CN109712061A - A kind of GPU command processor robustness operation management method - Google Patents
A kind of GPU command processor robustness operation management method Download PDFInfo
- Publication number
- CN109712061A CN109712061A CN201811510181.7A CN201811510181A CN109712061A CN 109712061 A CN109712061 A CN 109712061A CN 201811510181 A CN201811510181 A CN 201811510181A CN 109712061 A CN109712061 A CN 109712061A
- Authority
- CN
- China
- Prior art keywords
- gpu
- command
- command processor
- processor
- graph
- 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
- 238000007726 management method Methods 0.000 title claims abstract description 20
- 238000012544 monitoring process Methods 0.000 claims abstract description 46
- 238000000034 method Methods 0.000 claims abstract description 28
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 28
- 238000012360 testing method Methods 0.000 claims description 12
- 238000003780 insertion Methods 0.000 claims description 8
- 230000037431 insertion Effects 0.000 claims description 8
- 230000000149 penetrating effect Effects 0.000 claims 1
- 238000013507 mapping Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
Landscapes
- Debugging And Monitoring (AREA)
Abstract
The present invention provides one kind, the invention belongs to area of computer graphics more particularly to a kind of GPU command processor robustness operation management method.This method mainly includes GPU command processor operational monitoring (3), monitoring resource (2), external monitoring (3).This method is directed to the operation characteristic and task feature of GPU command processor, monitors the operation overall process of GPU command processor, achievees the purpose that GPU command processor robustness management.
Description
Technical field
The invention belongs to computer graphics techniques fields, are related to a kind of GPU command processor robustness operation management method.
Background technique
Command processor in GPU is first functional unit for handling graph command, needs to handle a large amount of graph commands,
Running robustness influences entire processor operating condition.The data of GPU command processor processing are different from general command process
Device needs to design special robustness monitoring method, it is ensured that normal operation.
Summary of the invention
The present invention provides one kind for graphics processor command process robustness operation management method, to guarantee at order
Manage the normal operation of device.
First invention, the present invention provide a kind of GPU command processor robustness operation management method, comprising:
When graph command inputs GPU command processor, insertion operational monitoring label, the monitoring mark is used in GPU
Operational monitoring is carried out during command processor processing graph command；
The graph command end of run output pattern flowing water control stream and data flow；
External monitoring is carried out to the graphical stream water management stream and data stream；
Monitoring resource is carried out to the process of GPU command processor processing graph command.
Optionally, it at the time of the insertion operational monitoring marks, specifically includes:
The moment is generated when front end generates graph command simultaneously；When GPU command processor receives graph command and successfully resolved
Generate the moment.
Optionally, the operational monitoring label, specifically includes: GPU command processor instructs self-test to mark, at GPU order
Manage device register self-test label, GPU command processor memory self-test label.
Optionally, described that external monitoring is carried out to the graphical stream water management stream and data stream, it specifically includes:
Graph command is entered after the completion of the input node and processing of GPU command processor and is exported from GPU command processor
Node be monitored.
Optionally, from GPU order after the completion of the input node and processing that GPU command processor is entered to graph command
The node of processor output is monitored, and is specifically included:
Runing time to the graph command in GPU command processor is monitored, and the runing time is less than described
The preset time of graph command.
Optionally, graph command is entered after the completion of the input node and processing of GPU command processor from GPU command process
The node of device output is monitored, and is specifically included:
The output pattern flowing water control stream of the output end output and the graph command of data flow and input terminal input
Mapping relations have to comply with the operation result of the graph command.
Optionally, the process to GPU command processor processing graph command carries out monitoring resource, specifically includes:
Monitor the I/O device state and current Graphics flowing water state of GPU graphics processor.
Optionally, the I/O device state of the GPU graphics processor, specifically includes:
I/O device accesses the bandwidth and performance of the correctness of storage device data, I/O device access data path.
Optionally, the current Graphics flowing water state, specifically includes:
Figure flowing water makes fruitless efforts state, figure the flowing water a certain grade processing time less than this grade of desired value.
