CN106227506A - A kind of multi-channel parallel Compress softwares system and method in memory compression system - Google Patents
A kind of multi-channel parallel Compress softwares system and method in memory compression system Download PDFInfo
- Publication number
- CN106227506A CN106227506A CN201510616502.1A CN201510616502A CN106227506A CN 106227506 A CN106227506 A CN 106227506A CN 201510616502 A CN201510616502 A CN 201510616502A CN 106227506 A CN106227506 A CN 106227506A
- Authority
- CN
- China
- Prior art keywords
- compression
- data
- internal memory
- memory
- parallel
- 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
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The present invention provides the multi-channel parallel Compress softwares system and method in a kind of memory compression system, including multiple parallel compression/decompression devices, for being compressed the data in internal memory/decompress concurrently.Multi-channel parallel Compress softwares system and method in the memory compression system of the present invention is effectively increased the data throughput of memory compression system;And accelerate internal memory response efficiency.
Description
Technical field
The present invention relates to the technical field of memory compression decompression, particularly relate to the multi-channel parallel pressure in a kind of memory compression system
Contracting decompression system and method.
Background technology
Along with the software system of the mobile device such as smart mobile phone, panel computer operation is constantly to intelligent development, the application of mobile terminal
Program is increasingly sophisticated, thus also grows with each passing day the demand of memory size and speed.Although Double Data Rate SDRAM (Double
Date Rate SDRAM, DDR SDRAM) technology also slowly improving, but its development progress is well below in moving chip
The development progress of other modules.Especially, mobile terminal exacerbates internal memory significantly to the demand of display, figure, image, video etc.
And the disequilibrium between processor.It addition, the price range of decrease of main flow DDR is also far below the fall of processor, cause
The price proportion that DDR internal memory accounts in a mobile device is more and more higher.
Therefore, the hot subject that memory compression system processing speed becomes the most urgently to be resolved hurrily how is improved.
Summary of the invention
The shortcoming of prior art in view of the above, it is an object of the invention to provide the multichannel in a kind of memory compression system also
Row Compress softwares system and method, replacement based on cache line, use multi-channel structure make multiple cache line time
Territory realizes parallel processing, thus the problem being greatly reduced internal storage access entirety time delay.
For achieving the above object and other relevant purposes, the present invention provides the multi-channel parallel compression system in a kind of memory compression system
System, including multiple parallel compressoies, for being compressed the data that will write internal memory concurrently.
According to the multi-channel parallel compressibility in above-mentioned memory compression system, wherein: be initially completed the compressor pressure of compression
After having contracted, the priority completed according to data compression, successively by the data write internal memory after compression in multiple parallel compressoies.
According to the multi-channel parallel compressibility in above-mentioned memory compression system, wherein: described compressor is in running order
And after data to be compressed have read compressor, it is compressed operation.
Meanwhile, the present invention also provides for the multi-channel parallel compression method in a kind of memory compression system, by multiple parallel compressions
The data that will write internal memory are compressed by device concurrently.
According to the multi-channel parallel compressibility in above-mentioned memory compression system, wherein: be initially completed the compressor pressure of compression
After having contracted, according to the priority compressed, successively by the data write internal memory after compression in multiple parallel compressoies.
It addition, the present invention also provides for the multi-channel parallel decompression system in a kind of memory compression system, including multiple parallel decompressions
The data of reading from internal memory, for often organizing after data read from internal memory, are decompressed by device concurrently.
According to the multi-channel parallel decompressing method in above-mentioned memory compression system, wherein: from internal memory read data time, according to
The bus priority of the equipment that cache lines is corresponding is read out successively.
According to the multi-channel parallel decompressing method in above-mentioned memory compression system, wherein: described decompression machine is in running order
And data to be decompressed are after DDR reads decompression machine, carry out decompression operations.
