CN107092676A - A kind of data processing method and device - Google Patents
A kind of data processing method and device Download PDFInfo
- Publication number
- CN107092676A CN107092676A CN201710253998.XA CN201710253998A CN107092676A CN 107092676 A CN107092676 A CN 107092676A CN 201710253998 A CN201710253998 A CN 201710253998A CN 107092676 A CN107092676 A CN 107092676A
- Authority
- CN
- China
- Prior art keywords
- data
- row
- pending
- parquet
- pending data
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000000034 method Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims description 21
- 241001269238 Data Species 0.000 claims description 19
- 238000005520 cutting process Methods 0.000 claims description 14
- 230000006837 decompression Effects 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 2
- ZLIBICFPKPWGIZ-UHFFFAOYSA-N pyrimethanil Chemical compound CC1=CC(C)=NC(NC=2C=CC=CC=2)=N1 ZLIBICFPKPWGIZ-UHFFFAOYSA-N 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/25—Integrating or interfacing systems involving database management systems
- G06F16/258—Data format conversion from or to a database
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/221—Column-oriented storage; Management thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
- G06F3/061—Improving I/O performance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0638—Organizing or formatting or addressing of data
- G06F3/064—Management of blocks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Human Computer Interaction (AREA)
- Software Systems (AREA)
- Devices For Executing Special Programs (AREA)
Abstract
The invention discloses a kind of data processing method and device, this method includes:The pending data stored in Distributed Computing Platform is converted into the Parquet forms by row storage;The pending data of Parquet forms is loaded into internal memory by row group using Spark, and pending data is decompressed and sent after unserializing to the corresponding calculating of FPGA Heterogeneous Computings device progress in internal memory.In technical scheme disclosed in the present application, the pending data stored in Distributed Computing Platform is first converted into the Parquet forms by row storage, the pending data of Parquet forms is loaded into internal memory according still further to row group and sent after relevant treatment to the progress correspondence calculating of FPGA Heterogeneous Computings device.The pending data of Parquet forms is read out according to row group in the application, the data processing speed of FPGA Heterogeneous Computing devices is substantially increased.
Description
Technical field
The present invention relates to deep learning isomery acceleration technique field, more specifically to a kind of data processing method and
Device.
Background technology
The development of big data and high-performance calculation platform greatly advances the paces of deep learning, in deep learning,
When data set and very big parameter model, culture large-scale model is just needed by distributed computing framework.Apache Spark
There is unsurpassed internal memory to calculate and expansible Distributed Calculation performance, can effectively improve the super ginseng in deep learning
Number tuning and the deployment of extensive model.
In two critical activitys of deep learning:, it is necessary to very high inherent degree of parallelism, substantial amounts of floating-point in classification and convolution
Computing capability and matrix computations.GPU and FPGA relies on its outstanding isomery speed-up computation and Floating-point Computation ability, greatly carries
The high speed and performance of deep learning, relative CPU processing have required server less, the characteristics of power consumption is lower.
Spark, by parsing after metadata, stores a relative displacement for the data of variable-length in internal memory
And length, the pointer for pointing to the row data stored with this length is served as using UnsafeRow examples, when handling data line,
Just digital independent into internal memory;And FPGA can only temporarily read in this data line when accelerating data processing.Invention human hair
Existing, this data reading mode accelerates Spark to calculate for FPGA can concurrency apparently without make full use of FPGA so that
FPGA data reading speed is slower, and the processing speed for further resulting in FPGA for data is slower.
In summary, how to provide it is a kind of can make full use of FPGA can concurrency to improve its data processing speed
Technical scheme, is current those skilled in the art's urgent problem to be solved.
The content of the invention
It is an object of the invention to provide a kind of data processing method and device, with make full use of FPGA can concurrency to carry
Its high data processing speed.
To achieve these goals, the present invention provides following technical scheme:
A kind of data processing method, including:
The pending data stored in Distributed Computing Platform is converted into the Parquet forms by row storage;
The pending data of the Parquet forms is loaded into internal memory by row group using Spark, and it is right in internal memory
The pending data carries out decompression and sent after unserializing to the corresponding calculating of FPGA Heterogeneous Computings device progress.
It is preferred that, pending data is converted into the Parquet forms by row storage, including:
Generation includes the Parquet files of whole pending datas, and in this document cuts whole pending datas
Be divided into lines group, wherein in each row group the row comprising all pending datas row block, and either rank block is stored with pair
The pending data of corresponding part in should arranging.
