# US 2017/0256025 A1 · Efficient low-power texture cache architecture

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

- Number: US 2017/0256025 A1
- Document ID: US20170256025A1
- Status: Published application
- Kind: A1
- Application number: 15/420,463
- Filed: 2017-01-31
- Published: 2017-09-07
- Earliest claimed: 2016-03-04
- Assignees: Samsung Electronics Co., Ltd.
- Applicants: Samsung Electronics Co., Ltd.
- Inventors: Santosh ABRAHAM; Karthik RAMANI; Woong SEO; Kwontaek KWON; Jeongae PARK
- Web page: https://patentgrep.com/patent/US20170256025A1

## CPC

- G06T 1/60: Memory management
- G06F 12/0875: with dedicated cache, e.g. instruction or stack
- G06F 12/0897: with two or more cache hierarchy levels (with multilevel cache hierarchies G06F12/0811)
- G06T 15/005: {General purpose rendering architectures}
- G06T 15/04: Texture mapping
- H04N 19/426: using memory downsizing methods
- H04N 19/44: Decoders specially adapted therefor, e.g. video decoders which are asymmetric with respect to the encoder
- G06F 2212/401: Compressed data
- G06F 2212/455: Image or video data
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION
- 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

A texture cache architecture includes a first texture cache to store compressed texel data and a second texture cache to store decompressed texel data. A controller schedules accesses to access texel data from the first or second texture cache. The second texture cache permits decompressed texel data to be reused for more than one texel access request.

## Claims

### Claim 1

A graphics processing unit comprising:
a first texture cache configured to store compressed texel data;
a second texture cache configured to store texel data that has been decompressed from the first texture cache: and
a controller configured to:
receive requests for texel data for a group of pixels, and
schedule accesses to the first or second texture caches for the texel data.

### Claim 2 (depends on 2)

The graphics processing unit of claim 2, wherein the controller is further configured to:
determine whether there is a cache hit or a cache miss for the requested texel data in the second texture cache;
access the first texture cache for the requested texel data in response to a determination of a cache miss; and
access the second texture cache for the requested texel data in response to a determination of a cache hit.

### Claim 3 (depends on 1)

The graphics process unit of claim 1, wherein data is organized into the second texture cache based on locality patterns that are present in a set of accesses.

### Claim 4 (depends on 1)

The graphics processing unit of claim 1, wherein the second texture cache has texel data grouped into cache lines that are organized in Morton order that correspond to contiguous two-dimensional blocks of texels.

### Claim 5 (depends on 1)

The graphics processing unit of claim 1, wherein the controller is further configured to partition the set of texel addresses of the request into a sequence of at least one non-conflicting memory access.

### Claim 6 (depends on 5)

The graphics processing unit of claim 5, wherein the at least one non-conflicting memory access does not have tag conflicts or data bank conflicts.

### Claim 7 (depends on 1)

The graphics processing unit of claim 1, wherein the controller is further configured to combine texel requests that satisfy a set of constraints based on at least one of a number of distinct cache lines or a number of dwords per bank.

### Claim 8 (depends on 7)

The graphics processing unit of claim 7, wherein the controller is further configured to:
find cache-line addresses required for the texel requests originating from each pixel of the group of pixels; and
combine texel requests satisfying constraints of no more than four distinct cache lines and no more than one dword from each bank.

### Claim 9 (depends on 1)

The graphics processing unit of claim 1, wherein the second texture cache has 4-way banked tag lookup and 16-way banked data store.

### Claim 10 (depends on 1)

The graphics process unit of claim 1, wherein a layout of texels in the second texture cache is selected to ensure that quad texels in a texel footprint are on distinct banks.

### Claim 11

A method of operating a graphics processing unit comprising:
storing compressed texel data in a first texture cache;
storing, in a second texture cache, texel data that has been decompressed from the first texture cache;
receiving requests for texel data for a group of pixels, and
scheduling accesses to the first or second texture caches for the texel data.

### Claim 12 (depends on 11)

The method of claim 11, wherein the scheduling comprises:
determining whether there is a cache hit or a cache miss for the requested texel data in the second texture cache;
accessing the first texture cache for the requested texel data in response to a determination of a cache miss; and
accessing the second texture cache for the requested texel data in response to a determination of a cache hit.

### Claim 13 (depends on 12)

The method of claim 12, further comprising:
organizing the texel data in the second texture cache into tiles within which cache lines are organized in Morton order such that a contiguous two-dimensional block of texels require less than a predefined number of distinct cache lines and tag-lookups.

### Claim 14 (depends on 11)

The method of claim 11, further comprising partitioning the set of texel addresses of the request into a set of non-conflicting sets of accesses.

### Claim 15

The method of 14, wherein the non-conflicting sets of accesses do not have tag conflicts or data bank conflicts.

### Claim 16 (depends on 11)

The method of claim 11, further comprising combining texel requests that satisfy a set of constraints for a number based on at least one of a number of distinct cache lines or a number of dwords per bank.

### Claim 17 (depends on 16)

The method of claim 16, wherein the combining texel requests comprises combining texel requests satisfying constraints of no more than four distinct cache lines and no more than one dword from each bank.

### Claim 18 (depends on 11)

The method of claim 11, wherein the second texture cache has 4-way banked tag lookup and 16-way banked data store.

### Claim 19 (depends on 11)

The method of claim 11, wherein a layout of texels in the second texture cache is selected to ensure that quad texels in a texel footprint are on distinct banks.

### Claim 20 (depends on 11)

The method of claim 11, wherein data is organized into the second texture cache based on locality patterns that are present in a set of accesses.