GPU command processor robustness operation management method provided by the invention, by inputting GPU order in graph command
When processor, insertion operational monitoring label, the monitoring mark is used for during GPU command processor handles graph command
Carry out operational monitoring；The graph command end of run output pattern flowing water control stream and data flow；To the graphical stream water control
System stream and data stream carry out external monitoring；Monitoring resource is carried out to the process of GPU command processor processing graph command, it is special to examine
The issuable failure of command processor and mistake that carry out a large amount of command analysis and processing are surveyed and handled, ensure that GPU order
The normal operation of processor.
Detailed description of the invention
Fig. 1 is the schematic diagram of GPU command processor robustness management method provided by 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.
Embodiment one
The present invention provides a kind of GPU command processor robustness operation management method, comprising:
Step 101: when graph command inputs GPU command processor, insertion operational monitoring label, the monitoring mark is used
In GPU command processor handle graph command during carry out operational monitoring；
Step 102: the graph command end of run output pattern flowing water control stream and data flow；
Step 103: external monitoring is carried out to the graphical stream water management stream and data stream；
Step 104: monitoring resource is carried out to the process of GPU command processor processing graph command.
Optionally, it at the time of the insertion operational monitoring marks, specifically includes:
The moment is generated when front end generates graph command simultaneously；When GPU command processor receives graph command and successfully resolved
Generate the moment.
Optionally, the operational monitoring label, specifically includes: GPU command processor instructs self-test to mark, at GPU order
Manage device register self-test label, GPU command processor memory self-test label.
Optionally, described that external monitoring is carried out to the graphical stream water management stream and data stream, it specifically includes:
Graph command is entered after the completion of the input node and processing of GPU command processor and is exported from GPU command processor
Node be monitored.
Optionally, from GPU order after the completion of the input node and processing that GPU command processor is entered to graph command
The node of processor output is monitored, and is specifically included:
Runing time to the graph command in GPU command processor is monitored, and the runing time is less than described
The preset time of graph command.
Optionally, graph command is entered after the completion of the input node and processing of GPU command processor from GPU command process
The node of device output is monitored, and is specifically included:
The output pattern flowing water control stream of the output end output and the graph command of data flow and input terminal input
Mapping relations have to comply with the operation result of the graph command.
Optionally, the process to GPU command processor processing graph command carries out monitoring resource, specifically includes:
Monitor the I/O device state and current Graphics flowing water state of GPU graphics processor.
Optionally, the I/O device state of the GPU graphics processor, specifically includes:
I/O device accesses the bandwidth and performance of the correctness of storage device data, I/O device access data path.
Optionally, the current Graphics flowing water state, specifically includes:
Figure flowing water makes fruitless efforts state, figure the flowing water a certain grade processing time less than this grade of desired value.
In conclusion GPU command processor robustness operation management method provided by the invention, by defeated in graph command
When entering GPU command processor, insertion operational monitoring label, the monitoring mark is used in GPU command processor processing figure life
Operational monitoring is carried out during order；The graph command end of run output pattern flowing water control stream and data flow；To described
Graphical stream water management stream and data stream carry out external monitoring；Resource is carried out to the process of GPU command processor processing graph command
Monitoring, special detection and processing carry out the issuable failure of command processor and mistake of a large amount of command analysis and processing, protect
The normal operation of GPU command processor is demonstrate,proved.
With reference to the accompanying drawing 1 and specific embodiment technical solution of the present invention is described in further detail.
A kind of GPU command processor robustness operation management method, comprising: external monitoring (1), monitoring resource (2), operation
It monitors (3), sphere of action is GPU command processor；
Illustratively, monitoring graph command inputs GPU command processor to output two nodes of GPU command processor.Prison
The target of survey includes but is not limited to be input to the time interval of output to must not exceed the desired value of the graph command, output and input
Mapping relations have to comply with the operation result of the graph command；
Illustratively, the I/O device and current Graphics flowing water state of GPU graphics processor are monitored.The target of monitoring include but
It is not limited to, I/O device needed for GPU command processor accesses expection of the time interval no more than the graph command of external module
Value, access correctness, the bandwidth of data path and the performance of storage device data, figure flowing water makes fruitless efforts state and figure flowing water
Corresponding time interval；
Illustratively, can be generate graph command when simultaneously generate, be also possible to GPU command processor reception figure
It orders and when successfully resolved generates, circulate with processing of the graph command in GPU command processor.The range of effect is
The overall process of GPU command processor operating, the target of monitoring include but is not limited to that GPU command processor instructs self-test, GPU life
Enable the self-test of processor register, GPU command processor memory self-test.
