CN105389220A - Method for performing parallel linear algebraic calculation in interactive R language platform - Google Patents
Method for performing parallel linear algebraic calculation in interactive R language platform Download PDFInfo
- Publication number
- CN105389220A CN105389220A CN201510755923.2A CN201510755923A CN105389220A CN 105389220 A CN105389220 A CN 105389220A CN 201510755923 A CN201510755923 A CN 201510755923A CN 105389220 A CN105389220 A CN 105389220A
- Authority
- CN
- China
- Prior art keywords
- language
- interactive
- platform
- parallel linear
- linear algebra
- 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
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000004364 calculation method Methods 0.000 title claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims abstract description 65
- 230000006870 function Effects 0.000 claims abstract description 25
- 238000005316 response function Methods 0.000 claims description 10
- 238000005516 engineering process Methods 0.000 claims description 9
- 238000013461 design Methods 0.000 claims description 5
- 230000007547 defect Effects 0.000 abstract 1
- 230000007613 environmental effect Effects 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 230000007812 deficiency Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/547—Remote procedure calls [RPC]; Web services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
- G06F9/5022—Mechanisms to release resources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/54—Indexing scheme relating to G06F9/54
- G06F2209/541—Client-server
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Devices For Executing Special Programs (AREA)
Abstract
The present invention discloses a method for performing parallel linear algebraic calculation in an interactive R language platform. The method comprises the following steps: providing two calculation platforms, wherein one calculation platform is an interactive R language platform, and the other calculation platform is a parallel linear algebraic calculation platform, and the two calculation platforms communicate with each other via a computer network; then, in the interactive R language platform, designing and implementing a parallel linear algebraic calculation application program interface; and finally, in a distributed matrix class of the parallel linear algebraic calculation application program interface, comprising a member variable of an R environment type, and in the initialization process of an object of the distributed matrix class, by using an reg.finalizer function of the R language, registering a garbage collection response mode of the member variable to a garbage collector of the interactive R language platform. According to the method for performing parallel linear algebraic calculation in the interactive R language platform, which is provided by the present invention, the defect that parallel linear algebraic calculation cannot be carried out in an existing interactive R language programming platform is overcome, and calculation capability of the interactive R language platform is extended.
Description
Technical field
The present invention relates to parallel computing, particularly relate to the method that one can carry out parallel linear algebra calculating in interactive (interactive) R language platform.
Background technology
R language is widely used a kind of programming language in data science field.R language is that computer user provides a large amount of conventional statistical computation function, and supports that the function of user oneself coding to R language is expanded.R language itself provides batch processing function library, and they constitute original R language platform.User can expand R language by oneself coding, and the program that user writes generally extends in R language platform with the form of R lingware bag.R language is programmed for main programming paradigm with functional expression, supports the modern procedures methods for designing such as Object-oriented Programming Design simultaneously.
R language platform supports batch processing (batch) to run simultaneously and interactive (interactive) runs two kinds of methods of operation.The interactive method of operation provides an interactively order line control desk for computer user, and user alternatively can input instruction in R platform, and R language platform carries out calculating and replying user according to instruction.The interactively method of operation allows user to design while adjust, and obtains feedback closely in real time, makes mistake and weak point be corrected in time and supplement, facilitates the performance history of R language program, extensively by the welcome of data science man.According to O ' Reilly publishing house of the U.S. and the investigation of KDNuggets website, R language is widely used in data science man community.Because R language platform is unicomputer Environment Design at first, therefore basic R language platform itself cannot make full use of the computation capability that multiple processor (CPU) or multiple stage computing machine provide.At the large data age that data scale is increasing, basic R language platform is limited to the computing power of unicomputer and cannot processes large-scale data analysis task.The limited main manifestations of basic R language platform processing power is exactly: basic R language platform cannot process large-scale linear algebra computational problem (such as extensive matrix multiplication).How to expand basic R language platform and can carry out extensive linear algebra calculating, become the major issue that one, computing technique field needs to solve.And extensive linear algebra calculating is mainly solved by the means of parallel computing at present.