It is preferred that, whole pending data cuttings are embarked on journey group in the Parquet files, including:
Whole pending data cuttings are embarked on journey group in the Parquet files, wherein the size of each row group
It is identical with the size of data block in the Distributed Computing Platform.
It is preferred that, the pending data of the Parquet forms is loaded into internal memory by row group using Spark, including:
Using advanced with Spark c language setting function by the pending data of the Parquet forms by row group
It is loaded into internal memory.
It is preferred that, will in internal memory carry out decompression and unserializing after pending data to FPGA Heterogeneous Computings device it
Afterwards, in addition to:
Obtain the result of calculation of the FPGA Heterogeneous Computings device return and the result of calculation returned into the Spark,
The result of calculation is that the FPGA Heterogeneous Computings device calculates what is obtained using the pending data progress correspondence received.
A kind of data processing equipment, including:
Modular converter, is used for:The pending data stored in Distributed Computing Platform is converted into by row storage
Parquet forms;
Processing module, is used for:The pending data of the Parquet forms is loaded into internal memory by row group using Spark
In, and in internal memory the pending data is decompressed and unserializing after send to FPGA Heterogeneous Computings device and carry out pair
It should calculate.
It is preferred that, the modular converter includes:
Converting unit, is used for:Generation includes the Parquet files of whole pending datas, and in this document will be complete
Portion's pending data cutting is embarked on journey group, wherein in each row group the row comprising all pending datas row block, it is and any
Row block is stored with the pending data of corresponding part in respective column.
It is preferred that, the converting unit includes:
Conversion subunit, is used for:Whole pending data cuttings are embarked on journey group in the Parquet files, wherein often
The size of the individual row group is identical with the size of data block in the Distributed Computing Platform.
It is preferred that, the processing module includes:
Loading unit, is used for:Utilize function the treating the Parquet forms that the setting of c language is advanced with Spark
Processing data is loaded into internal memory by row group.
It is preferred that, in addition to:
Module is returned, is used for:The pending data after decompression and unserializing will be carried out in internal memory to FPGA isomery meters
Calculate after device, obtain the result of calculation of the FPGA Heterogeneous Computings device return and return to the result of calculation described
Spark, the result of calculation is that the FPGA Heterogeneous Computings device carries out corresponding calculate using the pending data received
Arrive.
The invention provides a kind of data processing method and device, wherein this method includes:By in Distributed Computing Platform
The pending data of storage is converted into the Parquet forms by row storage;Using Spark by the Parquet forms wait locate
Reason data are loaded into internal memory by row group, and in internal memory the pending data is decompressed and unserializing after send to
FPGA Heterogeneous Computings device carries out correspondence calculating.In technical scheme disclosed in the present application, it will first be stored in Distributed Computing Platform
Pending data be converted into by row storage Parquet forms, the pending data of Parquet forms is added according still further to row group
It is loaded into internal memory send after relevant treatment to FPGA Heterogeneous Computings device and carries out correspondence calculating.Different from needing in the prior art
Want to read the pending data of Parquet forms according to row group in the reading manner of data line data line, the application
Take, can make full use of FPGA Heterogeneous Computing devices can concurrency so that the digital independent speed of FPGA Heterogeneous Computing devices
Degree is very fast, and then substantially increases the data processing speed of FPGA Heterogeneous Computing devices.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to be briefly described, it should be apparent that, drawings in the following description are only this
The embodiment of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can also basis
The accompanying drawing of offer obtains other accompanying drawings.
Fig. 1 is a kind of flow chart of data processing method provided in an embodiment of the present invention;
Fig. 2 is a kind of structural representation of data processing equipment provided in an embodiment of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
Referring to Fig. 1, it illustrates a kind of flow chart of data processing method provided in an embodiment of the present invention, can wrap
Include:
S11:The pending data stored in Distributed Computing Platform is converted into the Parquet forms by row storage.
Whether it is by row storage that the pending data being stored in Distributed Computing Platform can be first determined in the application
Parquet forms, if it is, not processing, if it is not, then needing these pending datas being converted into by row storage
Parquet forms.The data of Parquet forms compare line be stored with reduction disk storage space, supporting vector computing, obtain
The advantage of more preferable scan performance.The executive agent of each step can be number in a kind of data processing method disclosed in the present application
According to processing unit, and the data processing equipment can be integrated in Spark, thus the application collect each step executive agent it is equal
Can be Spark.
