Packet switched cache coherent multiprocessor system

• US 5,634,068 A
• Filed: 03/31/1995
• Issued: 05/27/1997
• Est. Priority Date: 03/31/1995
• Status: Expired due to Fees

• Current Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Nishtala, Satyanarayana, Coffin, Louis F. III, Kohn, Leslie, Normoyle, Kevin, Van Loo, William C., Ebrahim, Zahir
• Primary Examiner(s): SHAH, ALPESH
• Application Number:
  US08/415,175
• Time in Patent Office: 788 Days
• Field of Search: 395/800, 395/775, 395/200.08, 395/200.15, 395/827, 395/840-841, 395/851, 395/853, 395/468-472, 395/474, 395/650, 364/131-134, 340/825.03, 340/826, 340/825.06, 340/825.5, 370/54, 370/60, 370/85.1, 370/94.1
• US Class Current: 711/141
• CPC Class Codes: G06F 12/0822 Copy directories local copy tags for implementing a bus snooping protocol G06F12/0831

Abstract

A multiprocessor computer system has a multiplicity of sub-systems and a main memory coupled to a system controller. An interconnect module, interconnects the main memory and sub-systems in accordance with interconnect control signals received from the system controller. All of the sub-systems include a port that transmits and receives data as data packets of a fixed size. At least two of the sub-systems are data processors, each having a respective cache memory and a respective set of master cache tags (Etags), including one cache tag for each data block stored by the cache memory. The system controller maintains a set of duplicate cache tags (Dtags) for each of the data processors. The data processors each include master cache logic for updating the master cache tags, while the system controller includes logic for updating the duplicate cache tags. Memory transaction request logic simultaneously looks up the second cache tag in each of the sets of duplicate cache tags corresponding to the memory transaction request. It then determines which one of the cache memories and main memory to couple to the requesting data processor based on the second cache states and the address tags stored in the corresponding second cache tags. Duplicate cache update logic simultaneously updates all of the corresponding second cache tags in accordance with predefined cache tag update criteria.

Claims

1. A computer system, comprising:

a system controller;

a multiplicity of sub-systems coupled to the system controller;

a main memory coupled to said system controller; and

a datapath, coupled to said system controller, interconnecting said main memory and said sub-systems in accordance with interconnect control signals received from said system controller;

a plurality of said sub-systems comprising data processors, at least one of said data processors having a respective cache memory that stores multiple blocks of data and a set of master cache tags (Etags), including one Etag for each data block stored by said cache memory;

at least one of said sub-systems including a port that transmits and receives data as data packets of a fixed size equal in size to said each data block;

said datapath and each said port having a datapath width smaller than said each data block;

said at least one of said data processors including a master interface, coupled to said system controller, for sending memory transaction requests to said system controller and for receiving cache access requests from said system controller corresponding to memory transaction requests by other ones of said data processors;

said system controller including memory transaction request logic for processing each said memory transaction request by a requesting one of said data processors, for determining which one of said cache memories and main memory to couple to the requesting data processor, for sending corresponding interconnect control signals to said datapath so as to couple the requesting data processor to said determined one of said cache memories and main memory, and for sending a reply message to said requesting data processor to prompt said requesting data processor to transmit/receive one data packet to/from said determined one of said cache memories and main memory.

2. The computer system of claim 1,

said system controller including a set of duplicate cache tags (Dtags) for each of said data processors, said set of duplicate cache tags for each data processor having an equal number of Dtags as the corresponding set of master cache tags;

each Etag denoting a Etag cache state and an address tag;

the Dtag corresponding to each Etag denoting a Dtag state and the same address tag as the corresponding Etag;

said data processors each including master cache logic for updating said Etags;

said system controller including Dtag cache logic for updating said Dtags so as to store in said Dtags Dtag states and address tags corresponding to said Etag states and address tags stored in said Etags; and

said memory transaction request logic including Dtag inspection logic for simultaneously looking up the Dtag in each of said duplicate cache indices corresponding to said each memory transaction request and for determining which one of said cache memories and main memory to couple to the requesting data processor based on said Dtag states and said address tags stored in said corresponding Dtags, and Dtag update logic for simultaneously updating all of said corresponding Dtags in accordance with predefined cache tag update criteria.

3. The computer system of claim 2, wherein

said memory transaction requests are transmitted from said data processors to said system controller over at least one request bus distinct from said datapath such that while a data packet for a first memory transaction request is being transmitted via said datapath, other memory transaction requests can be transmitted via said at least one request bus to said system controller.

4. The computer system of claim 2, wherein

said memory transaction requests are transmitted from said data processors to said system controller over a plurality of request busses distinct from said datapath such that while a data packet for a first memory transaction request is being transmitted via said datapath, other memory transaction requests can be transmitted via said request busses to said system controller.

5. The computer system of claim 2,

said Etag state being selected from the set of states consisting essentially of:

Exclusive Modified (M), Shared Modified (O), Exclusive Clean (E), Shared Clean (S), and Invalid (I);

said Dtag state being selected from the set of states consisting essentially of;

Exclusive Modified (M), Shared Modified (O), Shared Clean (S), and Invalid (I); and

wherein said Dtag state stored in said Dtags never indicates said Exclusive Clean (E) state and when each data processor modifies data stored in its cache memory in a cache line whose Etag thereby transitions from said E state to said M state, said data processor does not generate a corresponding transaction request and the corresponding Dtag remains unchanged with a Dtag state equal to said M state.

6. The computer system of claim 2,

wherein said main memory is a reflective memory;

said Etag state being selected from the set of states consisting essentially of;

Exclusive Modified (M), Exclusive Clean (E), Shared Clean (S), and Invalid (I);

said Dtag state being selected from the set of states consisting essentially of;

Exclusive Modified (M), Shared Clean (S), and Invalid (I); and

wherein said Dtag state stored in said Dtags never indicates said Exclusive Clean (E) state and when each data processor modifies data stored in its cache memory in a cache line whose Etag thereby transitions from said E state to said M state, said data processor does not generate a corresponding transaction request and the corresponding Dtag remains unchanged with a Dtag state equal to said M state.

7. A method for operating a cache coherent multiprocessor system having a system controller coupled to a main memory and to a plurality of data processors at least one of which has a cache memory, comprising the steps of:

storing master cache tags (Etags) in said at least one of said data processors, including one Etag for each cache line in said cache memory, said Etag for each cache line storing an address index and an Etag state value that indicates whether a data block stored in said cache line includes data modified its respective data processor;

sending memory transaction requests from said data processors to said system controller, said memory transaction requests including read requests and write requests;

each memory transaction request specifying an address for an associated data block to be read or written;

at said at least one of said data processors, receiving cache access requests from said system controller corresponding to memory transaction requests by other ones of said data processors;

providing a datapath that interconnects said main memory and said sub-systems in accordance with interconnect control signals from said system controller;

at each of said data processors, transmitting and receiving data as data packets of a fixed size equal in size to said each data block;

said datapath and each said port having a datapath width smaller than said each data block; and

at said system controller, while processing a memory transaction request from a requesting one of said data processors, determining which one of said cache memories and main memory to couple to the requesting data processor, sending corresponding interconnect control signals to said datapath so as to couple the requesting data processor to said determined one of said cache memories and main memory, and sending a reply message to said requesting data processor to prompt said requesting data processor to transmit/receive one data packet to/from said determined one of said cache memories and main memory.

8. The method of claim 7,

storing in said system controller a set of duplicate cache tags (Dtags) for each said cache memory, each Dtag corresponding to one of said Etags and including a Dtag state value and the same address index as the corresponding Etag;

said Dtag state value indicating whether said data block stored in the corresponding cache line includes data modified by a respective data processor;

at each data processor, storing in each Etag a Etag cache state and an address tag; and

in said system controller;

storing in each Dtag a Dtag state and the same address tag as the corresponding Etag; and

while processing each memory transaction request from a requesting one of said data processors, simultaneously looking up the Dtag in each of said duplicate cache indices corresponding to said each memory transaction request and determining which one of said cache memories and main memory to couple to the requesting data processor based on said Dtag states and said address tags stored in said corresponding Dtags, and then simultaneously updating all of said corresponding Dtags in accordance with predefined cache tag update criteria.

10. The method of claim 8,

said Etag state being selected from the set of states consisting essentially of:

Exclusive Modified (M), Shared Modified (O), Exclusive Clean (E), Shared Clean (S), and Invalid (I);

said Dtag state being selected from the set of states consisting essentially of;

Exclusive Modified (M), Shared Modified (O), Shared Clean (S), and Invalid (I); and

wherein said Dtag state stored in said Dtags never indicates said Exclusive Clean (E) state and when each data processor modifies data stored in its cache memory in a cache line whose Etag thereby transitions from said E state to said M state, said data processor does not generate a corresponding transaction request and the corresponding Dtag remains unchanged with a Dtag state equal to said M state.

11. The method of claim 8,

wherein said main memory is a reflective memory;

said Etag state being selected from the set of states consisting essentially of;

Exclusive Modified (M), Exclusive Clean (E), Shared Clean (S), and Invalid (I);

said Dtag state being selected from the set of states consisting essentially of;

Exclusive Modified (M), Shared Clean (S), and Invalid (I); and

wherein said Dtag state stored in said Dtags never indicates said Exclusive Clean (E) state and when each data processor modifies data stored in its cache memory in a cache line whose Etag thereby transitions from said E state to said M state, said data processor does not generate a corresponding transaction request and the corresponding Dtag remains unchanged with a Dtag state equal to said M state.

9. The method of 8,

transmitting said memory transaction requests from said data processors to said system controller over at least one request bus distinct from said datapath such that while a data packet for a first memory transaction request is being transmitted via said datapath, other memory transaction requests are transmitted via said at least one request bus to said system controller.