first SVD address;

receiving from a plurality of threads, other results, each result indicating whether an SVD address associated with the SVD key in a particular remote address cache associated with a particular task matches the first SVD address;

if the results indicate that at least one of the SVD addresses does not match the first SVD address, updating, by the runtime optimizer, the remote address cache of the second task such that a previously sum-checked SVD address is used to correspond with the SVD key within the remote address cache of the second task; and

if the results indicate that all of the SVD addresses match the first SVD address, updating, by the runtime optimizer, the remote address cache of the second task such that the updated first SVD address is used to correspond with the SVD key within the remote address cache.

2. The method of claim 1 further comprising creating, by the runtime optimizer, an individual compare-and-swap operation message header for each entry of a second SVD.

3. The method of claim 2 wherein creating an individual compare-and-swap operation message header for each entry of the second SVD includes selecting a header type and data schedule for each individual compare-and-swap operation message header.

4. The method of claim 1 further comprising broadcasting to the plurality of threads, by the runtime optimizer, the compare-and-swap operation message header.

5. The method of claim 4 wherein broadcasting to the plurality of threads, by the runtime optimizer, the compare-and-swap operation message header includes transmitting an active message using a parallel active message interface (PAMI).

6. The method of claim 1 wherein updating, by the runtime optimizer, the remote address cache of the second task such that the previously sum-checked SVD address is used to correspond with the SVD key within the remote address cache of the second task comprises storing the previously sum-checked SVD address in the remote address cache of the second task at an entry corresponding to the SVD key.