S12:The pending data of Parquet forms is loaded into internal memory by row group using Spark, and it is right in internal memory
Pending data carries out decompression and sent after unserializing to the corresponding calculating of FPGA Heterogeneous Computings device progress.
Pending data is converted into after the Parquet forms by row storage, then when needing to load it in internal memory then
It directly can be added into internal memory by row group using Spark, and then the correspondence provided in Parquet function libraries is provided in internal memory
Interface decompressed and the relevant treatment such as unserializing after send to FPGA Heterogeneous Computings device so that FPGA Heterogeneous Computing devices
Corresponding calculate is carried out to the data received.
In technical scheme disclosed in the present application, first the pending data stored in Distributed Computing Platform is converted into by row
The pending data of Parquet forms, is loaded into internal memory and carries out relevant treatment by the Parquet forms of storage according still further to row group
After send to FPGA Heterogeneous Computings device carry out correspondence calculating.Reading different from needing data line data line in the prior art
Take in mode, the application and the pending data of Parquet forms is read out according to row group, FPGA can be made full use of different
Structure computing device can concurrency so that the data reading speed of FPGA Heterogeneous Computing devices is very fast, and then substantially increases
The data processing speed of FPGA Heterogeneous Computing devices.
A kind of data processing method provided in an embodiment of the present invention, pending data is converted into by row storage
Parquet forms, can include:
Generation includes the Parquet files of whole pending datas, and in this document cuts whole pending datas
Be divided into lines group, wherein in each row group the row comprising all pending datas row block, and either rank block is stored with respective column
The pending data of middle corresponding part.
It is then specifically in Parquet it should be noted that pending data is converted into the Parquet forms by row storage
All pending data levels are cut into Row group (row group) in file, a Row group includes correspondence in all row
Partial column chunk (row block), a column chunk stores the pending data of the row, namely horizontal cutting is got
To row group, and group inside is gone then by row cutting, so that the part row block of each all row of the row group comprising pending data, and institute
There is row group to be grouped together and can be obtained by whole pending datas;Two are cut into if desired for the pending data for arranging 10 rows 4
Individual row group, then corresponding data are as first row group in arranging preceding 5 row and this 5 row 4, and then 5 rows and this 5 row are in 4 arrange
Corresponding data as second row group, finally give two be 5 rows 4 row row group.So as to quick square in this way
Just pending data is converted into Parquet forms, further ensure data processing method provided in an embodiment of the present invention
Smooth realization.
A kind of data processing method provided in an embodiment of the present invention, cuts whole pending datas in Parquet files
Be divided into lines group, can include:
Whole pending data cuttings are embarked on journey group in Parquet files, wherein the size of each row group with it is distributed
The size of data block is identical in calculating platform.
It should be noted that row group size can be configured according to actual needs, be set in the application with
Data block size in Distributed Computing Platform is identical, without carrying out cutting or combination to data block, simplifies operating process.
A kind of data processing method provided in an embodiment of the present invention, using Spark by the pending data of Parquet forms
It is loaded into internal memory, can includes by row group:
The pending data of Parquet forms is loaded by row group using the function that the setting of c language is advanced with Spark
Enter in internal memory.
It should be noted that technical scheme provided in an embodiment of the present invention can be realized based on c language, wherein can be advance
Corresponding function realization is write using c language pending data is read and is loaded into internal memory and subsequently treated by capable organize
Processing transmission of processing data etc..Specifically, the application realizes a kind of data model being present in internal memory with c language,
The model is based on parquet storage formats, and with reference to Java Native Interface, (JNI, JNI provide scala and C
The communication of language and OpenCL, it is allowed to which code and data are interacted between different language), it is possible to achieve Parquet files
The functions such as metadata parsing, digital independent, decompression, unserializing simultaneously can provide the interface interacted with Scala.So as to
Using c language it is simple and quick realize technical scheme disclosed in the present application.
A kind of data processing method provided in an embodiment of the present invention, will carry out decompression and treating after unserializing in internal memory
After processing data to FPGA Heterogeneous Computing devices, it can also include:
Obtain the result of calculation of FPGA Heterogeneous Computings device return and the result of calculation is returned into Spark, result of calculation
Correspondence is carried out using the pending data received calculate what is obtained for FPGA Heterogeneous Computings device.