Meanwhile, the present invention also provides for the multi-channel parallel decompressing method in a kind of memory compression system, by multiple parallel decompressions
Device, is often organizing after data read from internal memory, is decompressing the data read from internal memory concurrently.
According to the multi-channel parallel decompressing method in above-mentioned memory compression system, wherein: from internal memory read data time, according to
The bus priority of the equipment that cache lines is corresponding is read out successively.
It addition, the present invention also provides for the multi-channel parallel compressibility in a kind of memory compression system, including
One caching, is used for storing data;
One internal memory, for storing the data of compression;With
Multiple parallel compression/decompression devices, after the data compression in will be stored in caching concurrently, are stored in internal memory, or use
After the compression data decompression that will read from internal memory concurrently, it is stored in caching.
Meanwhile, the present invention also provides for the multi-channel parallel compressibility in a kind of memory compression system, including
One caching, is used for storing data;
One internal memory, for storing the data of compression;With
Multiple parallel compression/decompression devices are connected, by a single channel bus and internal memory chain by a multichannel bus and caching
Connecing, the plurality of parallel compression/decompression device is for reading, by multichannel bus parallel, the data being stored in caching, and will read
It is stored in internal memory by single channel bus after the data compression taken, or reads compression data by single channel bus from internal memory, and will
After the compression data decompression that internal memory reads, it is stored in caching by multichannel bus parallel.
As it has been described above, the multi-channel parallel Compress softwares system and method in the memory compression system of the present invention, have following useful
Effect:
(1) data throughput of memory compression system it is effectively increased;
(2) internal memory response efficiency is accelerated.
Accompanying drawing explanation
The structural representation of the multi-channel parallel Compress softwares system that Fig. 1 is shown as in the memory compression system of the present invention;
The schematic diagram of one embodiment of data write when Fig. 2 is shown as in internal memory not having compressor;
The schematic diagram of the embodiment that data when Fig. 3 is shown as in internal memory employing single-channel compression device write;
The signal of the preferred embodiment that data when Fig. 4 is shown as in the internal memory of the present invention employing multichannel compressor write
Figure;
The schematic diagram of one embodiment of digital independent when Fig. 5 is shown as in internal memory not having decompression machine;
The schematic diagram of one embodiment of digital independent when Fig. 6 is shown as in internal memory using single channel decompression machine;
The signal of one preferred embodiment of digital independent when Fig. 7 is shown as in the internal memory of the present invention using multichannel decompression machine
Figure.
Detailed description of the invention
Below by way of specific instantiation, embodiments of the present invention being described, those skilled in the art can be by disclosed by this specification
Content understand other advantages and effect of the present invention easily.The present invention can also be added by the most different detailed description of the invention
To implement or application, the every details in this specification can also be based on different viewpoints and application, in the essence without departing from the present invention
Various modification or change is carried out under god.
It should be noted that the diagram provided in the present embodiment illustrates the basic conception of the present invention the most in a schematic way, the most graphic
In component count, shape and size time only display with relevant assembly in the present invention rather than is implemented according to reality draw, its reality
During enforcement, the kenel of each assembly, quantity and ratio can be a kind of random change, and its assembly layout kenel is likely to increasingly complex.
Fig. 1 shows the structural representation of the multi-channel parallel Compress softwares system in a memory compression system.This memory compression system
At least include one-level or multistage caching (Cache) 120, one or more compression/decompression module (such as module 131,
132,133 etc.), and one internal memory 140.This memory compression system can be a computer system (such as notebook electricity
Brain, panel computer, or smart mobile phone etc.) the computing of CPU 110 its required data are provided.One-level, two grades, or
Three grades of cachings (being referred to as Cache 120) can be directly connected to CPU 110, for reading or the storage data of CPU 110.
