CN1204090A - Half-line communication processor supporting parallel computation of computer group - Google Patents
Half-line communication processor supporting parallel computation of computer group Download PDFInfo
- Publication number
- CN1204090A CN1204090A CN 98114065 CN98114065A CN1204090A CN 1204090 A CN1204090 A CN 1204090A CN 98114065 CN98114065 CN 98114065 CN 98114065 A CN98114065 A CN 98114065A CN 1204090 A CN1204090 A CN 1204090A
- Authority
- CN
- China
- Prior art keywords
- communication
- module
- buffer
- data
- bus
- 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
- 230000006854 communication Effects 0.000 title claims abstract description 159
- 238000004891 communication Methods 0.000 title claims abstract description 152
- 239000000872 buffer Substances 0.000 claims abstract description 170
- 238000012545 processing Methods 0.000 claims abstract description 15
- 230000005540 biological transmission Effects 0.000 claims description 39
- 230000003139 buffering effect Effects 0.000 claims description 9
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- 238000000547 structure data Methods 0.000 claims description 3
- 230000007257 malfunction Effects 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 7
- 239000013256 coordination polymer Substances 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000000034 method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 2
- -1 address counting JS Substances 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 238000005562 fading Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012856 packing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Landscapes
- Computer And Data Communications (AREA)
- Multi Processors (AREA)
Abstract
The present invention relates to an open-ended computer group interconnected network unit, in particular it is a support computer group parallel communication processor, consisting of communication controller, register, communication status register, communication protocol register, hardware code error checking device, hardware malfunction register and control circuit. It possesses several processing node interfaces, can simultaneously form several communication channels, and each communication possesses its respective communication buffer zone, and every buffer zone includes independent receiving buffer zone and sending buffer zone, and all the buffer zones simultaneusly receive or send data so as to implement on-network multi-computer parallel communication.
Description
The present invention relates to a kind of scalable computer group of planes interconnection network device, particularly a kind of support computer group parallel computation parallel communications processor.
Along with the development of VLSI technology, last decade comes the development of concurrent computational system very rapid, and a lot of commercial massively parallel computer systems (hereinafter to be referred as KPP) put goods on the market.Be accompanied by the development of MPP recent years, a kind of new parallel and distributed computing technology, and computer group technology (computer cluster technology) has caused people's very big concern, has become very active research field.Group of planes technology is intended to a group computing machine (as workstation, micro computer, large scale computer etc.) with network to make full use of each computer resource with certain interconnected, and uniform dispatching, Coordination Treatment realize the high-level efficiency parallel computation.Group of planes parallel computation has following seven principal features:
1, the ratio of performance to price is higher than MPP;
2, customer investment risk is little, particularly to having the user of a lot of computer systems;
3, programming is convenient;
4, structure is flexible;
5, can make full use of the stand-alone computer resource;
6, loose couplings structure extensibility is good;
7, has height I/O concurrency.
These seven characteristics have represented the glamour of group of planes concurrent computational system.External a lot of scholar and manufacturer's prophesy, group of planes parallel computer will be run neck and neck with MPP, the development of leading parallel computing.
The performance of each stand-alone computer has not been a problem in the group of planes parallel computation.An efficiency that key issue is interconnection network of group of planes technology.When group of planes parallel computer carried out parallel computation, each computing machine need carry out a large amount of communication and synchronous operation by interconnection network.If the efficient of interconnection network is very low, communication will be blocked, prolong the parallel processing time, reduce speed-up ratio, influence the efficient of group of planes parallel computation.Therefore, network service is a bottleneck of group of planes parallel computation.
The interconnection network of group of planes parallel computer as ETHERNET, TOKEN-RING etc., also have some to use high speed networks such as FDDI, ATM mainly based on local area computer network at present.All these networks all have the characteristics of sequenced communications, can not solve the network bottleneck problem well.In order to solve the network bottleneck problem, people have carried out some research work, as the bus bridge interconnection network of BJ University of Aeronautics ﹠ Astronautics
[6]Patent Office of the People's Republic of China is in disclosed node interconnecting adapter patent of invention on Dec 11st, 1.This patent is to be installed on the distributed bus expansion slot of expanding each node in the parallel Network of Workstation, the Bus Interface Unit that Bus Interface Unit that is joined by the bus duct with this node computer and interconnection network join, the dispensing unit of bi-directional data register, command register, status register and XBH, ruling logical block and interrupt control logic unit are formed, be structure based on the BBP agreement, the degree of coupling is adjustable parallel computer Network of Workstation critical component.Yet, this machine use bi-directional data register respectively with data/address bus and this node computer internal bus and the address wire of BBP interface unit, this machine is only supported a kind of topological structure, and only is applicable to workstation, speed is slower when data transmission.