Pending data after will be treated is sent to FPGA Heterogeneous Computing devices, the calculation in FPGA Heterogeneous Computing devices
Method is carried out after corresponding data processing, feeds back to Spark, is easy to Spark to be based on result of calculation and is carried out next step operation.
It is further to note that in embodiments of the invention for corresponding technical scheme realization principle in the prior art
Consistent part is simultaneously unspecified, in order to avoid excessively repeat.
The embodiment of the present invention additionally provides a kind of data processing equipment, as shown in Fig. 2 can include:
Modular converter 11, is used for:The pending data stored in Distributed Computing Platform is converted into by row storage
Parquet forms;
Processing module 12, is used for:The pending data of Parquet forms is loaded into internal memory by row group using Spark,
And pending data is decompressed in internal memory and sent after unserializing to the corresponding calculating of FPGA Heterogeneous Computings device progress.
A kind of data processing equipment provided in an embodiment of the present invention, modular converter can include:
Converting unit, is used for:Generation includes the Parquet files of whole pending datas, and in this document will be complete
Portion's pending data cutting is embarked on journey group, wherein in each row group the row comprising all pending datas row block, and either rank block
Be stored with the pending data of corresponding part in respective column.
A kind of data processing equipment provided in an embodiment of the present invention, converting unit can include:
Conversion subunit, is used for:Whole pending data cuttings are embarked on journey group in Parquet files, wherein each row
The size of group is identical with the size of data block in Distributed Computing Platform.
A kind of data processing equipment provided in an embodiment of the present invention, processing module can include:
Loading unit, is used for:Using advancing with the function of c language setting in Spark by the pending of Parquet forms
Data are loaded into internal memory by row group.
A kind of data processing equipment provided in an embodiment of the present invention, can also include:
Module is returned, is used for:The pending data after decompression and unserializing will be carried out in internal memory to FPGA isomery meters
Calculate after device, obtain the result of calculation of FPGA Heterogeneous Computings device return and the result of calculation is returned into Spark, calculate knot
Fruit carries out correspondence using the pending data received for FPGA Heterogeneous Computings device and calculates what is obtained.
The explanation of relevant portion refers to the embodiment of the present invention in a kind of data processing equipment provided in an embodiment of the present invention
The detailed description of corresponding part, will not be repeated here in a kind of data processing method provided.
The foregoing description of the disclosed embodiments, enables those skilled in the art to realize or using the present invention.To this
A variety of modifications of a little embodiments will be apparent for a person skilled in the art, and generic principles defined herein can
Without departing from the spirit or scope of the present invention, to realize in other embodiments.Therefore, the present invention will not be limited
It is formed on the embodiments shown herein, and is to fit to consistent with features of novelty with principles disclosed herein most wide
Scope.
Claims (10)
1. a kind of data processing method, it is characterised in that including:
The pending data stored in Distributed Computing Platform is converted into the Parquet forms by row storage;
The pending data of the Parquet forms is loaded into internal memory by row group using Spark, and to described in internal memory
Pending data carries out decompression and sent after unserializing to the corresponding calculating of FPGA Heterogeneous Computings device progress.
2. according to the method described in claim 1, it is characterised in that pending data is converted into the Parquet by row storage
Form, including:
Generation includes the Parquet files of whole pending datas, and is in this document cut into whole pending datas
Row group, wherein in each row group the row comprising all pending datas row block, and either rank block is stored with respective column
The pending data of middle corresponding part.
3. method according to claim 2, it is characterised in that by whole pending datas in the Parquet files
Cutting is embarked on journey group, including:
Whole pending data cuttings are embarked on journey group in the Parquet files, wherein size and the institute of each row group
The size for stating data block in Distributed Computing Platform is identical.
4. method according to claim 3, it is characterised in that utilize Spark by the pending number of the Parquet forms
It is loaded into according to by row group in internal memory, including:
The pending data of the Parquet forms is loaded by row group using the function that the setting of c language is advanced with Spark
Enter in internal memory.
5. method according to claim 4, it is characterised in that decompression and treating after unserializing will be carried out in internal memory and is located
After reason data to FPGA Heterogeneous Computing devices, in addition to:
Obtain the result of calculation of the FPGA Heterogeneous Computings device return and the result of calculation is returned into the Spark, it is described
Result of calculation is that the FPGA Heterogeneous Computings device calculates what is obtained using the pending data progress correspondence received.