Claims (9)
1. a kind of GPU command processor robustness operation management method, which is characterized in that the described method includes:
When graph command inputs GPU command processor, insertion operational monitoring label, the monitoring mark is used in GPU order
Operational monitoring is carried out during processor processing graph command；
The graph command end of run output pattern flowing water control stream and data flow；
External monitoring is carried out to the graphical stream water management stream and data stream；
Monitoring resource is carried out to the process of GPU command processor processing graph command.
2. method described in claim 1, which is characterized in that at the time of the insertion operational monitoring marks, specifically include:
The moment is generated when front end generates graph command simultaneously；GPU command processor generates when receiving graph command and successfully resolved
Moment.
3. method as claimed in claim 2, which is characterized in that the operational monitoring label specifically includes: GPU command processor
Instruct self-test label, GPU command processor register self-test label, GPU command processor memory self-test label.
4. method described in claim 1, which is characterized in that described to carry out outside to the graphical stream water management stream and data stream
Monitoring, specifically includes:
Enter the section exported after the completion of the input node and processing of GPU command processor from GPU command processor to graph command
Point is monitored.
5. method as claimed in claim 4, which is characterized in that the input section for entering GPU command processor to graph command
It is monitored, specifically includes from the node that GPU command processor exports after the completion of point and processing:
Runing time to the graph command in GPU command processor is monitored, and the runing time is less than the figure
The preset time of order.
6. method as claimed in claim 4, which is characterized in that graph command enter GPU command processor input node and
It is monitored, specifically includes from the node that GPU command processor exports after the completion of processing:
The graph command that the output pattern flowing water control stream and data flow of the output end output are inputted with the input terminal is reflected
The relationship of penetrating has to comply with the operation result of the graph command.
7. method described in claim 1, which is characterized in that it is described to GPU command processor processing graph command process into
Row monitoring resource, specifically includes:
Monitor the I/O device state and current Graphics flowing water state of GPU graphics processor.
8. method of claim 7, which is characterized in that the I/O device state of the GPU graphics processor specifically includes:
I/O device accesses the bandwidth and performance of the correctness of storage device data, I/O device access data path.
9. method of claim 7, which is characterized in that the current Graphics flowing water state specifically includes:
Figure flowing water makes fruitless efforts state, figure the flowing water a certain grade processing time less than this grade of desired value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811510181.7A CN109712061A (en) | 2018-12-11 | 2018-12-11 | A kind of GPU command processor robustness operation management method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811510181.7A CN109712061A (en) | 2018-12-11 | 2018-12-11 | A kind of GPU command processor robustness operation management method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109712061A true CN109712061A (en) | 2019-05-03 |
Family
ID=66255614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811510181.7A Pending CN109712061A (en) | 2018-12-11 | 2018-12-11 | A kind of GPU command processor robustness operation management method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109712061A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070139421A1 (en) * | 2005-12-21 | 2007-06-21 | Wen Chen | Methods and systems for performance monitoring in a graphics processing unit |
| US7600155B1 (en) * | 2005-12-13 | 2009-10-06 | Nvidia Corporation | Apparatus and method for monitoring and debugging a graphics processing unit |
| US20170177458A1 (en) * | 2015-12-18 | 2017-06-22 | Stephen Viggers | Methods and Systems for Monitoring the Integrity of a GPU |
