# US 2017/0255554 A1 · Cache memory and operation method thereof

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

- Number: US 2017/0255554 A1
- Document ID: US20170255554A1
- Status: Published application
- Kind: A1
- Application number: 15/241,902
- Filed: 2016-08-19
- Published: 2017-09-07
- Earliest claimed: 2016-03-02
- Assignees: Electronics and Telecommunications Research Institute
- Applicants: Electronics and Telecommunications Research Institute
- Inventors: Jin Ho HAN; Young-Su KWON; Kyung Jin BYUN; Nak Woong EUM
- Web page: https://patentgrep.com/patent/US20170255554A1

## CPC

- G06F 12/0811: with multilevel cache hierarchies
- G06F 12/084: with a shared cache
- G06F 12/0842: for multiprocessing or multitasking
- G06F 12/0875: with dedicated cache, e.g. instruction or stack
- G06F 2212/6042: Allocation of cache space to multiple users or processors
- G06F 2212/62: Details of cache specific to multiprocessor cache arrangements
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

Provided is a cache memory. The cache memory includes a first to Nth level-1 caches configured to correspond to first to Nth cores, respectively, a level-2 sharing cache configured to be shared by the first to Nth level-1 caches, and a coherence controller configured to receive an address from each of the first to Nth cores and allocate at least a partial area in an area of the level-2 sharing cache to one of the first to Nth level-1 caches based on the received address.

## Claims

### Claim 1

A cache memory comprising:
first to Nth level-1 caches configured to correspond to first to Nth cores, respectively (where N is a natural number greater than 1);
a level-2 sharing cache configured to be shared by the first to Nth level-1 caches; and
a coherence controller configured to receive an address from each of the first to Nth cores and allocate at least a partial area of an area of the level-2 sharing cache to one of the first to Nth level-1 caches based on the received address.

### Claim 2 (depends on 1)

The cache memory of claim 1, wherein at least the partial area corresponds to a data unit accessed by each of the first to Nth cores.

### Claim 3 (depends on 1)

The cache memory of claim 1, wherein the coherence controller comprises:
a storage circuit configured to store map data; and
first and Nth map controllers configured to receive the address from each of the first to Nth cores, allow the first to Nth level-1 caches to access data corresponding to the received address, or allocate at least a portion of the level-2 sharing cache to one of the plurality of level-1 caches.

### Claim 4 (depends on 3)

The cache memory of claim 3, wherein the first to Nth map controllers are configured to communicate with the plurality of level-1 caches through a sharing bus.

### Claim 5 (depends on 3)

The cache memory of claim 3, wherein the first map controller among the first to Nth map controllers receives the address from the first core among the first to Nth cores, and when information corresponding to at least a portion of the received address is not included in the map data, allocates an area where page data corresponding to the received address is stored in an area of the level-2 sharing cache to the first level-1 cache.

### Claim 6 (depends on 5)

The cache memory of claim 5, wherein the first map controller stores information on an ID of the first level-1 cache and an ID of the page data in the map data.

### Claim 7 (depends on 3)

The cache memory of claim 3, wherein the first map controller among the first to Nth map controllers comprises:
a map reader configured to receive the address from the first core and determine whether at least the portion of the received address is included in the map data; and
a page allocation unit, when at least the portion of the received address is not included in the map data, configured to allocate an area where page data corresponding to the receive address in an area of the level-2 sharing cache to one of the first to Nth level-1 caches on the basis of an allocation information table.

### Claim 8 (depends on 7)

The cache memory of claim 7, wherein the coherence controller further comprises an allocation information table comprising page allocation information allocated to each of the first to Nth level-1 caches,
wherein the page allocation unit is configured to allocate an area where the page data is stored to one of the first to Nth level-1 caches on the basis of the allocation information table.

### Claim 9 (depends on 8)

The cache memory of claim 8, wherein when page allocation information on the first level-1 cache is less than a threshold, the page allocation unit allocates an area where the page data is stored to the first level-1 cache, and when the page allocation information on the first level-1 cache is greater than the threshold, allocates the area where the page data is stored to a level-1 cache having the smallest page allocation information among the second to Nth level-1 caches.

### Claim 10 (depends on 1)

The cache memory of claim 1, wherein each of the first to Nth level-1 caches comprises:
an instruction cache configured to store an instruction used in each of the first to Nth cores; and
a data cache configured to temporarily store data used in the first to Nth cores.

### Claim 11 (depends on 1)

The cache memory of claim 1, wherein each of the first to Nth level-1 caches operates based on a clock frequency different from that of the level-2 sharing cache, and each of the first to Nth level-1 caches communicates with the level-2 sharing cache through a bus having a clock domain crossing (CDC) function.

### Claim 12 (depends on 1)

The cache memory of claim 1, further comprising a cache eraser configured to receive a cache clear signal from at least one of the first to Nth cores, and erase the first to Nth level-1 caches and the level-2 sharing cache sequentially in response to the received cache clear signal, wherein the cache eraser erases the first to Nth level-1 caches and the level-2 sharing cache sequentially on the basis of at least one among a plurality of address sets.

### Claim 13

An operation method of a cache memory comprising a plurality of level-1 caches and a level-2 sharing cache, the method comprising:
receiving an address from an external core;
determining whether at least a portion of the received address is included in map data;
if at least the portion of the received address is not included in the map data, allocating an area where page data corresponding to the received address in an area of the level-2 sharing cache to one of the plurality of level-1 caches;
storing a cache ID of a level-1 cache where the area is allocated among the plurality of level-2 caches and at least a portion of the received address in the map data; and
performing an access operation on one of the plurality of level-1 caches on the basis of the map data.

### Claim 14 (depends on 13)

The method of claim 13, wherein if at least the portion of the received address is not included in the map data, the allocating of the area where the page data corresponding to the received address in the area of the level-2 sharing cache to the one of the plurality of level-1 caches comprises allocating the area where the page data is stored to one of the plurality of level-1 caches on the basis of page allocation information comprising information on the number of pages allocated to the plurality of level-1 caches.

### Claim 15 (depends on 14)

The method of claim 14, wherein the allocating of the area where the page data is stored to the one of the plurality of level-1 caches on the basis of the page allocation information comprising the information on the number of pages allocated to the plurality of level-1 caches comprises:
when allocation information on a level-1 cache corresponding to the external core among the plurality of level-1 caches is less than a threshold, allocating the area where the page data is stored to a level-1 cache corresponding to the external core; and
when the allocation information on the level-1 cache corresponding to the external core among the plurality of level-1 caches is greater than the threshold, allocating the area where the page data is stored to one of the remaining level-1 caches.

### Claim 16 (depends on 15)

The method of claim 15, wherein when the allocation information on the level-1 cache corresponding to the external core among the plurality of level-1 caches is greater than the threshold, the allocating of the area where the page data is stored to the one of the remaining level-1 caches comprises allocating the area where the page data is stored to a level-1 cache having the smallest allocation information among the remaining level-1 caches.