On the other hand, the parallel computing of existing Effect-based operation passing interface (MPI), can solve extensive linear algebra computational problem.The software library ScaLAPACK of Effect-based operation passing interface (MPI) technology provides one group of application programming interfaces (API) function, covers most linear algebra computation requirement.ScaLAPACK software library, by using MPI parallel computing, breaches the restriction of unicomputer computing power, can make full use of the computing power of multiple stage computing machine.Existing pbdR project (network address of this project is http://r-pbd.org/) utilizes ScaLAPACK software library to carry out Function Extension to R language platform at present, makes R language users carry out parallel linear algebra calculating.But that write based on ScaLAPACK software library or based on pbdR item development program can only be run in the mode of batch processing, and they cannot directly run in interactive R language platform.This restriction makes that the user of interactive R language platform cannot utilize ScaLAPACK storehouse, pbdR project carries out large-scale parallel linear algebra calculating.There is no method at present and can carry out parallel linear algebra calculating in interactive R language platform.
Summary of the invention
Goal of the invention: in order to overcome the deficiency cannot carrying out parallel linear algebra calculating in existing interactive R language platform, the invention provides a kind of method expanding interactive R language platform, the method makes user can carry out parallel linear algebra calculating in interactive R language platform, and without the need to understanding the specific implementation details that parallel linear algebra calculates, solve the deficiency that existing interactive R Programming with Pascal Language platform cannot carry out parallel linear algebra calculating, extend the computing power of interactive R.
Technical solution of the present invention is: in order to realize foregoing invention object, the technical solution used in the present invention is a kind of computing method based on client-server (Client-Server) model, interactive computing platform and parallel linear algebra computing platform are structurally separated by the method, are realized the cooperated computing of two platforms by computer network communication.Whole technical scheme comprises the following steps:
(1) method of the present invention provides two computing platforms, and one is interactive R language platform, and another one is parallel linear algebra computing platform, and two computing platforms are undertaken communicating by computer network, cooperated computing;
(2) in interactive R language platform, define a distributed matrix class, such provides the application programming interfaces of the parallel linear algebra calculating that can interactively run;
(3) in distributed matrix class, the member variable of R environment (environment) type is comprised;
(4) be the member variable registration garbage reclamation response function of R environmental form in distributed matrix class.
Further, two computing platforms in described step (1) are respectively: one, interactive R language platform, this platform is the interactive R language platform of a standard, and this interactive R language platform is loaded with the expansion software package that the present invention realizes, and directly and computer user carry out alternately; Two, parallel linear algebra computing platform, this computing platform be one based on MPI(message passing interface) technology and ScaLAPACK software library, achieve the computing platform of parallel linear algebra computing function.Two computing platforms are communicated mutually by computer network.Interactive R language platform accepts interactive calculation command from computer user, and send corresponding computations to parallel linear algebra computing platform, the latter carries out concrete parallel linear algebra and calculates and result of calculation returned to interactive R language platform, and by interactive R language platform by result feedback to computer user.
Further, in described step (2), " distributed matrix class " refers to S3 class or the S4 class of a R language.Such uses R language compilation, and is loaded in interactive R language platform, for user.This distributed matrix class provides one group of R language function calculated for parallel linear algebra, and user, by calling corresponding computing function on the object of this distributed matrix class, has carried out parallel linear algebra calculation task.
Further, described step indicates in (3) that " distributed matrix class " should comprise a special member variable, and the type of this member variable is R environment (environment) type.The major function of this member variable is: when the object of Garbage Collector to distributed matrix class in interactive R language platform reclaims, the response function that Garbage Collector is registered by this member variable, notice parallel linear algebra computing platform, synchronously deletes corresponding matrix data.
Further, in described step (4), the process of registration garbage reclamation response function completes in the constructed fuction of distributed matrix class, realized by the reg.finalizer function of R language itself.