And the data in Cache 120 can pass through one or more compression/decompression module 131,132, or 133 be converted to pressure
After contracting data, it is stored in internal memory 140.Compression data in internal memory 140 can also pass through compression/decompression module 131,132,
Or after 133 decompressions, become uncompressed data and be stored in Cache 120.In other words, storage in Cache 120
It is unpressed data.And store is the data of compression in internal memory 140.In certain embodiments, also have other compression/decompression
Module (embodying the most in FIG) will store internal memory 140 after the data compression that other channels (such as hard disk, network) obtains
In, or send out after the compression data decompression in internal memory 140 from these other channels.
In certain embodiments, Cache 120 is three grades of cachings (L3 Cache), is a kind of for miss after reading L2 cache
Data and the caching that designs.The operation principles of L3 is to use storage device faster to retain a from storage device at a slow speed
Data streams read and copying in (such as internal memory 140), when there is a need to read and write data from slower storage body again, energy
The action enough making read-write first completes on quick device, thus improves the response speed of system.Due to the introducing of L3, this reality
The memory compression technology in example of executing can make the demand data piezometric in internal memory 140 be reduced to original about 1/2, it is also possible to many
Under number scene, allow computer system keep performance constant, even under a few cases, cause the lifting of performance.
In certain embodiments, due to need to access internal memory 140 CPU 110 (can be described as in a mobile device Host or
Core) having multiple, these CPU 110 can dish out the Address requests of multiple access internal memory 140 simultaneously.If an internal memory pressure
In compression system, all of CPU 110 allows to access internal memory 140 by L3 Cache 120, and at Cache 120 and interior
Depositing only one group of compression/decompression module 131 between 140, those all operations relevant to Address requests are required for one in order
Individual it be compressed by compression/decompression module 131/decompress.Owing to compression/decompression module 131 becomes bottleneck, its time delay
(latency) being superimposed with the consumption accessing internal memory 140, final result is to considerably increase the visit of internal memory 140
Ask time delay.
In one embodiment, memory compression system arranges multiple compression/decompression module 131 between Cache 120 and internal memory 140,
132, and 133.Wherein, Cache 120 and compression/decompression module 131, between 132, and 133, " multichannel bus " is set,
Allow each compression/decompression module 131,132, and 133 can receive data simultaneously by this multichannel bus from Cache 120,
Or transfer data to Cache120.Meanwhile, in compression/decompression module 131,132, and arrange between 133 and internal memory 140
" single channel bus ".So, a moment, only one of which compression/decompression module 131,132, or 133 can by should
Data are stored in internal memory 140 by single channel bus, or read data from internal memory 140.In certain embodiments, compression/decompression mould
Block 131,132, or 133 can each implement compression or decompression operation.
The memory compression system of the present invention, in Cache 120 and internal memory 140 use multichannel, and include multiple can be parallel
The compression/decompression module of running, for concurrently the data that will write internal memory 140 being compressed, or will be from internal memory 140
The compression data of middle reading decompress.When being simultaneously received multiple request that data are stored in internal memory 140 from Cache 120
Time, the memory compression system of the present invention can allow multiple compression/decompression module 131,132,133 compress simultaneously and ask phase with these
Corresponding data, and the priority completed according to data compression, the data write internal memory 140 that will first compress successively, without
Will be according to the write internal memory 140 the most later of the request receiving storage data.And, compression/decompression module is only being in work
When state and data to be processed are the most ready, just it is compressed operation.
The difference of the present invention and prior art is illustrated below by specific embodiment.
As in figure 2 it is shown, have in Cache 120 three groups of unpressed data (write_1, write_2 and write_3) according to
Priority orders needs to write in internal memory 140.In other words, memory compression system successively obtains one or more CPU 110
The write request gone out, it is desirable to data write_1 in Cache 120, write_2 and write_3 are write internal memory 140.?
In Fig. 2, by data write_1, write_2 and write_3 write internal memory 140 needed for time, with contain write_1,
Three rectangular length of write_2 and write_3 are expressed.