6. a kind of data processing equipment, it is characterised in that including:
Modular converter, is used for:The pending data stored in Distributed Computing Platform is converted into the Parquet lattice by row storage
Formula;
Processing module, is used for:The pending data of the Parquet forms is loaded into internal memory by row group using Spark, and
The pending data is decompressed in internal memory and unserializing after send to FPGA Heterogeneous Computings device carry out to accrued
Calculate.
7. device according to claim 6, it is characterised in that the modular converter includes:
Converting unit, is used for:Generation includes the Parquet files of whole pending datas, and will all treat in this document
Processing data cutting is embarked on journey group, wherein in each row group the row comprising all pending datas row block, and either rank block
Be stored with the pending data of corresponding part in respective column.
8. device according to claim 7, it is characterised in that the converting unit includes:
Conversion subunit, is used for:Whole pending data cuttings are embarked on journey group in the Parquet files, wherein each institute
The size for stating row group is identical with the size of data block in the Distributed Computing Platform.
9. device according to claim 8, it is characterised in that the processing module includes:
Loading unit, is used for:Using advancing with the function of c language setting in Spark by the pending of the Parquet forms
Data are loaded into internal memory by row group.
10. device according to claim 9, it is characterised in that also include:
Module is returned, is used for:The pending data after decompression and unserializing to FPGA Heterogeneous Computings will be carried out in internal memory to fill
After putting, obtain the result of calculation of the FPGA Heterogeneous Computings device return and the result of calculation is returned into the Spark, institute
State result of calculation and carry out what correspondence calculating was obtained using the pending data received for the FPGA Heterogeneous Computings device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710253998.XA CN107092676A (en) | 2017-04-18 | 2017-04-18 | A kind of data processing method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710253998.XA CN107092676A (en) | 2017-04-18 | 2017-04-18 | A kind of data processing method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107092676A true CN107092676A (en) | 2017-08-25 |
Family
ID=59638522
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710253998.XA Pending CN107092676A (en) | 2017-04-18 | 2017-04-18 | A kind of data processing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107092676A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107680144A (en) * | 2017-10-16 | 2018-02-09 | 郑州云海信息技术有限公司 | A kind of method and device of WebP files conversion |
CN109542889A (en) * | 2018-10-11 | 2019-03-29 | 平安科技(深圳)有限公司 | Stream data column storage method, device, equipment and storage medium |
CN109918347A (en) * | 2019-03-07 | 2019-06-21 | 网宿科技股份有限公司 | A kind of method and apparatus generating parquet file |
CN110187829A (en) * | 2019-04-22 | 2019-08-30 | 上海蔚来汽车有限公司 | A kind of data processing method, device, system and electronic equipment |
CN110689475A (en) * | 2019-09-10 | 2020-01-14 | 浪潮电子信息产业股份有限公司 | Image data processing method, system, electronic equipment and storage medium |
CN111092947A (en) * | 2019-12-19 | 2020-05-01 | 北京金山云网络技术有限公司 | Data transmission method, client, intermediate server, node and electronic equipment |
CN112269765A (en) * | 2020-11-13 | 2021-01-26 | 中盈优创资讯科技有限公司 | Method and device for improving data source reading performance of Spark structured stream file |
CN112835867A (en) * | 2021-01-11 | 2021-05-25 | 中国农业银行股份有限公司 | Data preprocessing method and device |
CN117591035A (en) * | 2024-01-18 | 2024-02-23 | 天津医康互联科技有限公司 | Data set processing method, device and computer readable storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810197A (en) * | 2012-11-09 | 2014-05-21 | 北京亿赞普网络技术有限公司 | Hadoop-based data processing method and system |
US20170090893A1 (en) * | 2015-09-29 | 2017-03-30 | Skytree, Inc. | Interoperability of Transforms Under a Unified Platform and Extensible Transformation Library of Those Interoperable Transforms |
-
2017
- 2017-04-18 CN CN201710253998.XA patent/CN107092676A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810197A (en) * | 2012-11-09 | 2014-05-21 | 北京亿赞普网络技术有限公司 | Hadoop-based data processing method and system |