The object of the present invention is to provide a kind of parallel communications processor, implemented the parallel communications of many microcomputers, multicomputer can physically transmit the computing machine of information simultaneously on the net.
In order to achieve the above object, the present invention is that each HCP has a plurality of processing node interfaces, can form many communication ports simultaneously, each communication port of HCP all has communication buffer separately, each buffer zone all comprises independently send buffer and transmission buffer zone, and each buffer zone can receive or send data simultaneously.
HCP is made up of eight modules, these eight modules are respectively module 01, module 02, module 03, module 1, module 2, module 3, module 4 and module 5, module 01 comprises a communication controler, module 02 is a register module, it comprises the communications status register, the communication protocol register, hardware error checking device, hardware fault register and control circuit.Module 01, the 02 common main control circuit that constitutes the parallel communications processor, main control circuit is responsible for the transaction scheduling control of each module, hardware communication protocols, fault processing, and the communication data transmission adopts the hardware circuit that can break away from communication controler to realize, module 03 is the host computer interface, it provides interface and data communication protocol between parallel communications processor and the host computer, module 1, module 2, module 3, module 4 and module 5 are five identical in structure data transmission modules, and each data transmission module is all by a send buffer, one sends buffer zone, three multidigit bus switchs and a formation such as multi-bit parallel I/O communication port.
Send buffer zone by sending bus by data buffer KF1-1, transceiver bus, impact damper KP1-1, output bus links to each other with communication port Port, this is to send the path of buffering to communication port Port output data, sending buffer zone links to each other with communication controler by data buffer ZK1, link to each other with the multidigit bus by impact damper KF1-2, be called the reception bus with the direct-connected bus of send buffer, send buffer is by data buffer KS1-1, transceiver bus, impact damper KP1-2, output bus links to each other with communication port port, and this is to receive the path of buffering to communication port port output data; Send buffer links to each other with the multidigit bus by data buffer KS1-2, link to each other with communication controler by data buffer ZK2, the multidigit bus links to each other with output bus by data buffer KP1-2 and also links to each other with communication port Port, sends bus, reception bus, multidigit bus and output bus and is 16 bit parallel buses.
The control line of communication controler is by transmit control device and transmission buffer zone, address counting FS, finishing control FS and data buffer KF1-1, KF1-2, KP1-1, KP1-2 link to each other, address counting FS, and finishing control FS provides 16 bit address of the autoincrement mode that sends buffer zone jointly; The control line of communication controler is by receiving controller and send buffer, and JS is counted in the address, and finishing control JS and data buffer KS1-1, KS1-2, KP1-1, KP1-2 link to each other.
Host computer links to each other with the transmission buffer zone with the transmission bus by impact damper KSZ1, this is host computer and the data path that sends buffer zone, host computer links to each other with send buffer with the reception bus by impact damper KSZ2, and they constitute jointly, and host computer leads to and the data path of send buffer.
Characteristics of the present invention are:
(1) breaks away from the constraint of LAN, realized supporting the communication processor of computer group parallel computation;
(2) broken through the limitation of bus type communication network, realized online multicomputer parallel communications, promptly multicomputer can physically transmit information simultaneously on the net;
(3) realize the multi-bit parallel information transmission of network physical link, overcome LAN (Local Area Network) physical link succession problem;
(4) realize the overlapping of calculating and communication process, improved the concurrency of system;
(5) communication protocol and the message passing mechanism of simple and effective are provided;
(6) support various topological structures: many microcomputers interconnection network that can be used for realizing various topological structures such as Hypercube, annular, tree-like, a peacekeeping two-dimensional array.
