CN107451022A - A kind of method and system for automatically adjusting linpack performance tests - Google Patents
A kind of method and system for automatically adjusting linpack performance tests Download PDFInfo
- Publication number
- CN107451022A CN107451022A CN201710686285.2A CN201710686285A CN107451022A CN 107451022 A CN107451022 A CN 107451022A CN 201710686285 A CN201710686285 A CN 201710686285A CN 107451022 A CN107451022 A CN 107451022A
- Authority
- CN
- China
- Prior art keywords
- linpack
- values
- cpu
- error
- performance
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/22—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
- G06F11/2289—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing by configuration test
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/22—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
- G06F11/2273—Test methods
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
Including a kind of method for automatically adjusting linpack performance tests, following steps are specifically included：Calculate the theoretical value and desired value of cpu performance；Constantly adjustment N values and NB values, realize automatically adjusting for linpack performance tests.Also include a kind of system for automatically adjusting linpack performance tests, including cpu performance value computing module, for reading CPU check figure, number and dominant frequency information, and calculate the theoretical value and desired value of cpu performance；Automated tuning module, for changing NB values, the N values of configuration file, N represents linpack computing exponent number, and NB represents the size of solution matrix piecemeal.The automated tuning of linpack performance tests can be realized, automatic configuration is realized, automatic running test, exports optimal result, it is not necessary to artificial modification configuration file, save the time resource of human resources and test, realize the simplification and automation of test.
Description
Technical field
The present invention relates to technical field of performance test, specifically a kind of side for automatically adjusting linpack performance tests
Method and system.
Background technology
Linpack (linear system software kit) has become most popular for testing high-performance calculation in the world now
The benchmark (benchmark) of machine system floating-point performance.
With the continuous upgrading of server product hardware, whole computing capability is also lifted with the speed of the order of magnitude.Weighing apparatus
An important indicator for measuring computing power is exactly to calculate peak value, such as Floating-point Computation peak value, and it refers to computer energy each second
The Floating-point Computation maximum times of completion.Actual measurement floating-point peak value refers to Linpack test values, that is to say, that is run on this machine
Linpack test programs, the optimal test result obtained by various tuning methods, the tuning method of most common of which are exactly
The configuration file for changing linpack carries out tuning.
Modification configuration file is manually modified mostly at present, not only loses time to be also easy to malfunction.
The content of the invention
In order to solve the problems, such as proposed in background technology, automatically adjust linpack performances the invention provides one kind and survey
The method and system of examination.
The technical scheme adopted by the invention to solve the technical problem is that：It is a kind of to automatically adjust linpack performance tests
Method, specifically include following steps：
Calculate the theoretical value and desired value of cpu performance；
Constantly adjustment N values and NB values, realize automatically adjusting for linpack performance tests.
Further, also needed to before above-mentioned steps obtain linpack path, call redhat systems under/
Proc/cpuinfo files, to obtain CPU check figure, number, dominant frequency.
Further, the computational methods of the theoretical value of cpu performance are：Theoretical value=CPU number * CPU check figure * master
Frequency * CPU each clock cycle performs the number of floating-point operation；
The computational methods of the desired value of cpu performance are：Desired value=theoretical value * 90%.
Further, realize that the method automatically adjusted specifically includes following steps：
The initial value of N values is set, performance test is carried out to linpack, with this determination NB optimal value；
According to NB optimal value, N values are constantly adjusted, linpack is reached best performance.
Further, the method for constantly adjustment N values specifically includes following steps：
Set error and derror (t)/dt initial value；
Error and derror (t)/dt is substituted into formula N1=N+Kp*error+KD*derror (t)/dt, obtains new N
Value：N1；In formula, KD is the rate of change of difference, and Kp is the direct ratio of the proportional parts and error in controller output, and error is
Error；
Linpack is tested with N1, obtains test result values；
Error=test result values-desired value is made, judges whether error is more than 0, if greater than 0, then output result；It is no
Then, perform and operate in next step；
New error values are substituted into formula N1=N+Kp*error+KD*derror (t)/dt, obtain new N values again：
N2, continue with N2 and linpack is tested, until the new error values constantly measured are more than 0.
A kind of system for automatically adjusting linpack performance tests, mould is calculated using described method, including cpu performance value
Block, for reading CPU check figure, number and dominant frequency information, and calculate the theoretical value and desired value of cpu performance；With,
Automated tuning module, for changing NB, N value of configuration file, N represents linpack computing exponent number, and NB represents to ask
The size of dematrix piecemeal.