US20170090893A1 (en) * | 2015-09-29 | 2017-03-30 | Skytree, Inc. | Interoperability of Transforms Under a Unified Platform and Extensible Transformation Library of Those Interoperable Transforms |
Non-Patent Citations (1)
Title |
---|
MUHUAN HUANG 等: "Programming and Runtime Support to Blaze FPGA Accelerator Deployment at Datacenter Scale", 《ACM》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107680144A (en) * | 2017-10-16 | 2018-02-09 | 郑州云海信息技术有限公司 | A kind of method and device of WebP files conversion |
CN109542889A (en) * | 2018-10-11 | 2019-03-29 | 平安科技(深圳)有限公司 | Stream data column storage method, device, equipment and storage medium |
CN109542889B (en) * | 2018-10-11 | 2023-07-21 | 平安科技(深圳)有限公司 | Stream data column storage method, device, equipment and storage medium |
CN109918347A (en) * | 2019-03-07 | 2019-06-21 | 网宿科技股份有限公司 | A kind of method and apparatus generating parquet file |
CN110187829A (en) * | 2019-04-22 | 2019-08-30 | 上海蔚来汽车有限公司 | A kind of data processing method, device, system and electronic equipment |
CN110689475A (en) * | 2019-09-10 | 2020-01-14 | 浪潮电子信息产业股份有限公司 | Image data processing method, system, electronic equipment and storage medium |
CN111092947A (en) * | 2019-12-19 | 2020-05-01 | 北京金山云网络技术有限公司 | Data transmission method, client, intermediate server, node and electronic equipment |
CN111092947B (en) * | 2019-12-19 | 2022-05-10 | 北京金山云网络技术有限公司 | Data transmission method, client, intermediate server, node and electronic equipment |
CN112269765A (en) * | 2020-11-13 | 2021-01-26 | 中盈优创资讯科技有限公司 | Method and device for improving data source reading performance of Spark structured stream file |
CN112835867A (en) * | 2021-01-11 | 2021-05-25 | 中国农业银行股份有限公司 | Data preprocessing method and device |
CN117591035A (en) * | 2024-01-18 | 2024-02-23 | 天津医康互联科技有限公司 | Data set processing method, device and computer readable storage medium |
CN117591035B (en) * | 2024-01-18 | 2024-05-10 | 天津医康互联科技有限公司 | Data set processing method, device and computer readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107092676A (en) | A kind of data processing method and device | |
US9830303B1 (en) | Optimized matrix multiplication using vector multiplication of interleaved matrix values | |
CN109997132A (en) | Low time delay matrix multiplication component | |
CN103927193B (en) | Loading method, service end virtual machine during the migration operation of java application function | |
Zhong et al. | Medusa: A parallel graph processing system on graphics processors | |
WO2022001014A1 (en) | Neural network model compilation method and apparatus, storage medium, and electronic device | |
CN111786688B (en) | Broadband parallel channelization receiving method based on embedded GPU | |
EP3608913A1 (en) | Bioinformatics systems, apparatuses, and methods executed on an integrated circuit processing platform | |
CN114217886A (en) | Function calling method, computing device and storage medium | |
CN108197075B (en) | Multi-core implementation method of Inceptation structure | |
MacWilliam et al. | CrowdCL: Web-based volunteer computing with WebCL | |
CN103309893A (en) | Character string comparing method and device | |
CN104281636B (en) | The concurrent distributed approach of magnanimity report data | |
Sun et al. | Enabling embedded inference engine with arm compute library: A case study | |
CN102340296A (en) | GPU (Graphics Processing Unit)-based method for realizing frequency-domain parallel processing for high-end digital FIR (Finite Impulse Response) filter | |
Carneiro et al. | Towards chapel-based exascale tree search algorithms: dealing with multiple GPU accelerators | |
CN114117992A (en) | Serialization and deserialization method and device and electronic equipment | |
CN112559197B (en) | Convolution calculation data reuse method based on heterogeneous many-core processor | |
Funasaka et al. | Fully parallelized LZW decompression for CUDA-enabled GPUs | |
CN105531602A (en) | System and method of implementing finite difference time domain models with multiple accelerated processing components (APCs) | |
CN107643892B (en) | Interface processing method, device, storage medium and processor | |
US20210334264A1 (en) | System, method, and program for increasing efficiency of database queries | |
Hui et al. | APL since 1978 | |
Pietroń et al. | Accelerating select where and select join queries on a GPU | |
CN103746771B (en) | Data format conversion method of channel coding and decoding based on GPP and SIMD technologies |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170825 |