Description of drawings:
Fig. 1 is the block scheme of parallel communications processor;
Fig. 2 is a parallel communications processor communication interface card block scheme;
Fig. 3 is four-dimensional super cube structure block scheme;
Fig. 4 is two adjacent processing nodes of super side's body and annexation block scheme thereof;
Fig. 5 is to use five data buffer storage circuits to carry out the data communication block scheme;
Fig. 6 has the independent communication processor block scheme that sends and receive buffer storage circuit.
Below in conjunction with accompanying drawing embodiments of the invention are described in further detail.
By Fig. 1-Fig. 6 as can be known, the present invention is that each HCP has a plurality of processing node interfaces, can form many communication ports simultaneously, each communication port of HCP all has communication buffer separately, each buffer zone all comprises independently send buffer and transmission buffer zone, and each buffer zone can receive or send data simultaneously.
HCP is made up of eight modules, these eight modules are respectively module 01, module 02, module 03, module 1, module 2, module 3, module 4 and module 5, module 01 comprises a communication controler, module 02 is a register module, it comprises the communications status register, the communication protocol register, hardware error checking device, hardware fault register and control circuit.Module 01, the 02 common main control circuit that constitutes the parallel communications processor, main control circuit is responsible for the transaction scheduling control of each module, hardware communication protocols, fault processing, and the communication data transmission adopts the hardware circuit that can break away from communication controler to realize, module 03 is the host computer interface, it provides interface and data communication protocol between parallel communications processor and the host computer, module 1, module 2, module 3, module 4 and module 5 are five identical in structure data transmission modules, and each data transmission module is all by a send buffer, one sends buffer zone, three multidigit bus switchs and a formation such as multi-bit parallel I/O communication port.
Send buffer zone by sending bus by data buffer KF1-1, transceiver bus, impact damper KP1-1, output bus links to each other with communication port Port, this is to send the path of buffering to communication port Port output data, sending buffer zone links to each other with communication controler by data buffer ZK1, link to each other with the multidigit bus by impact damper KF1-2, be called the reception bus with the direct-connected bus of send buffer, send buffer is by data buffer KS1-1, transceiver bus, impact damper KP1-2, output bus links to each other with communication port Port, and this is to receive the path of buffering to communication port Port output data; Send buffer links to each other with the multidigit bus by data buffer KS1-2, link to each other with communication controler by data buffer ZK2, the multidigit bus links to each other with output bus by data buffer KP1-2 and also links to each other with communication port Port, sends bus, reception bus, multidigit bus and output bus and is 16 bit parallel buses.
The control line of communication controler is by transmit control device and transmission buffer zone, address counting FS, finishing control FS and data buffer KF1-1, KF1-2, KP1-1, KP1-2 link to each other, address counting FS, finishing control FS provides 16 bit address of the autoincrement mode that sends buffer zone jointly: the control line of communication controler is by receiving controller and send buffer, address counting JS, finishing control JS and data buffer KS1-1, KS1-2, KP1-1, KP1-2 link to each other.
Host computer links to each other with the transmission buffer zone with the transmission bus by impact damper KSZ1, this is host computer and the data path that sends buffer zone, host computer links to each other with send buffer with the reception bus by impact damper KSZ2, and they constitute jointly, and host computer leads to and the data path of send buffer.
Described module 1-5 is five interface cards, and five interface cards are connected with module 01,02,03 by five interface sockets.
Described five interface cards can be connected with five host computers respectively, are used for realizing many microcomputers interconnection network of various topological structures such as Hypercube, annular, tree-like, a peacekeeping two-dimensional array.
Described multicomputer can be simultaneously on the net physically multi-bit parallel transmission information, realized the overlapping of calculating and communication process.
Principle of work of the present invention and function:
Communication controler
The microprocessor that adopts one 16 is as the communication controler among the HCP, and establishing CP is a HCP, and MP is the communication controler of CP, and the function of MP is as follows:
(1) state of each buffer zone of MP dynamic monitoring and each register.
(2) when the HCP of adjacency computing machine when the reception buffer storage circuit of CP is sent data, the host computer of MP notice CP receives this batch data or transmits this batch data to other HCP in abutting connection with computing machine.
(3) send data to the transmission buffer storage circuit of CP when the host computer of CP, and require when the neighbour presses the computer sends the data, MP is responsible for packing data and data is sent in the corresponding reception buffer storage circuit of corresponding HCP in abutting connection with computing machine.