| CN108021487A (en) * | 2017-11-24 | 2018-05-11 | 中国航空工业集团公司西安航空计算技术研究所 | A kind of GPU graphics process performance monitoring and analysis method |
| CN108140234A (en) * | 2015-10-23 | 2018-06-08 | 高通股份有限公司 | GPU operation algorithms selection based on order flow label |
| CN108241532A (en) * | 2016-12-23 | 2018-07-03 | 北京奇虎科技有限公司 | The management distribution method of GPU resource and management distributor |
-
2018
- 2018-12-11 CN CN201811510181.7A patent/CN109712061A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7600155B1 (en) * | 2005-12-13 | 2009-10-06 | Nvidia Corporation | Apparatus and method for monitoring and debugging a graphics processing unit |
| US20070139421A1 (en) * | 2005-12-21 | 2007-06-21 | Wen Chen | Methods and systems for performance monitoring in a graphics processing unit |
| CN108140234A (en) * | 2015-10-23 | 2018-06-08 | 高通股份有限公司 | GPU operation algorithms selection based on order flow label |
| US20170177458A1 (en) * | 2015-12-18 | 2017-06-22 | Stephen Viggers | Methods and Systems for Monitoring the Integrity of a GPU |
| CN108241532A (en) * | 2016-12-23 | 2018-07-03 | 北京奇虎科技有限公司 | The management distribution method of GPU resource and management distributor |
| CN108021487A (en) * | 2017-11-24 | 2018-05-11 | 中国航空工业集团公司西安航空计算技术研究所 | A kind of GPU graphics process performance monitoring and analysis method |
Non-Patent Citations (4)
| Title |
|---|
| ARNAU JM等: "Boosting Mobile GPU Performance with a Decoupled Access /Execute Fragment Processor", 《ACM SIGAＲCH COMPUTER ARCHITECTURE NEWS》 * |
| 张骏等: "面向GPU统一染色阵列的并行自适应看门狗", 《航空计算技术》 * |
| 朱宇兰: "基于GPU通用计算的并行算法和计算框架的实现", 《山东农业大学学报(自然科学版)》 * |
| 李荣振等: "基于飞腾平台的GPU图形加速驱动设计与实现", 《计算机工程与应用》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106339058B (en) | Dynamic manages the method and system of power supply | |
| US10073683B2 (en) | System and method for providing software build violation detection and self-healing | |
| US8635484B2 (en) | Event based correlation of power events | |
| US10098258B2 (en) | Minimizing leakage in liquid cooled electronic equipment | |
| WO2020080518A1 (en) | Systems and methods for dynamically identifying data arguments and instrumenting source code | |
| CN111145076B (en) | Data parallelization processing method, system, equipment and storage medium | |
| CN110083146B (en) | Fault determination method and device for autonomous vehicle, equipment and storage medium | |
| EP3068205A1 (en) | Minimizing leakage in liquid cooled electronic equipment | |
| US20140223229A1 (en) | Data processing apparatus and method for analysing transient faults occurring within storage elements of the data processing apparatus | |
| CN108491875A (en) | A kind of data exception detection method, device, equipment and medium | |
| US10884895B2 (en) | Capture of software element state changes during software application runtime and application modification based on state changes | |
| US20200250027A1 (en) | Time series forecasting classification | |
| CN113127050B (en) | Application resource packaging process monitoring method, device, equipment and medium | |
| US11126490B2 (en) | Apparatus and methods for fault detection in a system consisted of devices connected to a computer network | |
| CN114168222A (en) | Method and device for acquiring starting time, terminal equipment and storage medium | |
| US7937347B2 (en) | Method and apparatus for component association inference, failure diagnosis and misconfiguration detection based on historical failure data | |
| KR102324804B1 (en) | Monitoring system by generating context information of object in image data | |
| CN105223494A (en) | A kind of system single particle effect detection method based on parallel testing and system | |
| US10452987B2 (en) | Detecting deviations between event log and process model | |
| CN105549411B (en) | Smart machine wireless monitoring method | |
| US20140372803A1 (en) | Apparatus and method for analyzing abnormal states of component-based system | |
| CN109712061A (en) | A kind of GPU command processor robustness operation management method | |
| CN109271009A (en) | A kind of method, apparatus that control server backboard powers on and CPLD | |
| CN103577284A (en) | Abnormity detecting and recovering method for non-transparent bridge chip | |
| US10962593B2 (en) | System on chip and operating method thereof |
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: 20190503 |
|
| RJ01 | Rejection of invention patent application after publication |