If some moving chip framework does not has L3 caching and compression/decompression module, asking of concurrent write internal memory 140
Ask and can access internal memory 140 by a bus arbitration according to the priority orders of each request initiating terminal through taking.And move at some
After adding single-channel compression/decompression module in chip architecture, although because data are compressed, the time of write internal memory 140 can be significantly
Reduce, but can be because introducing compression/decompression module and time delay in whole reading path.
As shown in Figure 3, it is assumed that have three groups of unpressed data write_1, write_2 and write_3 to need according to priority orders
Internal memory 140 is write after being compressed successively.The first three rows of Fig. 3 represents that memory compression system is respectively by these three groups of data pressures respectively
After contracting/decompression module first carries out a squeeze operation, then write the time required for internal memory 140.If using single-channel compression solution
Die block, a compression/decompression module to compress data one by one according to precedence.Fourth line represents that three groups of arrays perform successively
Write internal memory 140 after compression to operate: after i.e. memory compression system first compresses the data of write_1 by its compression/decompression module, will
Write_1 data write internal memory 140 after compression;The most again after the data with compression/decompression module compression write_2, will compression
After write_2 data write internal memory 140;Finally after the data with compression/decompression module compression write_3, after compression
Write_3 data write internal memory 140.The above all operations desired time is T1.From the figure 3, it may be seen that T1 is three groups of data
It is compressed the summation of write operation wanted time respectively.Owing to using single-channel compression/decompression module, the compression behaviour of each group data
Work cannot synchronize to perform, and causes writing write_1, write_2, and write_3 and creates long time delay.
In one embodiment, some moving chip framework has the internal memory pressure of multichannel compression/decompression module as described in Figure 1
Compression system.When multichannel compression/decompression module in this memory compression system carries out parallelly compressed, CPU 110 is first by be compressed
Data are stored in Cache 120, and Cache 120 is passed through " multichannel bus " by data to be compressed by this memory compression system again
It is sent to 3 parallel compression modules (131,132 and 133) concurrently.3 parallel compression modules 131,132, and
133 are compressed operation simultaneously to data.After the compression module being initially completed compression completes compression, complete according to data compression
Priority, by " single channel bus " successively will in 3 parallel compressoies compression after data write internal memory 140 in.
As shown in Figure 4, in a preferred embodiment in the memory compression system of the present invention, including three compression modules
Compressor1, Compressor2 and Compressor3, it is possible to concurrently the data that will write internal memory 140 are compressed.
Assume have three groups of unpressed data write_1, write_2 and write_3 to write after needing to be compressed successively according to priority orders
Enter internal memory 140.Three row of Fig. 4 represent respectively memory compression system by these three groups of data compression module Compressor1,
After Compressor2 and Compressor3 first carries out squeeze operation, then write the time required for internal memory 140.Specifically, three
Data can be compressed by individual compression module Compressor1, Compressor2 and Compressor3 the most respectively.
Wherein, first Compressor2 completes compression, is followed by Compressor3 and Compressor1.Therefore, Compressor2 is worked as
When first completing compression, the priority completed according to data compression, memory compression system is first by the data write of the write_2 of compression
Deposit 140.Then, Compressor3 completes compression, and memory compression system is by the data write internal memory 140 of the write_3 of compression.
Finally, after the compression of Compressor1 completes, then by the data write internal memory 140 of the write_1 of compression.As seen from the figure,
After using the multi-channel parallel in the memory compression system of the present invention to compress and store three groups of data, required time T2 is much smaller than
T1, thus it is greatly reduced the overall time delay of memory storage.
It is to say, when memory compression system performs the operation writing data into internal memory 140, the order of data write depends on
The compression of each compression module completes order.After each compression module completes compression, memory compression system can be according to completing compression
Order, successively by each compression module compress after data write internal memory 140.It should be noted that in first pressure of write
During data after the compression of contracting module, if there being second compression module to complete compression, because compression module and internal memory 140
Between be single pass reason, after the data after the compression of first compression module will be waited to have write, compression memory system starts again
Implement the write operation of the data after second compression module compression.
