JP2010287110A - Information processor, information processing method, program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information processor and information processing method for increasing processing speed, in the information processor for executing graphics drawing and video encoding or transcoding in parallel. <P>SOLUTION: A first processing unit to be executed first in video encoding or transcoding is executed by a parallel data processing part and a general data processing part. The time when the parallel data processing part terminates the first processing unit is compared with the time when the general data processing part terminates the first processing unit. Whether the parallel data processing part or the general data processing part is used to execute the second and subsequent processing units after the first processing unit, is determined. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to high-speed data processing, and more particularly to an information processing apparatus capable of speeding up processing by load distribution in an information processing apparatus that simultaneously processes graphics rendering processing and video encoding processing and transcoding processing. The present invention relates to an information processing method, a program, and a recording medium.

  Conventionally, a graphics processing unit (GPU) with high parallel processing capability has been used to perform graphics drawing processing that requires three-dimensional operations at high speed. GPGPU (General Purpose Computation on Graphics Processing Unit) has emerged as a technique used for general-purpose numerical computations. Since GPGPU has the processing characteristics of GPU, it is suitable for parallel processing of relatively simple processes using a large amount of data compared to a highly versatile CPU (Central Processing Unit). While effective for arithmetic processing, etc., on the other hand, a program that does not match its characteristics cannot be expected to perform.

  An apparatus that distributes graphics rendering processing to a plurality of GPUs and balances the load is described in, for example, Patent Document 1.

JP 2008-71261 A

  By the way, the above-mentioned GPGPU is suitable for processing a large amount of simple data in parallel. For this reason, by performing the video encoding processing and transcoding processing, which have been conventionally performed by the CPU, instead of the CPU, the processing can be speeded up and the burden on the CPU can be reduced.

  However, in GPGPU, even if it is a program process that produces an effect when executed alone, such as video encoding process or transcoding process, a graphics drawing process with high load and unpredictable load variation, for example, 3D video If the video encoding process is executed in parallel in the situation where the vertex calculation for generating the three-dimensional image and the arithmetic process represented by the shadow calculation are executed as the drawing process of the image, the process of performing the GPGPU calculation The H / W that performs the three-dimensional calculation process with the part is time-shared with the whole or a part, so that the GPGPU is in a high load state and the process is not in time. As a result, the smoothness of the moving image is lost in the rendering process of the three-dimensional moving image, and a frame drop occurs in the encoding process of the video that requires real-time performance.

  In other words, generally in a multitasking environment, if there are heavy tasks, other task processing will be affected, but by processing the video encoding processing that was conventionally processed by the CPU with GPGPU, the original GPGPU It is a problem that drawing processing of a display screen that requires real-time performance is not in time.

  Therefore, the present invention estimates the load state of the GPGPU and CPU in an information processing apparatus that executes graphics rendering processing, video encoding processing, and transcoding processing in parallel, and further encodes video based on the characteristics of the processing target. Information processing apparatus and information processing method capable of distributing processing load and speeding up processing by deciding whether each processing step in processing or transcoding processing is allocated as GPGPU or CPU processing The purpose is to do.

  The information processing apparatus of the present invention executes a first processing unit to be processed first, which is a processing target composed of a plurality of data processing means and a plurality of processing units, by the plurality of data processing means. Based on the comparison result of the processing time comparing means for comparing the required time and the processing time comparing means, the second processing unit to be processed next to the first processing unit is one of the plurality of data processing means. And control means to be executed in step (b).

  The information processing apparatus of the present invention executes a first processing unit to be processed first, which is a processing target composed of a plurality of data processing means and a plurality of processing units, by the plurality of data processing means. In the comparison result of the processing time comparing means for comparing the required time and the processing time comparing means, if the time differences required for the processing of the plurality of data processing means are all within a predetermined time, the first processing unit The second processing unit to be processed next is executed by the plurality of data processing means, and the comparison result of the second processing time comparing means for comparing the time required for the processing and the second processing time comparing means And a control unit that causes a third processing unit to be processed next to the second processing unit to be executed by any one of the plurality of data processing units.

  According to the information processing method of the present invention, a plurality of data processing units and a first processing unit to be processed first, which is a processing target composed of a plurality of processing units, are executed by the plurality of data processing units. Based on the comparison result of the processing time comparing means for comparing the required time and the processing time comparing means, the second processing unit to be processed next to the first processing unit is one of the plurality of data processing means. And control means to be executed in step (b).

