CN109697157A - A kind of GPU statistical analysis of performance method based on data flow model - Google Patents
A kind of GPU statistical analysis of performance method based on data flow model Download PDFInfo
- Publication number
- CN109697157A CN109697157A CN201811518715.0A CN201811518715A CN109697157A CN 109697157 A CN109697157 A CN 109697157A CN 201811518715 A CN201811518715 A CN 201811518715A CN 109697157 A CN109697157 A CN 109697157A
- Authority
- CN
- China
- Prior art keywords
- stage
- unit
- processing
- vertex
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3452—Performance evaluation by statistical analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Generation (AREA)
Abstract
The invention belongs to area of computer graphics, provide a kind of GPU statistical analysis of performance method based on data flow model, comprising: (1) be divided into multiple stages according to graphics processing pipeline, each stage is handled by least one unit;Interactive information between each stage is graph data, and graphics command resolution phase parses graphics command, and the process object in vertex stage is vertex, and the process object in geometry stage is geometric graphic element, and the process object in pixel stage is pixel;(2) processing capacity for counting each unit of different phase, is counted by object of unit;(3) processing capacity for counting each stage, if a stage is composed of multiple units, the processing capacity in the stage is codetermined by the processing capacity of its multiple unit, the comprehensive processing capacity for forming the stage;(4) whole figure stream treatment system capability is analyzed, the bottleneck point of system is calculated according to the Performance Analysis in each stage, each unit and is optimized.
Description
Technical field
The invention belongs to area of computer graphics more particularly to a kind of GPU statistical analysis of performance based on data flow model
Method.
Background technique
The performance quality of GPU directly affects the application of this graphics processor, and statistical analysis of performance has become current figure
The research hotspot in shape field is mostly system-level performance evaluation in presently disclosed resource, finds cell level, stage grade
Statistical analysis of performance method, herein for cell level, the stage grade performance, propose a kind of GPU based on data flow model
Energy statistical method, covers the performance indicator in assembly line each stage, improves good data foundation for the positioning of bottleneck stage.
Summary of the invention
The purpose of the present invention is:
The GPU statistical analysis of performance method based on data flow model that present invention generally provides a kind of.
Solution of the invention is:
A kind of GPU statistical analysis of performance method based on data flow model, comprising:
(1) multiple stages, including graphics command parsing, vertex stage, geometry rank are divided into according to graphics processing pipeline
Section and pixel stage, each stage are handled by least one unit;Interactive information between each stage is graph data,
Graphics command resolution phase parses graphics command, and the process object in vertex stage is vertex, the processing pair in geometry stage
As being pixel for the process object of geometric graphic element, pixel stage;
(2) processing capacity for counting each unit of different phase, is counted by object of unit;
(3) processing capacity in each stage, including instruction generation phase ability are counted, resolution phase ability, vertex are instructed
Processing stage ability, geometric graphic element processing stage ability and processes pixel stage ability.If a stage is composed of multiple units,
Then the processing capacity in the stage is codetermined by the processing capacity of its multiple unit, the comprehensive processing capacity for forming the stage;
(4) whole figure stream treatment system capability is analyzed, according to each stage, the Performance Analysis meter of each unit
It calculates the bottleneck point for the system that obtains and optimizes.
Geometric graphic element includes point, line, triangle.
Step (2) specifically: the vertex stage counts the vertex quantity of each processing unit each second, corresponding capacity index
For Vertex/s, the geometry stage counts the pel quantity of each processing unit each second, and corresponding capacity index is
Primitive/s, pixel stage count the pixel quantity of each processing unit each second, and corresponding capacity index is Pixel/s.
Step (4) specifically:
The bottleneck stage of processing is found, the unit in bottleneck stage is positioned, optimizes unit algorithm or adding unit quantity,
The final process performance of raising system.
The invention has the advantages that traditional GPU statistical analysis of performance method is only for system-level, GPU proposed in this paper
Energy statistical method realizes that cover each stage of data flow model processing, specific aim provides not same order for cell level, stage grade
The performance indicator of section has so as to quickly position the bottleneck which is graphics process in stage in the research and development of GPU and test phase
There is important role.
Detailed description of the invention
Fig. 1 is method schematic diagram of the invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
Technical solution of the present invention is described in further detail in the following with reference to the drawings and specific embodiments.