Correspondingly, the multi-channel parallel compression method in the memory compression system of the present invention includes herein below:
By multiple parallel compressoies, concurrently the data that will write internal memory are compressed.Wherein, pressure is completed first
After the compressor compresses of contracting completes, according to the priority compressed, successively the data after compression in multiple parallel compressoies are write
Enter.
Multi-channel parallel decompression system in the memory compression system of the present invention includes multiple parallel decompression module, for from interior
Deposit after reading compression data in 140, concurrently the compression data read from internal memory 140 are decompressed.
Wherein, when memory compression system reads the data in internal memory 140, can be according to the CPU's 110 of Cache 120 correspondence
Bus priority is read out successively.Specifically, the total capacity of L3 Cache 120 can be configured to 2-32Mbytes, and is divided
Become the cache line of multiple a length of 1KBytes.When multiple cache line need to be replaced, memory compression system is according to each
The bus priority of CPU 110 corresponding for cache line accesses internal memory 140 successively.And, decompression module is only being in work
State and data to be decompressed, when internal memory 140 reads decompression module, just carry out decompression operations.Multiple parallel decompressions
Module is successively read compression data according to certain priority orders, and it is complete that the decompression module of each duty reads compression data at it
Decompression operations can be carried out concurrently after Biing.
As it is shown in figure 1, memory compression system is when carrying out multi-channel parallel decompression, 3 parallel decompression module (131,132
With 133) from internal memory 140, read compression data by single channel bus according to certain priority orders;3 parallel decompressions
Module 131,132, and 133 after in compression, digital independent completes, and i.e. carries out respective data expansion step;Each decompression module
131,132, and 133 after completing the data decompression of their own, will decompression data write Cache 120 by multichannel bus
In, read for CPU 110.
The difference of the present invention and prior art is illustrated below by specific embodiment.
As it is shown in figure 5, have three groups of data compressed (read_1, read_2 and read_3) in internal memory 140 according to priority
Order needs decompression and is loaded in Cache 120.In other words, memory compression system successively obtains one or more CPU 110
After the read requests sent, find that the data of CPU110 needs are not in Cache 120.Therefore, memory compression system will be by
Data read_1, read_2 and read_3 in internal memory 140 read in Cache 120.In Figure 5, read from internal memory 140
Time needed for data read_1, read_2 and read_3, by three rectangles containing read_1, read_2 and read_3
Length express.
If do not have L3 caching and compression/decompression module in some moving chip framework, concurrent reads asking of internal memory 140
Ask and can access internal memory 140 by a bus arbitration according to the priority orders of each request initiating terminal through taking.And move at some
After chip architecture adds single-channel compression/decompression module, although the data in internal memory 140 are compressed, and the reading time can subtract significantly
Few, but can be because introducing compression/decompression module and time delay in whole reading path.
As shown in Figure 6, it is assumed that have three groups of data read_1 compressed, read_2 and read_3 to need according to priority orders successively
Carry out decompression write Cache 120.The first three rows of Fig. 6 represent respectively memory compression system respectively by these three groups of data from internal memory 140
After reading out, carry out a decompression operations by a compression/decompression module, then write the time required for Cache 120.If
Using single-channel compression decompression module, the fourth line of Fig. 6 represents that three groups of arrays write Cache 120 after performing reading-decompression successively
Operation: i.e. memory compression system is first after internal memory 140 reads read_1, by the data of compression/decompression module decompression read_1;
Then after internal memory 140 reads read_2, by the data of compression/decompression module compression read_2;Finally read from internal memory 140
After read_3, by the data of compression/decompression module compression read_3.The above all operations desired time is T3.By Fig. 6
Understanding, T3 is read respectively by three groups of data and carries out decompressing the summation of wanted time.Owing to using single-channel compression/decompression module,
The squeeze operation of each group data cannot synchronize to perform, and causes reading read_1, read_2, and read_3 to create and prolongs for a long time
Time.