The invention has the beneficial effects as follows: (1), by interactive R language platform and parallel linear algebra computing platform being separated, makes both can run with the different methods of operation simultaneously.By computer network communication, the function enabling interactive R language platform call parallel linear algebra computing platform to provide, and do not lose its interactivity.The invention solves the problem can not carrying out parallel linear algebra calculating in interactive R language platform.(2) by being packaged in the middle of a distributed matrix class by all parallel linear algebra computing functions, the present invention provides a user-friendly use interface in interactive R language platform.Computer user as operation with traditional R language unit matrix general operation distributed matrix, can carry out parallel linear algebra calculating.The R language Accounting Legend Code that the present invention writes keeps highly consistent with primary R language Accounting Legend Code, alleviates the learning cost of computer user when using of the present invention.(3) the present invention is by the reg.finalizer function of R language to Garbage Collector registration garbage reclamation response function, solves the stationary problem of the garbage reclamation of interactive R language platform and the garbage reclamation of parallel linear algebra computing platform.The technical scheme that the present invention simultaneously uses avoids in similar technique to be needed to use C language to carry out the problem of programming, and the present invention all can be realized with R Programming with Pascal Language, reduce the implementation complexity of whole technical scheme.(4) after use the present invention expands interactive R language platform, interactive R language platform can break through the processing power restriction of unicomputer, and carries out more massive linear algebra calculating.
Accompanying drawing explanation
Fig. 1 is bulk treatment schematic flow sheet of the present invention.
Fig. 2 is the performance comparison figure of the unit linear algebra computing system that the present invention and standard R language platform provide.
Embodiment
Below in conjunction with the drawings and specific embodiments, illustrate the present invention further, these embodiments should be understood only be not used in for illustration of the present invention and limit the scope of the invention, after having read the present invention, the amendment of those skilled in the art to the various equivalent form of value of the present invention has all fallen within the application's claims limited range.
Technical scheme of the present invention forms primarily of two software modules: one is interactively R language platform, and another one is parallel linear algebra computing platform.Interactive R language platform is the R language computing platform of a standard, and come from the realization (project network address is https: //www.r-project.org/) of RProject project, this software does not belong to content of the present invention.Parallel linear algebra computing platform is then a computing platform meeting following feature: (1) this computing platform can receive computations by computer network, and can resolve instruction, performs corresponding parallel linear algebra calculation task; (2) this platform can storage matrix data in a distributed manner; (3) this computing platform is by MPI(message passing interface) technology and ScaLAPACK software library carry out parallel linear algebra calculating (MPI technology and ScaLAPACK software library do not belong to content of the present invention, putting forward " ScaLAPACK software library " all acute pyogenic infection of finger tip " with the software library of ScaLAPACK application programming interfaces operating such " in instructions of the present invention); (4) identifier of result of calculation or result of calculation is returned to the computations person of sending by computer network by this computing platform.The computing platform all meeting above-mentioned 4 features all can be considered to the parallel linear algebra computing platform meeting the technology of the present invention feature.
User uses the present invention to carry out the idiographic flow of parallel linear algebra calculating as shown in Figure 1.Interactively R language platform directly carries out alternately with final computer user, receives the instruction that user gives, performs calculating, and result of calculation is returned to user by computer screen.For the parallel linear algebra computations that user assigns, interactive R language platform passes through computer network, this instruction is sent to parallel linear algebra computing platform, calculating operation is performed by ScaLAPACK software library by parallel linear algebra computing platform, and result of calculation is returned to interactive R language platform by computer network, finally by interactive R language platform, result is returned to computer user.
The initialization Booting sequence of two computing platforms that the present invention relates to comprises the following steps:
(1) user starts an interactively R language platform on their computer;
(2) user loads the R software package realized by technical solution of the present invention in interactive R language platform, and sends the instruction starting parallel linear algebra and calculate to interactive R language platform;
(3) interactive R language platform is by the Operation control mechanism of MPI technology, starts the MPI operation that parallel linear algebra computing platform is corresponding;
(4), after parallel linear algebra computing platform starts, interactive R language platform is set up computer network communication with the master computing node of parallel linear algebra computing platform and is linked;
(5) interactive R language platform creates an empty queue as overall garbage reclamation queue, and then notify that computer user's software language of the present invention platform loads complete, interactive R language platform enters the wait instruction stage;
(6) user performs parallel linear algebra calculation task by interactive R language platform.
