Method and apparatus for data processing using graphic processing unit

• US 9,355,428 B2
• Filed: 07/12/2013
• Issued: 05/31/2016
• Est. Priority Date: 07/18/2012
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Lee, Changsung, Lee, Ilho, Ahn, Youngwoo, Yeo, Inchoon
• Primary Examiner(s): Crawford, Jacinta M
• Application Number:
  US13/940,947
• Publication Number:
  US 20140022268A1
• Time in Patent Office: 1,054 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining
• CPC Class Codes: G06T 1/60 Memory management

Abstract

A method of processing data using a Graphic Processing Unit (GPU) is provided. The method includes obtaining operand data and operator information to be applied to the operand data, partitioning the operand data by a channel element size unit of a texel of the GPU, transmitting an instruction which controls the GPU to return an operation result by performing an operation for the partitioned data and transmitting the partitioned data, and obtaining an operation result value from the GPU.

Claims

1. A method of processing data using a Graphic Processing Unit (GPU), the method comprising:

obtaining, by a data handler, operand data and operator information to be applied to the operand data;

partitioning, by the data handler, the operand data by a channel element size unit of a texel of the GPU;

transmitting, by a frame buffer manager, an instruction which controls the GPU to return an operation result by performing an operation for the partitioned data and transmitting the partitioned data; and

obtaining, by the frame buffer manager, an operation result value from the GPU,

wherein the operation result value is a first operation result value when the GPU is operating in a first mode and a second operation result value different from the first operation result value when the GPU is operating in a second mode, and

wherein the instruction controls the GPU to restore the partitioned data into an original operand data, perform an operation for the restored operand data, and store an operation result into a frame buffer of the GPU.

2. The method of claim 1, further comprising:

determining a size of a frame buffer of the GPU to which the operation result is returned according to a number of a row of the operand data, a number of a column, and a kind of an operation to be applied to the operand data,

wherein the operand data is a matrix.

3. The method of claim 1, further comprising:

determining a size of a frame buffer of the GPU to which the operation result is returned according to a number of a row of a first operand data and a number of a column of a second operand data among the operand data.

4. The method of claim 1, wherein the partitioning of the operand data comprises assigning two or more operand data to one texel, when a size of the operand data is two times a size of the channel element of the texel.

5. The method of claim 1, wherein the partitioning of the operand data comprises selecting a type of texture for storing the operand data according to a size of the operand data.

6. The method of claim 1, wherein the partitioning of the operand data comprises:

determining whether a quick processing mode is selected, when a size of the operand data is two times a size of the channel element of the texel;

assigning two or more operand data to one texel, when the quick processing mode is selected; and

selecting a type of texture for storing the operand data according to a size of the operand data, when the quick processing mode is not selected.

13. A non-transitory processor readable medium for storing a computer program of instructions configured to be readable by at least one processor for instructing the at least one processor to execute a computer process for performing the method as recited in claim 1.

7. An apparatus for processing data using a Graphic Processing Unit (GPU), the apparatus comprising:

a data handler configured to;

obtain operand data and operator information to be applied to the operand data, and

partition the operand data by a channel element size unit of a texel of the GPU; and

a frame buffer manager configured to;

transmit an instruction which controls the GPU to return an operation result by performing an operation for the partitioned data and transmit the partitioned data, and

obtain an operation result value from the GPU,

wherein the operational result value is a first operational result value when the GPU is operating in a first mode and a second operation result value different from the first operation result value when the GPU is operating in a second mode, and

wherein the instruction controls the GPU to restore the partitioned data into an original operand data, perform an operation for the restored operand data, and store an operation result into a frame buffer of the GPU.

8. The apparatus of claim 7, wherein the operand data is a matrix, and the frame buffer manager determines a size of a frame buffer of the GPU to which the operation result is returned according to a number of a row of the operand data, a number of a column, and a kind of an operation to be applied to the operand data.

9. The apparatus of claim 7, wherein the frame buffer manager determines a size of a frame buffer of the GPU to which the operation result is returned according to a number of a row of a first operand data and a number of a column of a second operand data among the operand data, when the operand data is a matrix and the kind of the operation is a matrix product.

10. The apparatus of claim 7, wherein the data handler assigns two or more operand data to one texel, when a size of the operand data is two times a size of the channel element of the texel.

11. The apparatus of claim 7, wherein the data handler selects a type of texture for storing the operand data according to a size of the operand data.

12. The apparatus of claim 7, wherein the data handler determines whether a quick processing mode is selected, when a size of the operand data is two times a size of the channel element of the texel, assigns two or more operand data to one texel, when the quick processing mode is selected, and selects a type of texture for storing the operand data according to a size of the operand data, when the quick processing mode is not selected.