As it is shown in fig. 7, in the memory compression system of the present invention in a preferred embodiment, including three decompression module
Decompressor1, Decompressor2 and Decompressor3, it is possible to concurrently the data read from internal memory 140 are carried out
Decompression.That is, by single channel bus according to certain priority orders read three sections compression data read_1, read_2, and
Read_3, and after every segment data reads, use corresponding decompression module to carry out decompression operations immediately.Three row of Fig. 7 divide
Not Biao Shi memory compression system by these three groups of data after internal memory 140 reads out, with decompression module Decompressor1,
Decompressor2 and Decompressor3 carries out decompression operations.Wherein, memory compression system, after having read read_1, is adopted
Decompress with Decompressor1, and the data decompressed are stored to Cache 120;After having read read_2, use
Decompressor 2 decompresses;After having read read_3, Decompressor3 is used to decompress.Wherein, three
Decompression machine Decompressor1, Decompressor 2 and Decompressor 3 is concurrent working, and total by multichannel
Line is connected to Cache 120, the most interrelated.As shown in Figure 7, the multichannel in the memory compression system of the present invention is used
Three groups of data of parallel decompression, required time T4 is much smaller than T3, thus is greatly reduced the overall time delay of internal storage access.
Correspondingly, the multi-channel parallel decompressing method in the memory compression system of the present invention includes herein below:
By multiple parallel decompression machines, after often organizing digital independent, concurrently reading data are decompressed.
Wherein, when reading the data in internal memory, it is read out successively according to the bus priority of equipment corresponding to cache lines.Specifically
Ground, the total capacity of L3 can be configured to 2-32Mbytes, it is possible to is divided into the cache line of multiple a length of 1KBytes.When
When multiple cache line need to be replaced, system accesses DDR successively according to the bus priority of each equipment corresponding for cache line.
In the present invention, compression/de-compression is consistent with the change direction of the frequency of internal memory 140.The generally compression/de-compression time
The time more than read/write memory 140.Therefore, solved by the multi-channel parallel compression in the memory compression system of the employing present invention
Pressure system and method, is effectively increased the data throughput of memory compression system;And accelerate the response efficiency of DDR internal memory.
Wherein, in the present invention, involved compressor and decompression machine are used compression and decompression algorithm are known algorithm.Such as,
The compression algorithm used is in " Parallel Compression with Cooperative Dictionary Construction "
The algorithm illustrated.
To those skilled in the art, certain existing compression/decompression algorithm is used to realize the Compress softwares of data to be
Ripe technology, therefore do not repeat them here.
It should be noted that all compressoies involved in the present invention and decompression machine all can concurrent workings.Therefore in the visit for internal memory
In asking, the compression/decompression of data can be performed simultaneously.
In sum, the multi-channel parallel Compress softwares system and method in the memory compression system of the present invention is effectively increased internal memory
The data throughput of compressibility;And accelerate internal memory response efficiency.So, the present invention effectively overcomes kind of the prior art
Plant shortcoming and have high industrial utilization.
The principle of above-described embodiment only illustrative present invention and effect thereof, not for limiting the present invention.Any it is familiar with this skill
Above-described embodiment all can be modified under the spirit and the scope of the present invention or change by the personage of art.Therefore, such as
All that in art, tool usually intellectual is completed under without departing from disclosed spirit and technological thought etc.
Effect is modified or changes, and must be contained by the claim of the present invention.
Claims (12)
1. the multi-channel parallel compressibility in a memory compression system, it is characterised in that: include multiple parallel compressor, be used for
Concurrently the data that will write internal memory are compressed.
Multi-channel parallel compressibility in memory compression system the most according to claim 1, it is characterised in that: it is being initially completed
After the compressor compresses of compression completes, the priority completed according to data compression, successively by after compression in multiple parallel compressoies
Data write internal memory.