User, in interactive R language platform, does not directly carry out programming and uses, but carry out programming realization towards distributed matrix class provided by the invention towards ScaLAPACK storehouse and MPI technology.The embodiment of distributed matrix class of the present invention is: first, uses a S3 class or S4 class to define this distributed matrix class, then provide concrete member variable by the mode of programming for such in R language; Such member variable comprises: (1) long-range matrix identifier variable, the type of this identifier in R language can be integer or character string, as long as different long-range matrix can be distinguished, the variable of (2) R environment (environment) types, this variable is used for the garbage reclamation of distributed matrix object; Then, by R language Object-oriented Programming Design mechanism, for such provides some member functions, and the existing linear algebra computing function of heavily loaded R language, make these computing functions support distributed matrix class.
In the member function or linear algebra computing function of each distributed matrix class, the step completing concrete parallel linear algebra calculation task is:
(1) interactive R language platform is from function call parameter, reads the long-range matrix identifier participating in the distributed matrix class object calculated;
(2) interactive R language platform is by computations corresponding for this linear algebra calculating operation, and all participate in calculate long-range matrix identifier, overall garbage reclamation queue, pass to parallel linear algebra computing platform by computer network;
(3) parallel linear algebra computing platform receives corresponding computations and participates in the long-range matrix identifier calculated, overall garbage reclamation queue.Parallel linear algebra computing platform then reads out its long-range matrix identifier of rubbish of preserving from overall garbage reclamation queue, from computing platform, then delete the matrix data that the long-range matrix identifier of these rubbish is corresponding.The then analytical Calculation instruction of parallel linear algebra computing platform, and according to the long-range matrix identifier participating in calculating, from internal memory, take out the distributed matrix operation handle be kept in this computing platform;
(4) parallel linear algebra computing platform carries out parallel linear algebra calculating by the distributed matrix that takes out in previous step operation handle and ScaLAPACK storehouse;
(5) parallel linear algebra computing platform is preserved this and is calculated the result produced, and this result is a distributed matrix, and is the long-range matrix identifier of distribution of results formula matrix allocation;
(6) the long-range matrix identifier that previous step generates by parallel linear algebra computing platform is packaged into replys in instruction, sends to interactive R language platform by computer network;
(7) interactive R language platform receives the reply instruction that parallel linear algebra computing platform is beamed back, and therefrom takes out long-range matrix identifier, the object of the distributed matrix class utilizing this identifier initialization one new, and this object is returned to user.User can use the object of the distributed matrix class returned, and carries out other calculation task.
When using R language definition distributed matrix class, needing to provide an initialization constructed fuction, making interactive R language platform correctly can process the garbage reclamation problem of the object of distributed matrix class.The workflow of the initialization constructed fuction of distributed matrix class comprises following steps:
(1) in calculator memory, initialization constructed fuction receives the input parameter of a long-range matrix identifier as this function;
(2) initialization constructed fuction newly-built empty distributed matrix object in internal memory;
(3) initialization constructed fuction utilizes the long-range matrix identifier member variable of distributed matrix object newly-built in the long-range matrix identifier initialization step (1) in input parameter;
(4) be the member variable of new R environment (environment) type of distributed matrix Object Creation one newly-built in step (1), and by the member variable of the long-range matrix identifier in input parameter stored in this R environmental form;
(5) by the reg.finalizer function of R language, R environmental form member variable described in (4) step is registered in the Garbage Collector of interactive R language platform, this R environmental form member variable is associated with a self-defining garbage reclamation response function;
(6) by create in step (1), again through the distributed matrix object of above-mentioned steps process, the rreturn value as constructed fuction returns.
Further, (5) step in the workflow of the initialization constructed fuction of above-mentioned distributed matrix class, refer to a self-defining garbage reclamation response function, the workflow of this function comprises following steps:
(1) self-defining garbage reclamation response function receives a R environmental form variable as input parameter;
(2) response function reads long-range matrix identifier from R environmental form variable, is joined in overall garbage reclamation queue by this identifier.