As shown in Figure 1, a kind of GPU performance accounting method based on data flow model, this method includes (1) graphics process number
According to flow model, (2) processing unit performance statistics, (3) processing stage performance statistics, (4) system performance analysis;
(1) graphics process data flow model,
Be divided into multiple stages according to graphics processing pipeline, including graphics command parsing, the vertex stage, the geometry stage and
In the pixel stage, wherein vertex stage, geometry stage and pixel stage are since function is various, structure is complicated, and can be subdivided into multiple
Processing unit.Interactive information between graphics process data flow model each stage is graph data, graphics command resolution phase
Mainly graphics command is parsed, the process object in vertex stage is vertex, and the process object in geometry stage is geometric graphic element,
Including ten kinds of pels such as point, line, triangle, the process object in pixel stage is pixel;
(2) processing unit performance statistics,
According to (1) graphics process data flow model, the processing capacity of each unit of different phase is counted, is pair with unit
As being counted.Specifically, the vertex stage counts the vertex quantity of each processing unit each second, corresponding capacity index is
Vertex/s, geometry stage count the pel quantity of each processing unit each second, and corresponding capacity index is Primitive/
S, pixel stage count the pixel quantity of each processing unit each second, and corresponding capacity index is Pixel/s;
(3) processing stage performance statistics,
According to (1) graphics process data flow model, each stage is counted on the basis of (2) processing unit performance statistics
Processing capacity.Including instructing generation phase ability, resolution phase ability, vertex processing stage ability, geometric graphic element processing are instructed
Stage ability and processes pixel stage ability.If a stage is composed of multiple units, the processing capacity in the stage is more by it
The processing capacity of a unit codetermines, the comprehensive processing capacity for forming the stage;
(4) system performance analysis,
Whole figure stream treatment system capability is analyzed, is calculated according to the Performance Analysis of each stage, each unit
It obtains the bottleneck point of system and optimizes.
The bottleneck stage of processing is found, the unit in bottleneck stage is positioned, optimizes unit algorithm or adding unit quantity,
The final process performance of raising system.
Claims (4)
1. a kind of GPU statistical analysis of performance method based on data flow model characterized by comprising
(1) multiple stages are divided into according to graphics processing pipeline, including graphics command parsing, the vertex stage, the geometry stage and
Pixel stage, each stage are handled by least one unit;Interactive information between each stage is graph data, figure
Instruction resolution phase parses graphics command, and the process object in vertex stage is vertex, and the process object in geometry stage is
Geometric graphic element, the process object in pixel stage are pixel;
(2) processing capacity for counting each unit of different phase, is counted by object of unit;
(3) processing capacity in each stage, including instruction generation phase ability are counted, resolution phase ability, vertex processing are instructed
Stage ability, geometric graphic element processing stage ability and processes pixel stage ability;It, should if a stage is composed of multiple units
The processing capacity in stage is codetermined by the processing capacity of its multiple unit, the comprehensive processing capacity for forming the stage;
(4) whole figure stream treatment system capability is analyzed, is calculated according to the Performance Analysis of each stage, each unit
It the bottleneck point of system and optimizes out.
2. a kind of GPU statistical analysis of performance method based on data flow model as described in claim 1, which is characterized in that several
What pel includes point, line, triangle.
3. a kind of GPU statistical analysis of performance method based on data flow model as described in claim 1, which is characterized in that step
Suddenly (2) specifically: the vertex stage counts the vertex quantity of each processing unit each second, and corresponding capacity index is Vertex/
S, geometry stage count the pel quantity of each processing unit each second, and corresponding capacity index is Primitive/s, pixel rank
The pixel quantity of each processing unit each second of Duan Tongji, corresponding capacity index is Pixel/s.
