Common memory programming

• US 8,438,341 B2
• Filed: 06/16/2010
• Issued: 05/07/2013
• Est. Priority Date: 06/16/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Finkler, Ulrich A., Hirsch, Steven N., Reindel, Harold E.
• Primary Examiner(s): RUTZ, JARED IAN
• Application Number:
  US12/816,588
• Publication Number:
  US 20110314238A1
• Time in Patent Office: 1,056 Days
• Field of Search: 711/152, 711/203, 719/312
• US Class Current: 711/152
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes

Abstract

A method for unidirectional communication between tasks includes providing a first task having access to an amount of virtual memory, blocking a communication channel portion of said first task'"'"'s virtual memory, such that the first task cannot access said portion, providing a second task, having access to an amount of virtual memory equivalent to the first task'"'"'s virtual memory, wherein a communication channel portion of the second task'"'"'s virtual memory corresponding to the blocked portion of the first task'"'"'s virtual memory is marked as writable, transferring the communication channel memory of the second task to the first task, and unblocking the communication channel memory of the first task.

Claims

1. A method for unidirectional communication between tasks, comprising:

blocking a communication channel portion of a first virtual memory of a first task, such that the first task cannot access said portion;

providing a second task, having access to a second amount of virtual memory with the same virtual address as the first virtual memory of the first task and having associated virtual table entries, wherein a communication channel portion of the second virtual memory corresponding to the blocked portion of the first virtual memory is usable;

mapping virtual table entries in the communication channel to physical memory;

transferring virtual table entries of the communication channel memory of the second task to the first task; and

unblocking the communication channel memory of the first task responsive to said transfer of virtual table entries.

2. The method of claim 1, wherein the second task is provided by cloning the first task.

3. The method of claim 1, wherein the second task is provided by negotiating to find a common unmapped virtual memory region.

4. The method of claim 1, further comprising modifying the communication channel memory of the second task before transferring.

5. The method of claim 4, wherein said modification maps physical memory with data content into the virtual memory region of the communication channel.

6. The method of claim 1, further comprising the step of mapping an unblocked portion of the first task'"'"'s virtual memory to physical memory.

7. The method of claim 6, wherein said step of providing a second task includes mapping an unblocked portion of the second task'"'"'s virtual memory to physical memory.

8. The method of claim 7, further comprising the step of modifying a portion of physical memory corresponding to at least one of said mappings.

9. The method of claim 8, further comprising the steps of:

terminating the second task; and

releasing all physical memory which was mapped exclusively by the second task.

10. The method of claim 1, wherein unblocking includes establishing regular copy-on-write semantics for a mapped part of the communication channel.

12. The computer readable storage medium of claim 10, wherein the second task is provided by cloning the first task.

13. The computer readable storage medium of claim 10, further comprising the step of modifying the communication channel memory of the second task before said step of transferring.

11. A non-transitory computer readable storage medium comprising a computer readable program, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

blocking a communication channel portion of a first virtual memory of a first task, such that the first task cannot access said portion;

providing a second task, having access to a second amount of virtual memory with the same virtual address as the first virtual memory of the first task and having associated virtual table entries, wherein a communication channel portion of the second virtual memory corresponding to the blocked portion of the first virtual memory is usable;

mapping virtual table entries in the communication channel to physical memory;

transferring virtual table entries of the communication channel memory of the second task to the first task; and

unblocking the communication channel memory of the first task responsive to said transfer of virtual table entries.

14. A system for parallel processing, comprising:

a physical memory;

a plurality of processes that are executed by one or more processors, including a first process and a second process, each process having a virtual memory that maps to the physical memory; and

a communication channel that establishes a temporary, unidirectional pathway for the transfer of pointer-based structures between the respective virtual memories of the first and second processes, wherein a writing process transfers virtual table entries of the communication channel to a receiving process and wherein the receiving process is blocked from accessing the communication channel until said virtual table entries are transferred.

15. The system of claim 14, wherein the communication channel does not maintain cache coherence.

16. The system of claim 14, wherein a portion of the virtual memory of the first task corresponding to the communication channel is not usable by the first task while the communication channel is active.

17. The system of claim 16, wherein said portion of virtual memory of the first task becomes usable by the first task after the communication channel closes.

18. The system of claim 14, wherein each process stores information that it alone references in physical memory that is not shared by any other process.

19. The system of claim 14, further comprising:

a message queue associated with one or more of said processes, the message queue accepting messages from one or more other processes; and

at least one pipe, allowing transfer of data between two processes.

20. The system of claim 14, wherein the second process is a clone of the first task.