The invention has the beneficial effects as follows, make computer user can carry out parallel linear algebra calculating in interactive R language platform; By the mode of multiple stage computing machine parallel computation, when the present invention can allow user process extensive linear algebra computational problem in interactive R language platform, obtain and calculate computing velocity faster than unit linear algebra.The present invention is based on more existing open source softwares and achieve a prototype system.According to technical scheme requirement of the present invention, prototype system comprises two computing platforms, wherein interactive R language platform uses the interactive R language platform that RProject project provides, parallel linear algebra computing platform is then the prototype computing platform that technical scheme according to the present invention is developed, the software that the pbdR of employing project under development (project home page http://r-pbd.org/) provides.The software that RProject project and pbdR project provide does not belong to content of the present invention.In addition according to technical scheme of the present invention, in prototype system, also comprise the distributed matrix class that runs on interactive R language platform.By using matrix multiplication operation (a kind of linear algebra calculating operation) as benchmark test, the unit linear algebra computing system that the prototype software system realize the present invention and existing R language platform provide is tested, and evaluation and test uses calculating consuming time as Measure Indexes.In evaluation and test, the prototype software system that the present invention realizes employs 10 computing machines and carries out parallel computation, and the unit linear algebra computing system that existing R language platform provides then is limited to its function, is merely able to use single computer to calculate.The result of evaluation and test is see Fig. 2.In fig. 2, solid line represents the evaluation result of the prototype software system that the present invention realizes, and dotted line represents the evaluation result of the linear computing system of unit that existing R language platform provides.Along with the increase participating in the matrix size calculated, the prototype software system that the present invention realizes is when completing identical calculation task, and spent time is fewer than previously described unit linear algebra computing system.Evaluation and test shows that prototype software system that the present invention realizes is when carrying out extensive matrix multiplication operation, calculate consuming time shorter, computing velocity is faster, demonstrate the validity of the method that the present invention proposes, demonstrate beneficial effect of the present invention.
Claims (5)
1. in interactive R language platform, carry out a method for parallel linear algebra calculating, comprise the following steps:
(1) provide two computing platforms, one is interactive R language platform, and another one is parallel linear algebra computing platform, and two computing platforms are communicated by computer network;
(2) in interactive R language platform, design a distributed matrix class calculated for parallel linear algebra, be supplied to computer user as application programming interfaces;
(3) in described distributed matrix class, the member variable of a R language environment type is comprised;
(4) be the member variable registration garbage reclamation response function approach of described R language environment type.
2. a kind of method of carrying out parallel linear algebra calculating in interactive R language platform according to claim 1, it is characterized in that: in described step (1), the workflow of parallel linear algebra computing platform is: storage matrix data first in a distributed manner, then computations is received and analytical Calculation instruction by computer network, then perform corresponding parallel linear algebra calculating operation by MPI technology and ScaLAPACK software library, finally by computer network, the identifier of result of calculation or result of calculation is returned to the person of sending of computations.
3. a kind of method of carrying out parallel linear algebra calculating in interactive R language platform according to claim 1, is characterized in that: distributed matrix class is defined within interactive R language platform, such application programming interfaces providing parallel linear algebra to calculate.
4. a kind of method of carrying out parallel linear algebra calculating in interactive R language platform according to claim 1, is characterized in that: described distributed matrix class is S3 class in R language or S4 class.