(4) when communication hardware or data transmission error take place, MP is responsible for sending error message to the HCP of adjacency computing machine or host computer, and does necessary processing.
(5) MP is responsible for being provided with or removing the user mode sign of each reception and transmission buffer storage circuit.
Each receives buffer storage circuit all is the static random memory circuit of 16K byte.Receiving buffer storage circuit is the buffer zone that receives data.Among HCP each receives buffer storage circuit and all shares by the communication controler of the host computer of this HCP, this HCP with in abutting connection with the communication controler of the HCP of computing machine.Receive buffer storage circuit and only be responsible for receiving the data of sending here in abutting connection with the HCP of computing machine.Each receive buffer storage circuit corresponding one in abutting connection with computing machine.Receive buffer storage circuit and can receive data, also can transmit data.
In receiving data procedures, the communication controler of HCP is responsible for checking whether respectively receive buffer storage circuit has new data to arrive, if receive buffer storage circuit new data is arranged, it is the data that send to the host computer of local HCP that communication controler is further distinguished data, or interim data.If send to the data of the host computer of local HCP, microprocessor notice host computer receives this batch data; If interim data is transmitted this batch data by communication controler to corresponding HCP in abutting connection with computing machine.
When the request that the host computer of a HCP only sends at the communication controler that receives this HCP receives data interruption, just have the right from receive buffer storage circuit, to receive data.Host computer need be put and receive the buffer storage circuit busy flag when reading the data that receive buffer storage circuit.
Receive and send buffer storage circuit and all have the full and buffering memory circuit dummy status of buffer storage circuit.The state note that receives buffer storage circuit is made S, and the state note that sends buffer storage circuit is made F, respectively by the corresponding positions record of status register.The full state of buffer storage circuit is 1, and dummy status is 0.If S=1, then the communication controler of HCP reads to receive the address information in the buffer storage circuit.If S is 0, then show the buffer storage circuit sky, the communication controler of the HCP of the computing machine of adjacency can send data to this buffer storage circuit.After sending ED, S to be changed to 1 in abutting connection with the communication controler of the HCP of computing machine.
Send buffer storage circuit
Each sends buffer storage circuit is the SRAM of 16K byte.Sending buffer storage circuit is the buffer zone that host computer sends data.They only are responsible for receiving the data of being sent here by host computer, will send to the computing machine of adjacency.Sending buffer storage circuit is shared by the communication controler of host computer and its HCP.When a host computer will to it in abutting connection with the computer sends the data the time, host computer at first sends the data that tendency to develop is sent to sending buffer storage circuit; HCP is from sending the buffer storage circuit reading of data and sending data to the corresponding input buffering memory circuit of the HCP of purpose computing machine then; Last purpose computing machine reads the data of sending from the corresponding input buffering memory circuit of its HCP.
In sending data procedures, the communication controler of HCP is responsible for checking the state F that respectively sends buffer storage circuit.Inquiry sends in the buffer storage circuit has no datat to send to the adjacency computing machine.If a state F=0 who sends buffer storage circuit then represents this buffer storage circuit free time, the host computer of HCP can send buffer storage circuit to this and send data.When sending ED, the host computer of HCP will put 1 to F.Have only when sending the state F=0 of buffer storage circuit for one, the host computer of HCP just has the right to send data to this transmission buffer storage circuit.If a state F=1 who sends buffer storage circuit, then the communication controler of HCP can read the data in this buffer storage circuit, sends to the HCP of the purpose computing machine of adjacency.Have only when F=1, the microprocessor of HCP just has the right to use the transmission buffer storage circuit to send data.When sending ED, the state F that the communication controler of HCP is put corresponding transmission buffer storage circuit is 0.
The communications status register
The communications status register holds receives and sends the state of buffer storage circuit.The communications status register is one 10 bit register, and wherein 5 is to receive the buffer storage circuit Status Flag, and 5 is to send the buffer storage circuit Status Flag.The microprocessor of the HCP of each HCP and adjacency thereof can read or be provided with the value of communications status register.The host computer of a HCP also can read and be provided with the value of the status register of this HCP.
The host computer of a HCP obtains the state of buffer storage circuit by the value of read status register, to determine and could operate buffer storage circuit.