Further, in addition to file path chooses module, for obtaining the absolute path of linpack softwares, and by road
Footpath with a variable return to backstage, call redhat systems under /proc/cpuinfo files, to obtain CPU check figure,
Number, dominant frequency；With,
As a result output module, for by linPack test result is exported to display device.
The effect provided in the content of the invention is only the effect of embodiment, rather than whole effects that invention is all, above-mentioned
A technical scheme in technical scheme has the following advantages that or beneficial effect：
The automated tuning of linpack performance tests can be realized, realizes automatic configuration, automatic running test, output is most
Excellent result, it is not necessary to artificial modification configuration file, save the time resource of human resources and test, realize the simplification of test
And automation.
Brief description of the drawings
Fig. 1 is the method flow schematic diagram of the embodiment of the present invention；
Fig. 2 is the system connection diagram of the embodiment of the present invention.
Embodiment
In order to the technical characterstic of clear explanation this programme, below by embodiment, and its accompanying drawing is combined, to this
Invention is described in detail.Following disclosure provides many different embodiments or example is used for realizing the different knots of the present invention
Structure.In order to simplify disclosure of the invention, hereinafter the part and setting of specific examples are described.In addition, the present invention can be with
Repeat reference numerals and/or letter in different examples.This repetition is that for purposes of simplicity and clarity, itself is not indicated
Relation between various embodiments are discussed and/or set.It should be noted that part illustrated in the accompanying drawings is not necessarily to scale
Draw.Present invention omits the description to known assemblies and treatment technology and process to avoid being unnecessarily limiting the present invention.
As shown in figure 1, a kind of method for automatically adjusting linpack performance tests, specifically includes following steps：
Step 1) calculates the theoretical value and desired value of cpu performance；
Step 2) constantly adjusts N values and NB values, realizes automatically adjusting for linpack performance tests.
Also needed to before above-mentioned steps obtain linpack path, call redhat systems under /proc/cpuinfo
File, to obtain CPU check figure, number, dominant frequency.Passage path is needed to choose the absolute road that module obtains linpack softwares
Footpath, so that user is tested.
The computational methods of the theoretical value of cpu performance are：Theoretical value=CPU number * CPU check figure * dominant frequency * CPU are each
Clock cycle performs the number of floating-point operation；If the CPU for testing purely platforms, CPU each clock cycle perform floating
The number of point processing is constant 32, and the number that CPU each clock cycle performs floating-point operation is relevant with platform.
The computational methods of the desired value of cpu performance are：Desired value=theoretical value * 90%.Server test passes through standard knots
Fruit is 90% or so of theoretical value, therefore it is the 90% of theoretical value to set the expected result tested.
Realize that the method automatically adjusted specifically includes following steps：
Step 21) sets the initial value of N values, performance test is carried out to linpack, with this determination NB optimal value；
Step 22) constantly adjusts N values, linpack is reached best performance according to NB optimal value.
The method of constantly adjustment N values specifically includes following steps：
Step 221) sets error and derror (t)/dt initial value；Error initial value is defaulted as 1, Derror
(t)/dt initial value is defaulted as 0.
Error and derror (t)/dt is substituted into formula N1=N+Kp*error+KD*derror (t)/dt by step 222),
Obtain new N values：N1；In formula, KD be difference rate of change, Kp be controller output in proportional parts and error just
Than error is error；
Step 223) is tested linpack with N1, obtains test result values；It is calculated after new N values, can be certainly
Under the dynamic catalogue into where linpack configuration files HPL.dat, HPL.dat configuration files are changed, parameter N value is changed
For the new N values being calculated, after modification preserves, linpack execution file is called to be tested, the result of test is reset
To be stored in/home catalogues below, linpack is completed in test can export end of test, and script judges to survey according to the character string
Examination is completed, into/home catalogues under read test destination file, file content is screened, first looked in file
Gflops printed words are positioned, and the last numeral for obtaining next line is result
Step 224) makes error=test result values-desired value, whether judges error more than 0, if greater than 0, then defeated
Go out result；Otherwise, perform and operate in next step；
New error values are substituted into formula N1=N+Kp*error+KD*derror (t)/dt by step 225), are obtained again
New N values：N2, continue with N2 and linpack is tested, until the new error values constantly measured are more than 0.
N is linpack computing exponent number, and N is bigger, and the effectively shared ratio of calculating is bigger, and system floating-point process performance is got over
Height, but N increase can cause the consumption of internal memory to increase, and once memory headroom is insufficient, using caching, performance can be greatly lowered.
NB is to solve for the size of the partitioning of matrix, to improve the locality of data, so as to improve overall performance, piecemeal it is big
Small to have a great impact to performance, NB*8 must be Cacheline multiple, and in being tested by experience, NB is generally 384,
192,256 three class values, it is array that NB is set in the software【384,192,256】.
For KD and KP value, rule of thumb can inquire about in table；Value such as KD can be inquired about from table 1：
| Rate of change error | <100 | 100-500 | 500-1800 | 1800-2500 | 2500> |
| <10 | PB | PB | PB | z | z |
| 10-20 | PB | PS | PS | z | z |
| 20-25 | PB | PS | z | NS | NB |
| 25-30 | z | z | NS | NS | NB |
| 30> | z | z | NB | NB | NB |
Table 1
NB, PB, Z, NS in table 1 are specific numerical value, are replaced herein using NB, PB, Z, NS.
KP value can inquire about from table 2：
Table 2
NB, PB, Z, NS in table 2 are concrete numerical value, are replaced herein using NB, PB, Z, NS.
As shown in Fig. 2 a kind of system for automatically adjusting linpack performance tests, including cpu performance value computing module, use
In reading CPU check figure, number and dominant frequency information, and calculate the theoretical value and desired value of cpu performance；With,
Automated tuning module, for changing NB values, the N values of configuration file, N represents linpack computing exponent number, and NB is represented
The size of solution matrix piecemeal；With,
File path chooses module, is returned for obtaining the absolute path of linpack softwares, and by path with a variable
To backstage, call redhat systems under /proc/cpuinfo files, to obtain CPU check figure, number, dominant frequency；With,
As a result output module, for by linPack test result is exported to display device.
Simply the preferred embodiment of the present invention described above, for those skilled in the art,
Without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications are also regarded as this hair
Bright protection domain.
Claims (7)
1. a kind of method for automatically adjusting linpack performance tests, it is characterized in that, specifically include following steps：
Calculate the theoretical value and desired value of cpu performance；
Constantly adjustment N values and NB values, realize automatically adjusting for linpack performance tests.
2. according to the method for claim 1, it is characterized in that, also needed to before above-mentioned steps obtain linpack path,
Under calling redhat systems /proc/cpuinfo files, to obtain CPU check figure, number, dominant frequency.
3. according to the method for claim 1, it is characterized in that, the computational methods of the theoretical value of cpu performance are：Theoretical value=
CPU number * CPU check figure * dominant frequency * CPU each clock cycle performs the number of floating-point operation；
The computational methods of the desired value of cpu performance are：Desired value=theoretical value * 90%.
4. according to the method for claim 1, it is characterized in that, realize that the method automatically adjusted specifically includes following steps：
The initial value of N values is set, performance test is carried out to linpack, with this determination NB optimal value；
According to NB optimal value, N values are constantly adjusted, linpack is reached best performance.
5. according to the method for claim 4, it is characterized in that, the method for constantly adjusting N values specifically includes following steps：
Set error and derror (t)/dt initial value；
Error and derror (t)/dt is substituted into formula N1=N+Kp*error+KD*derror (t)/dt, obtains new N values：
N1；In formula, KD is the rate of change of difference, and Kp is the direct ratio of the proportional parts and error in controller output, and error is to miss
Difference；
Linpack is tested with N1, obtains test result values；
Error=test result values-desired value is made, judges whether error is more than 0, if greater than 0, then output result；Otherwise,
Perform and operate in next step；
New error values are substituted into formula N1=N+Kp*error+KD*derror (t)/dt, obtain new N values again：N2, after
It is continuous that linpack is tested using N2, until the new error values constantly measured are more than 0.
6. a kind of system for automatically adjusting linpack performance tests, using the method described in claim 1 to 5 any one, its
It is characterized in, including cpu performance value computing module, for reading CPU check figure, number and dominant frequency information, and calculate cpu performance
Theoretical value and desired value；With,
Automated tuning module, for changing NB values, the N values of configuration file, N represents linpack computing exponent number, and NB represents to solve
The size of the partitioning of matrix.
