Extended fast memory access in a multiprocessor computer system

• US 9,069,672 B2
• Filed: 06/12/2009
• Issued: 06/30/2015
• Est. Priority Date: 06/12/2009
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Abts, Dennis C., Alverson, Robert, Froese, Edwin, Scott, Steven L., Pritchard, Howard
• Primary Examiner(s): Tiv, Backhean
• Application Number:
  US12/483,936
• Publication Number:
  US 20100318626A1
• Time in Patent Office: 2,209 Days
• Field of Search: 709/216, 709/213, 709/203, 711/6, 711/200, 711/202, 711/203
• US Class Current: 1/1
• CPC Class Codes: G06F 12/0813 with a network or matrix configuration
• CPC Class Codes: G06F 12/1009 using page tables, e.g. page table structures
• CPC Class Codes: G06F 12/109 for multiple virtual address spaces, e.g. segmentation G06F12/1036 takes precedence
• CPC Class Codes: G06F 13/4009 with data restructuring
• CPC Class Codes: G06F 15/167 using a common memory, e.g. mailbox
• CPC Class Codes: G06F 2212/163 Server or database system
• CPC Class Codes: G06F 9/3834 Maintaining memory consistency
• CPC Class Codes: H04L 67/00 Network arrangements or protocols for supporting network services or applications user-to-user messaging H04L51/00; network arrangements, protocols or services for supporting real-time applications in data packet communications networks H04L65/00

Abstract

A multiprocessor computer system comprises a first node operable to access memory local to a remote node by receiving a virtual memory address from a requesting entity in node logic in the first node. The first node creates a network address from the virtual address received in the node logic, where the network address is in a larger address space than the virtual memory address, and sends a fast memory access request from the first node to a network node identified in the network address.

Claims

1. A method of a first node accessing memory in a remote node in a multiprocessor computer network, wherein the remote node includes one or more processing elements, the method comprising:

receiving a virtual memory address from a requesting entity in the first node;

creating a network address from the virtual address received in the first node, where the network address is in a larger address space than the virtual memory address, and the network address includes a processing element field, a memory domain handle, and an offset; and

sending a fast memory access request from the first node to the remote node identified in the network address, wherein the fast memory access request includes the network address;

wherein the memory domain handle is associated with the memory in the remote node, and the processing element field is associated with the remote node.

2. The method of accessing memory in a remote node in a multiprocessor computer network of claim 1, further comprising receiving in the first node a fast memory access reply comprising the result of at least one PUT or GET operation from the remote node identified in the network address.

3. The method of accessing memory in a remote node in a multiprocessor computer network of claim 1, wherein the requesting entity is a processor local to the first node.

4. The method of accessing memory in a remote node in a multiprocessor computer network of claim 1, wherein the virtual memory address comprises:

a fast memory access (FMA) descriptor; and

an offset window used to generate network endpoint and offset components of the network address.

5. The method of accessing memory in a remote node in a multiprocessor computer network of claim 1, wherein the network logical address comprises:

a protection tag.

6. The method of accessing memory in a remote node in a multiprocessor computer network of claim 1, wherein fast memory access is used to provide load/store access to remote memory, and a block transfer engine is used to copy remote memory.

7. The method of accessing memory in a remote node in a multiprocessor computer network of claim 1, wherein creating a network address from the virtual address comprises looking up at least a portion of the virtual address in a fast memory access table.

8. A multiprocessor computer system, comprising a first node, including one or more processing elements, operable to access memory local to a remote node by:

receiving a virtual memory address from a requesting entity in the first node;

creating a network address from the virtual address received in the first node, where the network address is in a larger address space than the virtual memory address, and the network address includes a processing element field, a memory domain handle, and an offset; and

sending a fast memory access request from the first node to the remote node identified in the network address, wherein the fast memory access request includes the network address;

wherein the memory domain handle is associated with the memory in the remote node, and the processing element field is associated with the remote node.

9. The multiprocessor computer system of claim 8, the first node further operable to receive a fast memory access reply comprising the result of at least one PUT or GET operation from the remote node identified in the network logical address.

10. The multiprocessor computer system of claim 8, wherein the requesting entity is a processor local to the first node.

11. The multiprocessor computer system of claim 8, wherein the physical memory address comprises:

a fast memory access (FMA) descriptor; and

an offset window used to generate network endpoint and offset components of the network address.

12. The multiprocessor computer system claim 8, wherein the network address comprises:

a protection tag.

13. The multiprocessor computer system of claim 8, wherein fast memory access is used to provide load/store access to remote memory, and a block transfer engine is used to copy remote memory.

14. The multiprocessor computer system of claim 8, wherein creating a network address from the virtual address comprises looking up at least a portion of the virtual address in a fast memory access table.

15. A multiprocessor computer node operable to access memory local to a remote node by:

receiving a virtual memory address from a requesting entity in the multiprocessor computer node;

creating a network address from the virtual address received, where the network address is in a larger address space than the virtual memory address, and the network address includes a processing element field, a memory domain handle, and an offset; and

sending a fast memory access request from the multiprocessor computer node to the remote node identified in the network address, wherein the fast memory access request includes the network address;

wherein the memory domain handle is associated with the memory in the remote node, and the processing element field is associated with the remote node.

16. The multiprocessor computer node of claim 15, the first node further operable to receive a fast memory access reply comprising the result of at least one put or get operation from the remote node identified in the network logical address.

17. The multiprocessor computer node of claim 15, wherein the requesting entity is a processor local to the first node.

18. The multiprocessor computer node of claim 15, wherein the physical memory address comprises:

a fast memory access (FMA) descriptor; and

an offset window used to generate network endpoint and offset components of the network address.

19. The multiprocessor computer node claim 15, wherein the network logical address comprises:

a protection tag.

20. The multiprocessor computer node of claim 15, wherein fast memory access is used to provide load/store access to remote memory, and a block transfer engine is used to copy remote memory.

21. The multiprocessor computer node of claim 15, wherein creating a network address from the virtual address comprises looking up at least a portion of the virtual address in a fast memory access table.