A HCP in abutting connection with the microprocessor of the HCP of computing machine value by read status register, obtain the state of buffer storage circuit, to determine and could carry out data transfer operation to buffer storage circuit.
When DTD, the microprocessor of HCP must respectively have off-position to carry out set or reset operation to status register.
The communication protocol register
The communication protocol register is used for supporting the communication protocol between the HCP, the design of simplified communication protocol software and realization.
Error checking
Reasons such as the coupling of the dry at random sound that communication line causes owing to electrical specification own, signal frequency and amplitude fading, phase distortion, adjacent lines is crosstalked, the interference of external environment medium-high magnetic field, electric field, power-supply fluctuation, can cause the variation of code element in the data communication process, make to produce error code in the communication process, cause garble.
In the garble that above-mentioned interference source causes, mainly be divided into randomness mistake and unexpected error.The variation of certain bit symbols that dry at random sound etc. causes is called the randomness mistake.The dry sound of power-supply fluctuation and impact waits the variation of the string code element that causes, is called unexpected error.
At unexpected error in the communication process and randomness mistake, communication processor adopts software approach to correct the mistake that produces in the communication process.The software error checking adopts the CRC check method.
The concurrency of data communication and the support of various topological structures
If only be provided with a communication buffer memory circuit among the HCP, then each passage, HCP and host computer thereof must use the communication buffer memory circuit in the mode of order queuing, in order to avoid the use conflict of communication buffer memory circuit.Like this, the concurrency of HCP data transmission can reduce greatly.In order to address this problem, HCP has adopted the multiple buffer technology.Each HCP has a plurality of communication port, can form many communication ports simultaneously.Each communication port all has communication buffer separately.Each buffer zone can receive or send data simultaneously.Because each channel all has communication buffer separately, they can receive or send data simultaneously, have avoided the use conflict of buffer storage circuit, have strengthened the concurrency of data transmission.Because each HCP has a plurality of communication port, the data that HCP has also realized hardware broadcasting and selected multicast function: HCP it any one can be sent in buffer storage circuit be done hardware broadcasting and choosing is broadcast to all or part of communication port simultaneously.
Fig. 5 has provided one and has used five examples that the data buffer storage circuit carries out data communication.BUFF1 among the figure, BUFF2, BUFF3, BUFF4, BUFF5 are five fully independently data communication buffer storage circuits, respectively with 5 processing node P
1 #-P
5 #Be connected.Data in buffer storage circuit are just by processing node P
1 #When outwards transmitting, P
2 #Can send data, P to BUFF2 simultaneously
2 #Passage also can send data to BUFF3 simultaneously, and P
4 #, P
5 #Number passage then also can outwards send the data among the BUFF4 simultaneously.Like this, five data communication buffers of five processing node channels memory circuit has been realized the parallel transmission of data.
In serial communication network, the general shared communication buffer memory circuit of Data Receiving and transmission, communication buffer is at a time only done and is received buffer storage circuit or only do the transmission buffer storage circuit, the reception of data and transmission timesharing ground use the communication buffer memory circuit, have influenced the concurrency of transceive data.HCP has solved this problem.We have adopted and have independently received buffer storage circuit and send buffer storage circuit, have realized the parallel receive data and have sent data.In a plurality of communication buffer memory circuits in HCP, each buffer storage circuit comprises that all one independently receives buffer storage circuit and one and independently sends buffer storage circuit, as shown in Figure 5.S1#, S2#, S3#, S4# and S5# are respectively No. 1, No. 2, No. 3, No. 4 and No. 5 and receive buffer storage circuit among the figure, and F1#, F2#, F3#, F4# and F5# are respectively No. 1, No. 2, No. 3, No. 4 and No. 5 and send buffer storage circuit.
Use after independent reception and the transmission buffer storage circuit, HCP can realize at an easy rate walking abreast and transmit and receive data.For example, when 1# processing node interface when S1# sends data, the host computer of parallel communications processor can send data to F1# simultaneously, make data between the computing machine send and receive and carry out simultaneously.
Because a plurality of independently communication port of parallel communications processor adopting can realize multiple communication network topology structure.For example, parallel communications processor with five separate ports can connect five channels simultaneously, can be used for realizing that tree network connects, the two-dimensional array network connects, ring network connects (only needing two processing node interfaces), five dimension Hypercube networks connect or complicated composite network connects.
