# US 2008/0307168 A1 · Latency Reduction for Cache Coherent Bus-Based Cache

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

- Number: US 2008/0307168 A1
- Document ID: US20080307168A1
- Status: Published application
- Kind: A1
- Application number: 11/758,219
- Filed: 2007-06-05
- Published: 2008-12-11
- Earliest claimed: 2007-06-05
- Assignees: Apple Inc.
- Inventors: Brian P. Lilly; Sridhar P. Subramanian; Ramesh Gunna
- Web page: https://patentgrep.com/patent/US20080307168A1

## CPC

- G06F 12/0831: using a bus scheme, e.g. with bus monitoring or watching means
- G06F 12/084: with a shared cache
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

In one embodiment, a system comprises a plurality of agents coupled to an interconnect and a cache coupled to the interconnect. The plurality of agents are configured to cache data. A first agent of the plurality of agents is configured to initiate a transaction on the interconnect by transmitting a memory request, and other agents of the plurality of agents are configured to snoop the memory request from the interconnect. The other agents provide a response in a response phase of the transaction on the interconnect. The cache is configured to detect a hit for the memory request and to provide data for the transaction to the first agent prior to the response phase and independent of the response.

## Claims

### Claim 1

A system comprising:
a plurality of agents configured to cache data, wherein the plurality of agents are coupled to an interconnect; and
a cache coupled to the interconnect;
wherein a first agent of the plurality of agents is configured to initiate a transaction on the interconnect by transmitting a memory request, and wherein other agents of the plurality of agents are configured to snoop the memory request from the interconnect and provide a response in a response phase of the transaction on the interconnect, and wherein the cache is configured to detect a hit for the memory request and to provide data for the transaction to the first agent prior to the response phase and independent of the response.

### Claim 2 (depends on 1)

The system as recited in claim 1 wherein, even if one or more of the other agents detect a hit for the memory request, the one or more other agents do not provide data if the cache detects the hit.

### Claim 3 (depends on 1)

The system as recited in claim 1 wherein the transaction results in a cache state in the first agent that would cause the first agent to provide data in response to a subsequent snoop, and wherein the cache is configured to invalidate the data from a cache entry storing the data in response to the transaction.

### Claim 4 (depends on 3)

The system as recited in claim 3 wherein the state comprises one of the exclusive, modified, or owned states.

### Claim 5 (depends on 1)

The system as recited in claim 1 wherein the plurality of agents, in response to evicting a cache block that is in an exclusive state but unmodified, is configured to generate a writeback transaction for the cache block.

### Claim 6 (depends on 5)

The system as recited in claim 5 wherein the cache is configured to allocate a cache entry in response to the writeback transaction and to accept the cache block into the entry.

### Claim 7 (depends on 1)

The system as recited in claim 1 further comprising a data arbiter configured to arbitrate requests for a data interconnect, and wherein the cache is coupled to the data arbiter and, in response to receiving the memory request from the interconnect, the cache is configured to speculatively arbitrate for the data interconnect.

### Claim 8 (depends on 7)

The system as recited in claim 7 wherein the data arbiter is configured to grant the data interconnect to the cache, and wherein the cache is configured to transmit a cancel indication on the data bus to indicate to the first agent that the speculation was incorrect and that the first agent is to disregard the data.

### Claim 9 (depends on 1)

The system as recited in claim 1 wherein the cache is configured to provide an early response to the first agent, prior to the response phase, to indicate that the cache will be providing the data.

### Claim 10 (depends on 1)

The system as recited in claim 1 wherein the first agent is a processor.

### Claim 11 (depends on 10)

The system as recited in claim 10 wherein at least one other agent of the plurality of agents is another processor.

### Claim 12 (depends on 1)

The system as recited in claim 1 wherein the cache is coupled to receive the memory request from the interconnect prior to the plurality of agents.

### Claim 13 (depends on 12)

The system as recited in claim 12 wherein at least some of the plurality of agents are coupled to receive the memory transaction through one or more staging flops.

### Claim 14

In a system comprising a plurality of agents coupled to an interconnect and a cache coupled to the interconnect, wherein the plurality of agents are also configured to cache data, a method comprising:
initiating a transaction by transmitting a memory request on the interconnect from a first agent of the plurality of agents;
snooping the memory request from the interconnect by other agents of the plurality of agents;
providing a response in a response phase of the transaction on the interconnect by the other agents;
detecting a hit for the memory request in the cache; and
providing data for the transaction to the first agent prior to the response phase and independent of the response, the data provided by the cache in response to detecting the hit.

### Claim 15 (depends on 14)

The method as recited in claim 14 wherein, even if one or more of the other agents detect a hit for the memory request, the one or more other agents do not provide data if the cache detects the hit.

### Claim 16 (depends on 14)

The method as recited in claim 14 further comprising:
storing the data for the transaction in the first agent;
storing a cache state in the first agent that would cause the first agent to provide data in response to a subsequent snoop in the first agent; and
invalidating the data from a cache entry storing the data in the cache in response to the memory transaction.

### Claim 17 (depends on 14)

The method as recited in claim 14 further comprising:
evicting a cache block that is in an exclusive state but unmodified from one of the plurality of agents; and
generating a writeback transaction for the cache block;
the cache allocating a cache entry in response to the writeback transaction; and
accepting the cache block into the entry.

### Claim 18 (depends on 14)

The method as recited in claim 14 further comprising the cache speculatively requesting a data interconnect in response to receiving the memory request from the interconnect prior to detecting the hit.

### Claim 19 (depends on 14)

The method as recited in claim 14 further comprising providing an early response from the cache to the first agent, prior to the response phase, to indicate that the cache will be providing the data.

### Claim 20

A system comprising:
a plurality of agents configured to cache data, wherein the plurality of agents are coupled to an interconnect; and
a cache coupled to the interconnect;
wherein the cache and the plurality of agents are configured to maintain coherence states such that, if the cache detects a hit for a memory request transmitted on the interconnect, the cache is able to provide the data regardless of the state of the data in the plurality of agents, and wherein the cache is configured to provide data prior to the response phase corresponding to the memory request is the hit is detected.

### Claim 21 (depends on 20)

The system as recited in claim 20 wherein, if a resulting state in one of the plurality of agents in response to the memory request permits the one of the plurality of agents to modify the data, the cache is configured to invalidate the data in the cache.