  In the information processing method of the present invention, a first processing unit to be processed first, which is a processing target composed of a plurality of data processing steps and a plurality of processing units, is executed at each of the plurality of data processing steps. In the comparison result of the processing time comparison step for comparing the required time and the comparison result of the processing time comparison step, when the difference between the times required for the processing of the plurality of data processing steps is all within a predetermined time, the second processing unit is The second processing unit is executed in each of the plurality of data processing steps, and the second processing unit is compared based on the comparison result of the second processing time comparison step for comparing the time required for processing and the second processing time comparison step. And a control step for executing the third processing unit to be processed next in any of the plurality of data processing steps.

  The program of the present invention executes a first processing unit to be processed first, which is a processing target composed of a plurality of data processing steps and a plurality of processing units, at each of the plurality of data processing steps, and takes a time required for the processing. A processing time comparison step for comparing the first processing unit and a second processing unit to be processed next to the first processing unit based on a comparison result of the processing time comparison step in any of the plurality of data processing steps And a control step for causing the computer to execute the control step.

  The program of the present invention executes a first processing unit to be processed first, which is a processing target composed of a plurality of data processing steps and a plurality of processing units, at each of the plurality of data processing steps, and takes a time required for the processing. In the comparison results of the processing time comparison step and the processing time comparison step, when all the time differences required for the processing of the plurality of data processing steps are within a predetermined time, the second processing unit is Next to the second processing unit based on the comparison result of the second processing time comparison step and the second processing time comparison step, each of which is executed in a plurality of data processing steps and compares the time required for processing. A control step of causing a computer to execute a third processing unit to be processed and the subsequent steps in any one of the plurality of data processing steps is characterized.

  The recording medium of the present invention is a computer-readable recording medium for recording the processing of the program according to the present invention.

  According to the present invention, an information processing apparatus and an information processing method capable of speeding up processing in an information processing apparatus, particularly in an information processing apparatus that executes graphics rendering processing and video encoding processing and transcoding processing in parallel. Can be provided.

It is a functional lineblock diagram of an information processor concerning an embodiment of the present invention. It is a figure for demonstrating the operation processing of the information processing apparatus which concerns on embodiment of this invention. It is a flowchart which shows the process outline | summary of the information processing apparatus which concerns on embodiment of this invention. It is a flowchart at the time of changing the timing which transfers to the general purpose data processing from the parallel data processing of the information processing apparatus which concerns on embodiment of this invention. It is a flowchart in the case of determining time comparison of parallel data processing and general-purpose data processing of the information processing apparatus according to the embodiment of the present invention in two stages.

  Embodiments of the present invention will be described below in detail with reference to the drawings. The embodiments described below are preferred embodiments of the present invention, and thus various technically preferable limitations are given. However, the scope of the present invention is particularly limited in the following description. As long as there is no description of the effect, it is not restricted to these aspects.

  FIG. 1 is a functional configuration diagram of an information processing apparatus according to an embodiment of the present invention. A display device 2 is connected to the information processing device 1.

  As shown in FIG. 1, the information processing apparatus 1 includes a general-purpose data processing unit 11, a data copy unit 12, a data storage unit 13, a processing time comparison unit 14, and a GPGPU 15. The GPGPU 15 includes a parallel data processing unit 151 and a data storage unit 152.

  The general-purpose data processing unit 11 has a function of performing various numerical calculations, information processing, device control, etc. (having a function of performing more general-purpose data processing than the parallel data processing unit 151).

  The data copy unit 12 has a function of copying all or part of data stored in the data storage unit 152 from the data storage unit 152 of the GPGPU 15 and from the data storage unit 13 to the data storage unit 152. Note that the data copy unit 12 may be configured as one of the functions of the GPGPU 15.

  The data storage unit 13 has a function of storing data copied from the data storage unit 152 of the GPGPU 15 and can be read and written as processing target data of the general-purpose data processing unit 11. The general-purpose data processing unit 11 reads out the data stored in the data storage unit 13 and performs processing, and writes the obtained processing result in the data storage unit 13 again.

  The processing time comparison unit 14 has a function of measuring time required for processing execution in the general-purpose data processing unit and the parallel data processing unit of the GPGPU 15 and comparing the measurement times.