Multi-channel parallel compressibility in memory compression system the most according to claim 1, it is characterised in that: described compressor
After in running order and to be compressed data have read compressor, it is compressed operation.
4. the multi-channel parallel compression method in a memory compression system, it is characterised in that: by multiple parallel compressoies, parallel
The data that will write internal memory are compressed by ground.
Multi-channel parallel compressibility in memory compression system the most according to claim 4, it is characterised in that: it is being initially completed
After the compressor compresses of compression completes, according to the priority compressed, successively by the number after compression in multiple parallel compressoies
According to write internal memory.
6. the multi-channel parallel decompression system in a memory compression system, it is characterised in that: include multiple parallel decompression machine, be used for
Often organizing after data read from internal memory, concurrently the data read from internal memory are being decompressed.
Multi-channel parallel decompressing method in memory compression system the most according to claim 6, it is characterised in that: read from internal memory
When fetching data, it is read out successively according to the bus priority of equipment corresponding to cache lines.
Multi-channel parallel decompressing method in memory compression system the most according to claim 6, it is characterised in that: described decompression machine
In in running order and to be decompressed data after DDR reads decompression machine, carry out decompression operations.
9. the multi-channel parallel decompressing method in a memory compression system, it is characterised in that: by multiple parallel decompression machines, often
After group data read from internal memory, concurrently the data read from internal memory are decompressed.
Multi-channel parallel decompressing method in memory compression system the most according to claim 9, it is characterised in that: from internal memory
During middle reading data, it is read out successively according to the bus priority of equipment corresponding to cache lines.
Multi-channel parallel compressibility in 11. 1 kinds of memory compression systems, it is characterised in that: include
One caching, is used for storing data;
One internal memory, for storing the data of compression;With
Multiple parallel compression/decompression devices, after the data compression in will be stored in caching concurrently, are stored in internal memory, or
Person, after the compression data decompression that will read from internal memory concurrently, is stored in caching.
Multi-channel parallel compressibility in 12. 1 kinds of memory compression systems, it is characterised in that: include
One caching, is used for storing data;
One internal memory, for storing the data of compression;With
Multiple parallel compression/decompression devices are connected by multichannel bus and caching, by a single channel bus and interior
Depositing link, the plurality of parallel compression/decompression device is used for reading, by multichannel bus parallel, the data being stored in caching,
And internal memory will be stored in by single channel bus after the data compression of reading, or read compression by single channel bus from internal memory
After data, and the compression data decompression that will read from internal memory, it is stored in caching by multichannel bus parallel.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510296194 | 2015-06-02 | ||
CN2015102961949 | 2015-06-02 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106227506A true CN106227506A (en) | 2016-12-14 |
Family
ID=57528729
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510616502.1A Pending CN106227506A (en) | 2015-06-02 | 2015-09-24 | A kind of multi-channel parallel Compress softwares system and method in memory compression system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106227506A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107124615A (en) * | 2017-05-15 | 2017-09-01 | 郑州云海信息技术有限公司 | A kind of method and device of WebP lossy compression methods |
CN108415668A (en) * | 2018-02-06 | 2018-08-17 | 珠海市杰理科技股份有限公司 | Chip motivational techniques, device, system, computer equipment and storage medium |
CN109445719A (en) * | 2018-11-16 | 2019-03-08 | 郑州云海信息技术有限公司 | A kind of date storage method and device |
CN112860323A (en) * | 2019-11-27 | 2021-05-28 | 珠海格力电器股份有限公司 | Method and device for loading file into memory |