In the computer group that uses HCP to constitute, every computing machine has a HCP.HCP is connected with affiliated computing machine by the host computer interface.If a HCP belongs to computing machine C, we claim that then computing machine C is the host computer of this HCP, and HCP is the subordinate communication processor of C.Every HCP can be connected by five computing machines of five communication interfaces and other.In abutting connection with computing machine, the transmission buffer storage circuit of a 16K byte and the reception buffer storage circuit of a 16K byte are arranged all among the HCP for each.Send and receive buffer storage circuit and be used for realizing host computer and communication buffer in abutting connection with computing machine.HCP has realized 16 bit parallel communications.
Claims (4)
1, a kind of parallel communications processor (HCP) of computer group parallel computation of supporting is installed on the bus expansion slot of each node computer that can expand the parallel computer Network of Workstation, it also comprises communication buffer, handles node interface, communication protocol register, controller, status register, it is characterized in that:
Each HCP has a plurality of processing node interfaces, can form many communication ports simultaneously, each communication port of HCP all has communication buffer separately, and each buffer zone all comprises independently send buffer and transmission buffer zone, and each buffer zone can receive or send data simultaneously
HCP is made up of eight modules, these eight modules are respectively module 01, module 02, module 03, module 1, module 2, module 3, module 4 and module 5, module 01 comprises a communication controler, module 02 is a register module, it comprises the communications status register, the communication protocol register, hardware error checking device, hardware fault register and control circuit.Module 01, the 02 common main control circuit that constitutes the parallel communications processor, main control circuit is responsible for the transaction scheduling control of each module, hardware communication protocols, fault processing, and the communication data transmission adopts the hardware circuit that can break away from communication controler to realize, module 03 is the host computer interface, it provides interface and data communication protocol between parallel communications processor and the host computer, module 1, module 2, module 3, module 4 and module 5 are five identical in structure data transmission modules, and each data transmission module is all by a send buffer, one sends buffer zone, three multidigit bus switchs and a formation such as multi-bit parallel I/O communication port.
Send buffer zone by sending bus by data buffer KF1-1, transceiver bus, impact damper KP1-1, output bus links to each other with communication port Port, this is to send the path of buffering to communication port Port output data, sending buffer zone links to each other with communication controler by data buffer ZK1, link to each other with the multidigit bus by impact damper KP1-2, be called the reception bus with the direct-connected bus of send buffer, send buffer is by data buffer KS1-1, transceiver bus, impact damper KP1-2, output bus links to each other with communication port Port, and this is to receive the path of buffering to communication port Port output data; Send buffer links to each other with the multidigit bus by data buffer KS1-2, link to each other with communication controler by data buffer ZK2, the multidigit bus links to each other with output bus by data buffer KP1-2 and also links to each other with communication port Port, sends bus, reception bus, multidigit bus and output bus and is 16 bit parallel buses.
The control line of communication controler is by transmit control device and transmission buffer zone, address counting FS, finishing control FS and data buffer KF1-1, KF1-2, KP1-1, KP1-2 link to each other, address counting FS, and finishing control FS provides 16 bit address of the autoincrement mode that sends buffer zone jointly; The control line of communication controler is by receiving controller and send buffer, and JS is counted in the address, and finishing control JS and data buffer KS1-1, KS1-2, KP1-1, KP1-2 link to each other.
Host computer links to each other with the transmission buffer zone with the transmission bus by impact damper KSZ1, this is host computer and the data path that sends buffer zone, host computer links to each other with send buffer with the reception bus by impact damper KSZ2, and they constitute jointly, and host computer leads to and the data path of send buffer.
2, according to the parallel communications processor of the described support computer group of claim 1 parallel computation, it is characterized in that: described module 1-5 is five interface cards, and five interface cards are connected with module 01,02,03 by five interface sockets.
3, according to the parallel communications processor of claim 1 or 2 described support computer group parallel computations, it is characterized in that: described five interface cards can be connected with five host computers respectively, are used for realizing many microcomputers interconnection network of various topological structures such as Hypercube, annular, tree-like, a peacekeeping two-dimensional array.
4, according to the parallel communications processor of the described support computer group of claim 1 parallel computation, it is characterized in that: multicomputer can be simultaneously on the net physically multi-bit parallel transmission information, realized the overlapping of calculating and communication process.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 98114065 CN1125406C (en) | 1998-06-08 | 1998-06-08 | Half-line communication processor supporting parallel computation of computer group |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 98114065 CN1125406C (en) | 1998-06-08 | 1998-06-08 | Half-line communication processor supporting parallel computation of computer group |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1204090A true CN1204090A (en) | 1999-01-06 |
CN1125406C CN1125406C (en) | 2003-10-22 |
Family
ID=5223745
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 98114065 Expired - Fee Related CN1125406C (en) | 1998-06-08 | 1998-06-08 | Half-line communication processor supporting parallel computation of computer group |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN1125406C (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1327337C (en) * | 2002-06-06 | 2007-07-18 | 因芬尼昂技术股份公司 | Processor and method for a simultaneous execution of a calculation and a copying process |
CN100359509C (en) * | 2002-11-26 | 2008-01-02 | 英特尔公司 | Asynchronous communication protocol using efficient data transfer formats |
CN100382070C (en) * | 2003-07-24 | 2008-04-16 | 国际商业机器公司 | Cluster data port services for clustered computer system |
CN106123943A (en) * | 2016-07-15 | 2016-11-16 | 苏州西斯派克检测科技有限公司 | A kind of flexible on-line detecting system based on EPA |
-
1998
- 1998-06-08 CN CN 98114065 patent/CN1125406C/en not_active Expired - Fee Related
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1327337C (en) * | 2002-06-06 | 2007-07-18 | 因芬尼昂技术股份公司 | Processor and method for a simultaneous execution of a calculation and a copying process |
CN100359509C (en) * | 2002-11-26 | 2008-01-02 | 英特尔公司 | Asynchronous communication protocol using efficient data transfer formats |
CN100382070C (en) * | 2003-07-24 | 2008-04-16 | 国际商业机器公司 | Cluster data port services for clustered computer system |
CN106123943A (en) * | 2016-07-15 | 2016-11-16 | 苏州西斯派克检测科技有限公司 | A kind of flexible on-line detecting system based on EPA |
CN106123943B (en) * | 2016-07-15 | 2019-05-21 | 苏州西斯派克检测科技有限公司 | A kind of flexible on-line detecting system based on Industrial Ethernet |
Also Published As
Publication number | Publication date |
---|---|
CN1125406C (en) | 2003-10-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ajima et al. | The tofu interconnect d | |
US8811422B2 (en) | Single chip protocol converter | |
US5303383A (en) | Multiprocessor computer system | |
CN1608255B (en) | Communicating transaction types between agents in a computer system using packet headers including an extended type/extended length field | |
CN1229739C (en) | Global interrupt and barrier network | |
CN1700700A (en) | Methods and apparatuses to effect a variable-width link | |
CN1608256A (en) | Communicating message request transaction types between agents in a computer system using multiple message groups | |
US20050132089A1 (en) | Directly connected low latency network and interface | |
JPH02196543A (en) | Star-type local area network | |
CN1493040A (en) | Global tree network for computing structures | |
CN1324499C (en) | Method for handling unexpected completion packets and completion packets with a non-successful completion status | |
CN1925453A (en) | Message transferring method and device | |
Yu et al. | High performance and reliable NIC-based multicast over Myrinet/GM-2 | |
CN1125406C (en) | Half-line communication processor supporting parallel computation of computer group | |
JP4088611B2 (en) | Single chip protocol converter | |
JPH0635874A (en) | Parallel processor | |
CN2819315Y (en) | MVB four-kind apparatus based on ARM7 kernel microprocessor | |
CN2350808Y (en) | Parallel communication processor capable of supporting parallel computation of computer group | |
Gupta et al. | Efficient collective operations using remote memory operations on VIA-based clusters | |
CN107102961A (en) | Accelerate the method and system of arm processor concurrent working | |
CN1979461A (en) | Multi-processor module | |
Lee | A virtual bus architecture for dynamic parallel processing | |
CN107122268B (en) | NUMA-based multi-physical-layer partition processing system | |
CN1464693A (en) | Method for controlling port interactive access of Ethernet switch chip | |
Li et al. | Fat-tree for local area multiprocessors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C19 | Lapse of patent right due to non-payment of the annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |