# US 2006/0053257 A1 · Resolving multi-core shared cache access conflicts

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2006/0053257 A1
- Document ID: US20060053257A1
- Status: Published application
- Kind: A1
- Application number: 11/045,525
- Filed: 2005-01-28
- Published: 2006-03-09
- Earliest claimed: 2004-09-09
- Assignees: Intel Corporation
- Inventors: Krishnakanth Sistla; Yen-Cheng Liu; George Cai; Jeffrey Gilbert
- Web page: https://patentgrep.com/patent/US20060053257A1

## CPC

- G06F 12/084: with a shared cache
- G06F 12/0815: Cache consistency protocols
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Preventing request conflicts within microprocessors and/or computer systems. More particularly, embodiments of the invention relate to a technique to manage request conflicts within a processor and/or computer system in which a number of accesses may be made to a particular cache or group of caches shared amongst a set of cores or processors or agents.

## Claims

### Claim 1

An apparatus comprising:
ordering logic to prevent cache access conflicts between a read access to a first line of cache and a writeback access to the first line of cache.

### Claim 2 (depends on 1)

The apparatus of claim 1 wherein the line of cache is within a shared inclusive cache memory.

### Claim 3 (depends on 1)

The apparatus of claim 1 wherein the read access to the first line of cache is a cross snoop access from a shared inclusive cache memory to a core cache memory.

### Claim 4 (depends on 1)

The apparatus of claim 1 wherein a cache line fill to the first line of cache causes an eviction of a line of cache within a shared inclusive cache memory.

### Claim 5 (depends on 1)

The apparatus of claim 1 wherein the read access is from a first core in a multi-core processor and the writeback access is from a second core in the multi-core processor.

### Claim 6 (depends on 1)

The apparatus of claim 1 wherein the read access is from a first processor in a multi-processor system and the writeback is from a first core in a multi-core processor.

### Claim 7 (depends on 1)

The apparatus of claim 1 further comprising a storage unit to temporarily store coherency state information pertaining to the first line of cache in response to the read access.

### Claim 8 (depends on 7)

The apparatus of claim 7 wherein the coherency state information includes at least one bit to indicate that the first line of cache is atomically invalidated as a result of the read access.

### Claim 9

A system comprising:
a first processor comprising a shared inclusive cache, the shared inclusive cache including a first cache line having an invalid state in response to a snoop from another core within the first processor, the invalid state to indicate to all transactions subsequent to the read access, but before the read access is complete, that the first cache line is invalid.

### Claim 10 (depends on 9)

The system of claim 9 comprising a storage unit to store coherency information of the first cache line in response to the snoop.

### Claim 11 (depends on 10)

The system of claim 10 comprising a plurality of processor cores each having at least one corresponding core cache.

### Claim 12 (depends on 11)

The system of claim 11 wherein the shared inclusive cache is a last level cache to store the same data stored within the at least one corresponding core cache.

### Claim 13 (depends on 12)

The system of claim 12 wherein the first processor comprises logic to set the invalid state and to store the coherency information in response to the snoop.

### Claim 14 (depends on 13)

The system of claim 13 wherein the first processor comprises an internal and external request queue to, respectively, store accesses from and to the plurality of processor cores and a second processor.

### Claim 15 (depends on 14)

The system of claim 14 wherein the first and second processors are coupled together via a point-to-point interconnect.

### Claim 16 (depends on 14)

The system of claim 14 wherein the first and second processors are coupled together via a front-side bus interconnect.

### Claim 17

A method comprising:
performing a plurality of accesses to the same line of an inclusive cache structure;
preventing conflicts among the plurality of accesses to the line from a plurality of processor cores, the plurality of accesses being caused by any two transactions of a group consisting of: an external snoop of the line, an eviction of the line, a cross snoop from the inclusive cache, and a writeback to the line.

### Claim 18 (depends on 17)

The method of claim 17 wherein the preventing comprises storing coherency information of the line in response to one of the plurality of accesses.

### Claim 19 (depends on 18)

The method of claim 18 wherein the preventing further comprises atomically invalidating the line in response to one of the plurality of accesses.

### Claim 20 (depends on 19)

The method of claim 19 further comprising restoring the coherency information of the line in response to the one of the plurality of accesses being completed.

### Claim 21 (depends on 20)

The method of claim 20 further comprising setting the line to a valid state in response to the one of the plurality of accesses being completed.

### Claim 22

A processor comprising:
means for preventing cache access conflicts between a read access to a first line of cache and a writeback access to the first line of cache.

### Claim 23 (depends on 22)

The processor of claim 22 wherein the line of cache is within a shared inclusive cache memory.

### Claim 24 (depends on 22)

The processor of claim 22 wherein the read access to the first line of cache is a cross snoop access from a shared inclusive cache memory to a core cache memory.

### Claim 25 (depends on 22)

The processor of claim 22 wherein a cache line fill to the first line of cache causes an eviction of a line of cache within a shared inclusive cache memory.

### Claim 26 (depends on 22)

The processor of claim 22 wherein the read access is from a first core in a multi-core processor and the writeback access is from a second core in the multi-core processor.

### Claim 27 (depends on 22)

The processor of claim 22 wherein the read access is from a first processor in a multi-processor system and the writeback is from a first core in a multi-core processor.

### Claim 28 (depends on 22)

The processor of claim 22 further comprising a storage unit to temporarily store coherency state information pertaining to the first line of cache in response to the read access.

### Claim 29 (depends on 28)

The processor of claim 28 wherein the coherency state information includes at least one bit to indicate that the first line of cache is atomically invalidated as a result of the read access.