5. a kind of method of carrying out parallel linear algebra calculating in interactive R language platform according to claim 1, it is characterized in that: the workflow of described step (4) is: in the object initialization process of distributed matrix class, use the reg.finalizer function of R language, by the member variable of the R language environment type in this object, bind with a garbage reclamation response function, and be registered in the Garbage Collector of interactive R language platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510755923.2A CN105389220B (en) | 2015-11-09 | 2015-11-09 | The method of parallel linear algebra calculating is carried out in interactive R language platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510755923.2A CN105389220B (en) | 2015-11-09 | 2015-11-09 | The method of parallel linear algebra calculating is carried out in interactive R language platform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105389220A true CN105389220A (en) | 2016-03-09 |
| CN105389220B CN105389220B (en) | 2019-02-15 |
Family
ID=55421527
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510755923.2A Active CN105389220B (en) | 2015-11-09 | 2015-11-09 | The method of parallel linear algebra calculating is carried out in interactive R language platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105389220B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022483A (en) * | 2016-05-11 | 2016-10-12 | 星环信息科技（上海）有限公司 | Method and equipment for conversion between machine learning models |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5845120A (en) * | 1995-09-19 | 1998-12-01 | Sun Microsystems, Inc. | Method and apparatus for linking compiler error messages to relevant information |
| CN1339743A (en) * | 2000-08-23 | 2002-03-13 | 国际商业机器公司 | Method and device for computer software analysis |
-
2015
- 2015-11-09 CN CN201510755923.2A patent/CN105389220B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5845120A (en) * | 1995-09-19 | 1998-12-01 | Sun Microsystems, Inc. | Method and apparatus for linking compiler error messages to relevant information |
| CN1339743A (en) * | 2000-08-23 | 2002-03-13 | 国际商业机器公司 | Method and device for computer software analysis |
Non-Patent Citations (2)
| Title |
|---|
| 胡长军，等: ""p-HPF支持多范例并行计算的并行编译技术"", 《计算机学报》 * |
| 谢军: ""P_HPF并行编译系统EXTRINSIC过程调用及对ScaLAPACK并行线性代数算法包的支持"", 《HTTP://XUESHU.BAIDU.COM/S?WD=PAPERURI%3A%285826E837B9EDCAC52EFC79230D3F0235%29&FILTER=SC_LONG_SIGN&TN=SE_XUESHUSOURCE_2KDUW22V&SC_VURL=HTTP%3A%2F%2FWWW.DOC88.COM%2FP-2045556984539.HTML&IE=UTF-8&SC_US=16634907125052348932》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022483A (en) * | 2016-05-11 | 2016-10-12 | 星环信息科技（上海）有限公司 | Method and equipment for conversion between machine learning models |
| CN106022483B (en) * | 2016-05-11 | 2019-06-14 | 星环信息科技（上海）有限公司 | The method and apparatus converted between machine learning model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105389220B (en) | 2019-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ashari et al. | Fast sparse matrix-vector multiplication on GPUs for graph applications | |
| Ballard et al. | Minimizing communication in numerical linear algebra | |
| Liu et al. | Speculative segmented sum for sparse matrix-vector multiplication on heterogeneous processors | |
| Böhm et al. | xSim: The extreme-scale simulator | |
| CN104156313A (en) | Method for automatically generating Web service testing case | |
| Hendrickson | Load balancing fictions, falsehoods and fallacies | |
| Athas | Fine grain concurrent computations | |
| JP2022177793A (en) | Operator registration method and apparatus for deep learning framework, device and storage medium | |
| Sun et al. | An I/O bandwidth-sensitive sparse matrix-vector multiplication engine on FPGAs | |
| Riener et al. | metaSMT: focus on your application and not on solver integration | |
| Izsó et al. | IncQuery-D: incremental graph search in the cloud. | |
| Kuo et al. | The state of the art in parallel production systems | |
| Sao et al. | A communication-avoiding 3D algorithm for sparse LU factorization on heterogeneous systems | |
| Zhou et al. | Towards product-level parallel computing of large-scale building information modeling data using graph theory | |
| Tribastone et al. | Automatic translation of UML sequence diagrams into PEPA models | |
| CN115858205A (en) | Memory blackboard mechanism-based simulation component interaction method, device and equipment | |
| Xiao et al. | OpenABLext: An automatic code generation framework for agent‐based simulations on CPU‐GPU‐FPGA heterogeneous platforms | |
| CN105389220A (en) | Method for performing parallel linear algebraic calculation in interactive R language platform | |
| CN112015714A (en) | Database-based data model generation method and device | |
| Davis et al. | Paradigmatic shifts for exascale supercomputing | |
| CN113778564B (en) | Method, equipment and storage medium for efficiently executing EVM intelligent contract | |
| Diez Dolinski et al. | Distributed simulation of P systems by means of map-reduce: first steps with Hadoop and P-Lingua | |
| Caragea et al. | Brief announcement: performance potential of an easy-to-program PRAM-on-chip prototype versus state-of-the-art processor | |
| Humble et al. | Software systems for high-performance quantum computing | |
| Haller et al. | A programming model and foundation for lineage-based distributed computation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Gu Rong Inventor after: Wang Zhaokang Inventor after: Huang Yihua Inventor after: Fan Shiqing Inventor before: Huang Yihua Inventor before: Wang Zhaokang Inventor before: Gu Rong Inventor before: Fan Shiqing |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |