# US 2006/0080398 A1 · Direct access of cache lock set data without backing memory

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

- Number: US 2006/0080398 A1
- Document ID: US20060080398A1
- Status: Published application
- Kind: A1
- Application number: 10/961,752
- Filed: 2004-10-08
- Published: 2006-04-13
- Earliest claimed: 2004-10-08
- Assignees: International Business Machines Corporation
- Inventors: Russell Hoover; Eric Mejdrich; Sandra Woodward
- Web page: https://patentgrep.com/patent/US20060080398A1

## CPC

- G06F 12/126: with special data handling, e.g. priority of data or instructions, handling errors or pinning
- G06F 12/084: with a shared cache
- G06F 2212/2515: being configurable for different purposes, e.g. as cache or non-cache memory
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Methods, apparatus, and systems for quickly accessing data residing in a cache of one processor, by another processor, while avoiding lengthy accesses to main memory are provided. A portion of the cache may be placed in a lock set mode by the processor in which it resides. While in the lock set mode, this portion of the cache may be accessed directly by another processor without lengthy “backing” writes of the accessed data to main memory.

## Claims

### Claim 1

A method of transferring data between multiple processors without writing the data to external memory, comprising:
allocating a plurality of cache lines of a cache residing on a first processor for use in a lock set;
storing, by the first processor, data in one or more cache lines allocated to the lock set;
providing, by the first processor, an indication of the availability of the data stored in the one or more cache lines; and
transferring data stored in the one or more cache lines from the first processor to the second processor without writing the data stored in the one or more cache lines to the external memory.

### Claim 2 (depends on 1)

The method of claim 1, wherein allocating the plurality of cache lines for use in the lock set comprises issuing a plurality of data cache block set to zero (DCBZ) instructions.

### Claim 3 (depends on 2)

The method of claim 2, further comprising placing the cache in a direct mapped mode prior to issuing the plurality of DCBZ instructions.

### Claim 4 (depends on 3)

The method of claim 3, further comprising placing the cache in a class set mode prior to storing the data in the one or more cache lines.

### Claim 5 (depends on 1)

The method of claim 1, wherein:
the external memory is mapped to a first address range; and
the cache lines of the lock set are mapped to a second address range.

### Claim 6 (depends on 1)

The method of claim 1, wherein providing, by the first processor, an indication of the availability of the data stored in the one or more cache lines comprises writing a value to a pointer residing on the second processor.

### Claim 7 (depends on 6)

The method of claim 6, wherein storing, by the first processor, data in one or more cache lines allocated to the lock set comprises storing data up to an address corresponding to a pointer, residing on the first processor, that is written to by the second processor.

### Claim 8

A method for transferring data generated by a central processing unit (CPU) to a graphics processing unit (GPU) without writing the generated data to external memory, comprising:
allocating a plurality of cache lines of a cache residing on the CPU for use in a lock set;
storing, by the CPU, data in one or more cache lines allocated to the lock set;
providing, by the CPU, an indication to the GPU of the availability of the data stored in the one or more cache lines;
receiving, by the CPU, a request from the GPU for data stored in one or more cache lines allocated to the lock set; and
without writing the data stored in the one or more cache lines to the external memory, transferring the data stored in the one or more cache lines from the CPU to the GPU and processing the data by the GPU.

### Claim 9 (depends on 8)

The method of claim 8, wherein providing, by the CPU, an indication to the GPU of the availability of the data stored in the one or more cache lines comprises writing, by the CPU, a value to a pointer on the GPU.

### Claim 10 (depends on 8)

The method of claim 8, wherein receiving, by the CPU, a request from the GPU for data stored in one or more cache lines allocated to the lock set comprises receiving a request to read with no intent to cache (RWNITC).

### Claim 11 (depends on 8)

The method of claim 8, wherein the memory controller is integrated with a graphics processing unit (GPU).

### Claim 12 (depends on 11)

The method of claim 11, further comprising receiving, by the memory controller, a request for data from a processing core of the GPU and routing the request to the CPU or external memory depending on an address supplied with the request.

### Claim 13 (depends on 12)

The method of claim 12, wherein routing the request to the CPU or external memory depending on the address supplied with the request comprises:
routing the request to the CPU if the address falls within a first address range reserved for the lock set; and
routing the request to the external memory if the address falls within a second address range reserved for the external memory.

### Claim 14

A processor comprising:
one or more processing cores; and
at least one shared cache accessible by the processing cores, the shared cache comprising a plurality of cache lines allocatable to a lock set, wherein data stored in cache lines allocated to the lock set is transferred directly to an external processor without writing the data to external memory.

### Claim 15 (depends on 14)

The processor of claim 14, wherein:
the shared cache comprises a plurality of associative sets; and
a portion or all of one or more of the associative sets may be allocated to the lock set.

### Claim 16 (depends on 14)

The processor of claim 14, further comprising a pointer containing a value indicating an address corresponding to the lock set, up to which the external processor has read data.

### Claim 17 (depends on 14)

The processor of claim 14, wherein the at least one processing core comprises a plurality of processing cores, each having an associated primary cache.

### Claim 18

A system, comprising:
main memory;
a memory controller for accessing the main memory;
a first processor capable of accessing the main memory via the memory controller, the first processor having at least one cache capable of being placed in a lock set mode wherein a plurality of cache lines of the cache are allocated to a lock set; and
a second processor capable of accessing data stored in one or more cache lines allocated to a lock set of the cache on the first processor without the data being written out to main memory.

### Claim 19 (depends on 18)

The system of claim 18, wherein:
a first range of addresses is allocated to the main memory; and
a second range of addresses is allocated to the lock set.

### Claim 20 (depends on 18)

The system of claim 18, wherein:
the first processor is a central processing unit (CPU) having one or more processor cores that share the cache;
the second processor is a graphics processing unit (GPU); and
the memory controller is integrated with the GPU.