4. a kind of GPU statistical analysis of performance method based on data flow model as described in claim 1, which is characterized in that step
Suddenly (4) specifically:
The bottleneck stage of processing is found, the unit in bottleneck stage is positioned, optimizes unit algorithm or adding unit quantity, is improved
The final process performance of system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811518715.0A CN109697157A (en) | 2018-12-12 | 2018-12-12 | A kind of GPU statistical analysis of performance method based on data flow model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811518715.0A CN109697157A (en) | 2018-12-12 | 2018-12-12 | A kind of GPU statistical analysis of performance method based on data flow model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109697157A true CN109697157A (en) | 2019-04-30 |
Family
ID=66231589
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811518715.0A Pending CN109697157A (en) | 2018-12-12 | 2018-12-12 | A kind of GPU statistical analysis of performance method based on data flow model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109697157A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111062855A (en) * | 2019-11-18 | 2020-04-24 | 中国航空工业集团公司西安航空计算技术研究所 | Graph pipeline performance analysis method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102449665A (en) * | 2009-06-02 | 2012-05-09 | 高通股份有限公司 | Displaying a visual representation of performance metrics for rendered graphics elements |
US9035957B1 (en) * | 2007-08-15 | 2015-05-19 | Nvidia Corporation | Pipeline debug statistics system and method |
CN106326047A (en) * | 2015-07-02 | 2017-01-11 | 超威半导体(上海)有限公司 | Method for predicting GPU performance and corresponding computer system |
CN107958436A (en) * | 2017-11-24 | 2018-04-24 | 中国航空工业集团公司西安航空计算技术研究所 | A kind of figure towards OpenGL loads quantified detection method |
US20180158168A1 (en) * | 2016-10-31 | 2018-06-07 | Imagination Technologies Limited | Performance Profiling in Computer Graphics |
-
2018
- 2018-12-12 CN CN201811518715.0A patent/CN109697157A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9035957B1 (en) * | 2007-08-15 | 2015-05-19 | Nvidia Corporation | Pipeline debug statistics system and method |
CN102449665A (en) * | 2009-06-02 | 2012-05-09 | 高通股份有限公司 | Displaying a visual representation of performance metrics for rendered graphics elements |
CN106326047A (en) * | 2015-07-02 | 2017-01-11 | 超威半导体(上海)有限公司 | Method for predicting GPU performance and corresponding computer system |
US20180158168A1 (en) * | 2016-10-31 | 2018-06-07 | Imagination Technologies Limited | Performance Profiling in Computer Graphics |
CN107958436A (en) * | 2017-11-24 | 2018-04-24 | 中国航空工业集团公司西安航空计算技术研究所 | A kind of figure towards OpenGL loads quantified detection method |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111062855A (en) * | 2019-11-18 | 2020-04-24 | 中国航空工业集团公司西安航空计算技术研究所 | Graph pipeline performance analysis method |
CN111062855B (en) * | 2019-11-18 | 2023-09-05 | 中国航空工业集团公司西安航空计算技术研究所 | Graphic pipeline performance analysis method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110309824B (en) | Character detection method and device and terminal | |
CN102509099B (en) | Detection method for image salient region | |
CN107958436B (en) | OpenGL-oriented graph load quantitative detection method | |
CN102999926B (en) | A kind of image vision significance computational methods merged based on low-level image feature | |
CN103886633A (en) | Tessellating Patches Of Surface Data In Tile Based Computer Graphics Rendering | |
CN114170482B (en) | Document pre-training model training method, device, equipment and medium | |
CN102930518A (en) | Improved sparse representation based image super-resolution method | |
WO2020125062A1 (en) | Image fusion method and related device | |
CN112505652B (en) | Target detection method, device and storage medium | |
CN110490203A (en) | Image partition method and device, electronic equipment and computer readable storage medium | |
CN111625668A (en) | Object detection and candidate filtering system | |
CN113439227B (en) | Capturing and storing enlarged images | |
CN114299284A (en) | Training method, using method, device, equipment and medium of segmentation model | |
CN109697157A (en) | A kind of GPU statistical analysis of performance method based on data flow model | |
CN101354793B (en) | Real time three-dimensional image smoothing process method based on pattern processor | |
CN111738252A (en) | Method and device for detecting text lines in image and computer system | |
CN107566344A (en) | A kind of CAN signal analytic method and system | |
CN111597845A (en) | Two-dimensional code detection method, device and equipment and readable storage medium | |
CN110796115B (en) | Image detection method and device, electronic equipment and readable storage medium | |
CN116137061A (en) | Training method and device for quantity statistical model, electronic equipment and storage medium | |
CN111368572A (en) | Two-dimensional code identification method and system | |
CN114201369A (en) | Server cluster management method and device, electronic equipment and storage medium | |
CN114120306A (en) | License plate character segmentation method and device and storage medium | |
CN110910364A (en) | Method for detecting electrical equipment easy to cause fire in three-section fire scene based on deep neural network | |
CN114240924A (en) | Power grid equipment quality evaluation method based on digitization technology |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190430 |
|
RJ01 | Rejection of invention patent application after publication |