CN113885949A (en) * | 2021-10-22 | 2022-01-04 | 瑞芯微电子股份有限公司 | Quick startup method and system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6523102B1 (en) * | 2000-04-14 | 2003-02-18 | Interactive Silicon, Inc. | Parallel compression/decompression system and method for implementation of in-memory compressed cache improving storage density and access speed for industry standard memory subsystems and in-line memory modules |
US6879266B1 (en) * | 1997-08-08 | 2005-04-12 | Quickshift, Inc. | Memory module including scalable embedded parallel data compression and decompression engines |
CN102122959A (en) * | 2011-03-29 | 2011-07-13 | 西安交通大学 | Data compression device for improving main memory reliability of computer, and method thereof |
-
2015
- 2015-09-24 CN CN201510616502.1A patent/CN106227506A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6879266B1 (en) * | 1997-08-08 | 2005-04-12 | Quickshift, Inc. | Memory module including scalable embedded parallel data compression and decompression engines |
US6523102B1 (en) * | 2000-04-14 | 2003-02-18 | Interactive Silicon, Inc. | Parallel compression/decompression system and method for implementation of in-memory compressed cache improving storage density and access speed for industry standard memory subsystems and in-line memory modules |
CN102122959A (en) * | 2011-03-29 | 2011-07-13 | 西安交通大学 | Data compression device for improving main memory reliability of computer, and method thereof |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107124615A (en) * | 2017-05-15 | 2017-09-01 | 郑州云海信息技术有限公司 | A kind of method and device of WebP lossy compression methods |
CN108415668A (en) * | 2018-02-06 | 2018-08-17 | 珠海市杰理科技股份有限公司 | Chip motivational techniques, device, system, computer equipment and storage medium |
CN109445719A (en) * | 2018-11-16 | 2019-03-08 | 郑州云海信息技术有限公司 | A kind of date storage method and device |
CN109445719B (en) * | 2018-11-16 | 2022-04-22 | 郑州云海信息技术有限公司 | Data storage method and device |
CN112860323A (en) * | 2019-11-27 | 2021-05-28 | 珠海格力电器股份有限公司 | Method and device for loading file into memory |
CN113885949A (en) * | 2021-10-22 | 2022-01-04 | 瑞芯微电子股份有限公司 | Quick startup method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10545860B2 (en) | Intelligent high bandwidth memory appliance | |
CN105843775B (en) | On piece data divide reading/writing method, system and its apparatus | |
US10140123B2 (en) | SIMD processing lanes storing input pixel operand data in local register file for thread execution of image processing operations | |
CN106227506A (en) | A kind of multi-channel parallel Compress softwares system and method in memory compression system | |
CN111199273B (en) | Convolution calculation method, device, equipment and storage medium | |
JP5121291B2 (en) | Data transfer system | |
CN111079917B (en) | Tensor data block access method and device | |
CN111031011B (en) | Interaction method and device of TCP/IP accelerator | |
US20220179823A1 (en) | Reconfigurable reduced instruction set computer processor architecture with fractured cores | |
US11487342B2 (en) | Reducing power consumption in a neural network environment using data management | |
US9692813B2 (en) | Dynamic assignment of transfers of blocks of data | |
CN115880132A (en) | Graphics processor, matrix multiplication task processing method, device and storage medium | |
CN111324294A (en) | Method and apparatus for accessing tensor data | |
CN112235579A (en) | Video processing method, computer-readable storage medium and electronic device | |
KR101788245B1 (en) | Multi-port cache memory apparatus and method for operating multi-port cache memory apparatus | |
CN114442908B (en) | Hardware acceleration system and chip for data processing | |
CN104598409A (en) | Method and device for processing input and output requests | |
CN108701102A (en) | Direct memory access controller, method for reading data and method for writing data | |
KR100895298B1 (en) | Apparatus, Method, Data Processing Element for efficient parallel processing of multimedia data | |
US11467973B1 (en) | Fine-grained access memory controller | |
US20230161835A1 (en) | Matrix operation method and accelerator | |
CN115344393A (en) | Service processing method and related equipment | |
CN111666253B (en) | Delivering programmable data to a system having shared processing elements sharing memory | |
US11210105B1 (en) | Data transmission between memory and on chip memory of inference engine for machine learning via a single data gathering instruction | |
US10412400B1 (en) | Memory access ordering for a transformation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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: 20161214 |