  The GPGPU 15 includes a parallel data processing unit 151 and a data storage unit 152. Data to be processed is stored in the data storage unit 152, and can be read and written as processing target data of the parallel data processing unit 151. The parallel data processing unit 151 has a function of reading and processing data stored in the data storage unit 152 and writing the obtained processing result to the data storage unit 152 again. The parallel data processing unit 151 is suitable for parallel processing of relatively simple processing using a large amount of data as compared with the general-purpose data processing unit 11. Higher processing capability is demonstrated in encoding processing and transcoding processing.

  However, for example, a process directed to the parallel data processing unit 151 (such as a graphic rendering process of a three-dimensional moving image) and a process not directed to the parallel data processing unit 151 (that is, a process directed to the general-purpose data processing unit 11) appear alternately. Assuming a program that needs to execute the following processing based on the processing result of, in order to efficiently execute this program, the processing for the parallel data processing unit 151 is processed by the parallel data processing unit 151, Processing that is not suitable for the parallel data processing unit 151 is likely to be performed by the general-purpose data processing unit 11 alternately. However, in practice, since the processing result of the preceding process is used as input data for the next process, data copy is repeatedly executed between the data storage unit 152 and the data storage unit 13 every time the processing unit is switched. There is a need. In other words, it is necessary to take into account the influence of processing overhead, and particularly when the data size increases, the copy processing time also increases, so the influence of processing overhead is significant. Therefore, when the data size to be processed is large and the copy processing cost exceeds the effect of increasing the processing cost in the parallel data processing unit 151 or the general-purpose data processing unit 11, data is transferred between the data storage unit 152 and the data storage unit 13. It is not efficient to execute copying repeatedly.

  FIG. 2 is a diagram for explaining an operation process of the information processing apparatus according to the embodiment of the present invention.

  Here, as an example, a case will be described in which processing for encoding one frame of video is divided into 12 processing units (A to L). In the example of FIG. 2, encoding processing is performed in the order of processing units A to L, the characteristics of the processing target data change from image to code, and the processing data size decreases as the processing proceeds from A to L. ing. These processes are repeatedly executed. For example, in a video of 30 frames per second, the processing units A to L are repeatedly executed in a cycle of 1/30 = 33.3 ms. Further, when processing unit A is processed alone (there is no need to consider the influence of other program processing, etc.), it is shorter when executed in parallel data processing unit 151 than in general data processing unit 11. There is a tendency to process efficiently in time. This tendency appears most prominently for the processing unit A, and the tendency decreases as the processing unit B and the processing unit C become. In the processing unit L, the general-purpose data processing unit 11 rather than the parallel data processing unit 151. The tendency to be able to perform processing more efficiently in a shorter time appears when it is executed. This is because the closer to the process A, the more likely the simple process can be executed in parallel for a large amount of data, and the closer to the process L, the greater the tendency of the sequential process to generate a code.

  FIG. 2 illustrates three processing patterns (patterns 1 to 3), and the black circles indicate which of the processing units is executed by the parallel data processing unit 151 or the general-purpose data processing unit 11. For example, since the first processing unit A in the video encoding processing flow has black circles in both the “parallel data processing unit” and the “general-purpose data processing unit”, the processing unit C is executed by both processing units. Thereafter (CL), in any of the three cases, there is only a black circle in either the “parallel data processing unit” or the “general-purpose data processing unit”. Show. That is, at the determination point DP1 or DP2 in FIG. 2, the parallel data processing unit compares the time at which the processing unit A or B ends with the time at which the general-purpose data processing unit ends the processing unit A or B, respectively. This indicates that the parallel data processing unit or the general-purpose data processing unit determines in what pattern the subsequent processing (processing units C to L) is executed. Note that the number and location of the determination points DP1 and DP2 are not limited in this embodiment. For example, the determination point DP1 may be performed after a plurality of processing units A to C are collectively executed.

  FIG. 3 is a flowchart showing an outline of processing of the information processing apparatus according to the embodiment of the present invention.

  Data to be processed is stored in the data storage unit 152 and becomes a processing target of the parallel data processing unit 151 (S301). Further, the data to be processed is copied to the data storage unit 13 and becomes a processing target of the general-purpose data processing unit 11 (S302). ). The order in which the data to be processed is stored in the data storage unit 152 and the data storage unit 13 (S301, S302) is not limited in this embodiment.

  The same processing unit A is executed by both processing units of the parallel data processing unit 151 and the general-purpose data processing unit 11 (S303, S304), and the processing times of the processing unit A in both processing units are compared (S305). Note that the processing times of both processing units may be compared after adjusting for a certain time.

As a result of comparing the processing time of the processing unit A, when the parallel data processing unit 151 completes the processing earlier, the parallel data processing unit 151 executes the processing units B to F (S306), and the processing result data is stored as data. The data is copied to the storage unit 13 (S307), and this time, the general data processing unit 11 executes the processing units G to L (S308). (Processing flow of pattern 2 in FIG. 2 (pattern not including a broken line: processing unit B is executed only in the parallel data processing unit))
On the other hand, as a result of comparing the processing times of the processing units A, when the general-purpose data processing unit 11 completes the process earlier, the general-purpose data processing unit 11 executes the subsequent processes B to L (S309). (Processing flow of pattern 1 in FIG. 2 (pattern not including a broken line: processing unit B is executed only in the general-purpose data processing unit))
FIG. 4 is a flowchart in a case where the timing of shifting from parallel data processing to general-purpose data processing of the information processing apparatus according to the embodiment of the present invention is changed.

  The processing target data is stored in the data storage unit 152 and becomes a processing target of the parallel data processing unit 151 (S401). Further, the processing target data is copied to the data storage unit 13 and becomes a processing target of the general-purpose data processing unit 11 (S402). ).

  The same processing unit A is executed by both processing units of the parallel data processing unit 151 and the general-purpose data processing unit 11 (S403, S404), and the processing time of the processing unit A in both processing units is compared (S405).

As a result of comparing the processing time of the processing unit A, when the parallel data processing unit 151 completes the processing earlier, the parallel data processing unit 151 executes the processing units B to J (S406), and the processing result data is stored as data. The data is copied to the storage unit 13 (S407), and this time, the general-purpose data processing unit 11 executes the processing units K to L (S408). (Processing flow of pattern 3 in FIG. 2 (pattern not including a broken line: processing unit B is executed only in the parallel data processing unit))
On the other hand, as a result of comparing the processing time of the processing unit A, when the general data processing unit 11 completes the processing earlier, the general data processing unit 11 executes the subsequent processes B to L (S409). (Processing flow of pattern 1 in FIG. 2 (pattern not including a broken line: processing unit B is executed only in the general-purpose data processing unit))
FIG. 5 is a flowchart when the time comparison between the parallel data processing and the general-purpose data processing of the information processing apparatus according to the embodiment of the present invention is determined in two stages.

  Data to be processed is stored in the data storage unit 152 and becomes a processing target of the parallel data processing unit 151 (S501). Further, the data to be processed is copied to the data storage unit 13 and becomes a processing target of the general-purpose data processing unit 11 (S502). ).

  The same processing unit A is executed by both the parallel data processing unit 151 and the general-purpose data processing unit 11 (S503, S504), and the processing times of the processing unit A in both processing units are compared (S505).

As a result of comparing the processing time of the processing unit A, when the parallel data processing unit 151 completes the processing earlier, the parallel data processing unit 151 executes the processing units B to J (S506), and the processing result data is stored as data. The data is copied to the storage unit 13 (S507), and this time, the general-purpose data processing unit 11 executes the processing units K to L (S508). (Processing flow of pattern 3 in FIG. 2 (pattern not including a broken line: processing unit B is executed only in the parallel data processing unit))
On the other hand, as a result of comparing the processing time of the processing unit A, when the general-purpose data processing unit 11 completes the processing earlier, the parallel processing unit 151 and the general-purpose data processing unit 11 further process the same processing unit B. (S509, S510), and the processing time of the processing unit B in both processing units is compared (S511).

As a result of comparing the processing times of the processing unit B, when the parallel data processing unit 151 completes the processing earlier, the parallel data processing unit 151 executes the processing units C to J (S512), and the processing result data is stored as data. The data is copied to the storage unit 13 (S507), and this time, the general-purpose data processing unit 11 executes the processing units K to L (S508). (Processing flow of pattern 3 in FIG. 2 (pattern including a broken line: processing unit B is executed by both processing units))
On the other hand, as a result of comparing the processing time of the processing unit B, when the general data processing unit 11 completes the processing earlier, the general data processing unit 11 executes the subsequent processes C to L (S513). (Processing flow of pattern 1 in FIG. 2 (pattern including a broken line: processing unit B is executed by both processing units))
Each of the above-described embodiments is a preferred embodiment of the present invention, and the scope of the present invention is not limited to only the above-described embodiments, and those skilled in the art do not deviate from the gist of the present invention. It is possible to construct a form in which various modifications are made by modifying or substituting the embodiment.

  For example, the control operation in the above-described embodiment can be executed by hardware, software, or a combined configuration of both.

  When executing processing by software, whether the program is read from the ROM storing the program recording the processing sequence to the memory (RAM) in the computer incorporated in the dedicated hardware and executed. Alternatively, the program can be installed and executed on a general-purpose computer capable of executing various processes.

DESCRIPTION OF SYMBOLS 1 Information processing apparatus 2 Display apparatus 11 General-purpose data processing part 12 Data copy part 13 Data storage part 14 Processing time comparison part 15 GPGPU
151 Parallel Data Processing Unit 152 Data Storage Unit

Claims (10)

A plurality of data processing means;
A first processing unit to be processed first, which is a processing target composed of a plurality of processing units, is executed by each of the plurality of data processing units, and a processing time comparison unit for comparing the time required for processing;
Control means for causing any one of the plurality of data processing means to execute a second processing unit to be processed next to the first processing unit based on the comparison result of the processing time comparison means. A characteristic information processing apparatus.   The information processing apparatus according to claim 1, wherein the first processing unit further includes a plurality of processing units. A plurality of data processing means;
A first processing unit to be processed first, which is a processing target composed of a plurality of processing units, is executed by each of the plurality of data processing units, and a processing time comparison unit for comparing the time required for processing;
In the comparison result of the processing time comparing means, when all the time differences required for the processing of the plurality of data processing means are within a predetermined time, a second processing unit to be processed next to the first processing unit, A second processing time comparing means for executing each of the plurality of data processing means and comparing the time required for the processing;
A control unit that causes a third processing unit to be processed next to the second processing unit to be executed by any one of the plurality of data processing units based on a comparison result of the second processing time comparison unit; An information processing apparatus comprising:   The information processing apparatus according to claim 3, wherein the first processing unit or the second processing unit further includes a plurality of processing units.   5. The information processing apparatus according to claim 1, wherein the plurality of data processing means include parallel data processing means and general-purpose data processing means. Multiple data processing steps;
A processing time comparison step of executing a first processing unit to be processed first, which is a processing target composed of a plurality of processing units, in each of the plurality of data processing steps and comparing the time required for the processing;
A control step for executing a second processing unit to be processed next to the first processing unit in one of the plurality of data processing steps based on a comparison result of the processing time comparison step. A characteristic information processing method. Multiple data processing steps;
A processing time comparison step of executing a first processing unit to be processed first, which is a processing target composed of a plurality of processing units, in each of the plurality of data processing steps and comparing the time required for the processing;
In the comparison result of the processing time comparison step, when all the time differences required for the processing of the plurality of data processing steps are within a predetermined time, the second processing unit is executed in each of the plurality of data processing steps. A second processing time comparison step for comparing the time required for processing;
A control step of executing, in any one of the plurality of data processing steps, a third processing unit to be processed next to the second processing unit based on a comparison result of the second processing time comparison step; An information processing method comprising: Multiple data processing steps;
A processing time comparison step of executing a first processing unit to be processed first, which is a processing target composed of a plurality of processing units, in each of the plurality of data processing steps and comparing the time required for the processing;
Based on the comparison result of the processing time comparison step, the computer executes a control step that causes a second processing unit to be processed after the first processing unit to be executed in any of the plurality of data processing steps. A program characterized by letting Multiple data processing steps;
A processing time comparison step of executing a first processing unit to be processed first, which is a processing target composed of a plurality of processing units, in each of the plurality of data processing steps and comparing the time required for the processing;
In the comparison result of the processing time comparison step, when all the time differences required for the processing of the plurality of data processing steps are within a predetermined time, the second processing unit is executed in each of the plurality of data processing steps. A second processing time comparison step for comparing the time required for processing;
A control step of executing, in any one of the plurality of data processing steps, a third processing unit to be processed next to the second processing unit based on a comparison result of the second processing time comparison step; A program characterized by being executed by a computer.   A computer-readable recording medium for recording the processing of the program according to claim 8 or 9.

Images