7. system according to claim 6, it is characterized in that, in addition to file path selection module, for obtaining linpack
The absolute path of software, and path is returned to from the background with a variable, under calling redhat systems /proc/cpuinfo texts
Part, to obtain CPU check figure, number, dominant frequency；With,
As a result output module, for by linPack test result is exported to display device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710686285.2A CN107451022B (en) | 2017-08-11 | 2017-08-11 | A kind of method and system automatically adjusting linpack performance test |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710686285.2A CN107451022B (en) | 2017-08-11 | 2017-08-11 | A kind of method and system automatically adjusting linpack performance test |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107451022A true CN107451022A (en) | 2017-12-08 |
| CN107451022B CN107451022B (en) | 2019-07-30 |
Family
ID=60491933
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710686285.2A Active CN107451022B (en) | 2017-08-11 | 2017-08-11 | A kind of method and system automatically adjusting linpack performance test |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107451022B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109976708A (en) * | 2019-03-28 | 2019-07-05 | 苏州中晟宏芯信息科技有限公司 | A kind of method and device extracted suitable for CPU floating-point operation theoretical value |
| CN115333948A (en) * | 2022-08-23 | 2022-11-11 | 四川通信科研规划设计有限责任公司 | Method for improving network utilization rate based on cloud computing and transmission network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521119A (en) * | 2011-11-15 | 2012-06-27 | 浪潮电子信息产业股份有限公司 | Method for rapidly detecting cluster parallel efficiency |
| CN105589778A (en) * | 2016-01-12 | 2016-05-18 | 浪潮电子信息产业股份有限公司 | Method for automatically calculating CPU floating point operation theoretical value under linux system |
| US20170052857A1 (en) * | 2015-08-20 | 2017-02-23 | Qsigma, Inc. | Simultaneous Multi-Processor Apparatus Applicable to Acheiving Exascale Performance for Algorithms and Program Systems |
-
2017
- 2017-08-11 CN CN201710686285.2A patent/CN107451022B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521119A (en) * | 2011-11-15 | 2012-06-27 | 浪潮电子信息产业股份有限公司 | Method for rapidly detecting cluster parallel efficiency |
| US20170052857A1 (en) * | 2015-08-20 | 2017-02-23 | Qsigma, Inc. | Simultaneous Multi-Processor Apparatus Applicable to Acheiving Exascale Performance for Algorithms and Program Systems |
| CN105589778A (en) * | 2016-01-12 | 2016-05-18 | 浪潮电子信息产业股份有限公司 | Method for automatically calculating CPU floating point operation theoretical value under linux system |
Non-Patent Citations (1)
| Title |
|---|
| 孟金涛，贺鹏程，刘涛: "Nehalem平台的Linpack参数训练与优化", 《华中科技大学学报（自然科学版）》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109976708A (en) * | 2019-03-28 | 2019-07-05 | 苏州中晟宏芯信息科技有限公司 | A kind of method and device extracted suitable for CPU floating-point operation theoretical value |
| CN109976708B (en) * | 2019-03-28 | 2023-07-11 | 北京市合芯数字科技有限公司 | Method and device suitable for extracting floating point operation theoretical value of CPU |
| CN115333948A (en) * | 2022-08-23 | 2022-11-11 | 四川通信科研规划设计有限责任公司 | Method for improving network utilization rate based on cloud computing and transmission network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107451022B (en) | 2019-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107861606A (en) | A kind of heterogeneous polynuclear power cap method by coordinating DVFS and duty mapping | |
| CN101996117B (en) | Computer test system and method thereof | |
| NZ587241A (en) | Method, computer program product and system for enabling clinical decision support | |
| CN104461459B (en) | Fixed-point processing method and device | |
| CN111723018A (en) | Performance pressure testing method, device, equipment and storage medium | |
| CN107451022A (en) | A kind of method and system for automatically adjusting linpack performance tests | |
| CN104794058A (en) | Testing method for cloud desktop virtual environment performance | |
| CN113626297A (en) | Server energy efficiency evaluation method and related device | |
| Zeni et al. | Optimized implementation of the hpcg benchmark on reconfigurable hardware | |
| CN110333933A (en) | A kind of HPL computation model emulation mode | |
| CN105022693B (en) | A kind of performance test methods and device of application program | |
| CN106611021A (en) | Data processing method and equipment | |
| CN112947080B (en) | Scene parameter transformation-based intelligent decision model performance evaluation system | |
| CN115794570A (en) | Pressure testing method, device, equipment and computer readable storage medium | |
| CN103116617B (en) | A kind of disposal route of integrated circuit test data | |
| CN205721749U (en) | Tool dynamically sets the structure of error correcting code configuration | |
| Onnebrink et al. | Black box power estimation for digital signal processors using virtual platforms | |
| CN107729193A (en) | A kind of Linpack test Automatic Optimal collocation methods and system based on Purley platforms | |
| CN114021833A (en) | Line loss prediction method, system, storage medium and computing device | |
| CN110298690B (en) | Object class purpose period judging method, device, server and readable storage medium | |
| CN103793339B (en) | Data Cache performance heuristic approach based on internal storage access storehouse distance | |
| CN110377971A (en) | A kind of chip drives EQ value optimal value determines method and device | |
| CN102385645B (en) | Method for correcting device mismatch of capacitor | |
| CN107408060A (en) | Multiple Multi-dimension calculation frame structures for big data application | |
| CN101470650A (en) | Method and apparatus for checking computer mainboard |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |