CN2569238Y - Reading local internal memory maintenance device by remote distance node in distributive shared internal memory system - Google Patents
Reading local internal memory maintenance device by remote distance node in distributive shared internal memory system Download PDFInfo
- Publication number
- CN2569238Y CN2569238Y CN 02240922 CN02240922U CN2569238Y CN 2569238 Y CN2569238 Y CN 2569238Y CN 02240922 CN02240922 CN 02240922 CN 02240922 U CN02240922 U CN 02240922U CN 2569238 Y CN2569238 Y CN 2569238Y
- Authority
- CN
- China
- Prior art keywords
- internal memory
- local
- remote node
- data
- local internal
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The utility model relates to a maintenance device for the reading of local inner memory by remote nodes in a distributed inner memory share system. The distinction types of a memory consistency directory which is maintained by the nodes of the original distributed inner memory share system are changed, and an HOME state is changed into an HOME-N state and an HOME-M state. When the remote node sends out a reading request of local inner memory lines, a DSM controller can known the storage position of latest data of the local inner memory lines according to the memory consistency directory and makes a response to send the latest data to the remote node within the shortest time without requiring to sending a system bus transaction again so as to acknowledge the actual storage position, and thus the data reading delay between the share inner memories so as to enhance the system efficiency.
Description
Technical field
The utility model relates to a kind of distributed shared memory (Distributed Shared Memory, be called for short DSM) multicomputer system (multiprocessors system), and be particularly related to a kind of under the DSM system architecture remote node read local internal memory attending device, especially a kind of distributed shared memory system node.
Background technology
In today that Science and Technology Day reaches increasingly, the informationization of various routine matters and business activity is handled has become the only effective means of promoting processing speed, also owing to so, cause the data processing amount of system huge day by day.For separate data,, just must rely on a kind of framework that is called the distributed shared memory multicomputer system with the purpose that reaches parallel processing and can share fast.
Please refer to shown in Figure 1ly, it is a kind of distributed shared memory (Distributed Shared Memory is hereinafter to be referred as DSM) system node framework.In general, each computer system architecture all can be considered a DSM system node 100.As shown in the figure, this computer system architecture can be the computer system architecture (certainly, the computer system architecture of a processor also is feasible) of multiprocessor.It has a plurality of processors 110,120, can be connected to an internal memory control chip 130 via a system bus 160, and is controlled the access action of 110,120 of local internal memory (Local Memory) 150 and a plurality of processors by internal memory control chip 130.And internal memory control chip 130 more can be connected to other device and reaches basic I/O (I/O) control.Usually, internal memory control chip 130 is a north bridge chips.
Wherein, local internal memory 150 is divided into a plurality of local internal memory lines (Local Memory Line), for store data.And the data access request that processor 110 and 120 can send the local internal memory line of access is to internal memory control chip 130 and reached the data access of local memory data bus by internal memory control chip 130.
Under DSM system node framework, this node (following local node that all is referred to as, Local Node) 100 is to reach data via a DSM controller (hereinafter to be referred as the DSM controller) 140 to transmit and receive with other node (the following remote node that all is referred to as, Remote Node).And DSM controller 140 is to be connected to system bus 160 and to be connected to internal memory control chip 130 with an internal bus 135 in this local node 100.
When processor 110,120 need read the data of internal memory line of remote node, just must utilize the DSM controller 140 in the local node 100 to link up, and then reach the data of the internal memory line in the access remote node with the DSM controller (not illustrating) in the remote node.
Because each local node 100 all can be shared the local internal memory 150 of oneself with other remote node, so must have a memory coherency directory (Memory CoherencyDirectory) in each node.This table of comparisons is the state that DSM controller 140 is used for writing down each local internal memory line in the local internal memory.When remote node desires to read the data of local internal memory line of local node 100, can be according to the state of the local internal memory line that writes down in the table of comparisons, decide and read action and how to carry out, to guarantee to read correct data.
Please refer to shown in Figure 2, its be according in the known memory coherency directory at the constitutional diagram of local internal memory line.Show among the figure that the known table of comparisons is divided into 4 kinds of states with local internal memory line, is respectively described below:
HOME (this locality): the data of this local internal memory line are read in representative without any remote node.
SHARED (sharing): existing other remote node of representative has read the data of local internal memory line, and the data of this local internal memory line are not modified as yet.
GONE (changing): the existing remote node of representative has read the data of local internal memory line, and the data of this local internal memory line are changed by remote node.
WASH: the data of the local internal memory line that was modified are passed back to the transition state of local node by remote node.
Next at the data variation of a local internal memory line and cooperate the change of these states to do an explanation.When original state, the state corresponding to this local internal memory line in the memory coherency directory in the local node is the HOME state.Under the HOME state, the data of this local internal memory line of processor access arbitrarily in this local node, and the data in this local internal memory line exist only in this local node.
Under the HOME state, shown in path 1,, represent remote node will read the data of local internal memory line when remote node sends long-range reading (remote readline).At this moment, the data of local internal memory line can be transferred into remote node, and therefore, the memory coherency directory in the local node can change the state of this local internal memory line into the SHARED state by the HOME state.That is under the SHARED state, the data of local internal memory line are stored in local node and remote node simultaneously.
Under the SHARED state, shown in path 2,, represent local node will change the data of local internal memory line when local node sends this locality when reading invalid (localread invalidate) or local invalid (local invalidate).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the HOME state by the SHARED state.Under the HOME state, the data of this local internal memory line of processor access arbitrarily in this local node, and the data in this local internal memory line exist only in this local node.And that therefore the former old data of the data rights and wrongs that are stored in the local internal memory line in the remote node can become is invalid.
Under the SHARED state, shown in path 4, when remote node sends long-range data (Remote rollout of shared copy) that discharge this local internal memory line, represent remote node to abandon having the data of this local internal memory line.At this moment, the memory coherency directory in the local node can also must change the state of this local internal memory line into the HOME state by the SHARED state.Under the HOME state, the data of this local internal memory line of processor access arbitrarily in this local node, and the data in this local internal memory line exist only in this local node.
Under the SHARED state, shown in path 7, when remote node sends long-range invalid (Remoteinvalidate), represent remote node will change the data of this local internal memory line.At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the GONE state by the SHARED state.Under the GONE state, representing the data of the local internal memory line in the local node is old incorrect data, and the data of the local internal memory line of correct change are stored in remote node now.
Under the HOME state, shown in path 6,, represent remote node will read and change the data of this local internal memory line when remote node sends long-rangely when reading invalid (Remoteread invalidate).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the GONE state by the HOME state.Under the GONE state, representing the data of the local internal memory line in the local node is old incorrect data, and the data of the local internal memory line of correct change are stored in remote node now.
Under the GONE state, shown in path 8, when remote node sends the data (Remote rollout of modified copy) of disengaging this local internal memory line, represent remote node the data of revising local internal memory line later will be released and be back to local node.Memory coherency directory in the local node can change the state of this local internal memory line into the HOME state by the GONE state.At this moment, the data of change local internal memory line are later got back to local node again.
Under the GONE state, shown in path 5, when local node sends this locality and reads (Local readline) or invalid (Local read invalidate) read in this locality, represent local node will read the data of this local memory data bus or read after to revise.At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the WASH state by the GONE state.The data of the local internal memory line of this WASH state system representative are passed back to the transition state of local node by remote node.
Under the WASH state, shown in path 9, read and finish the reception of the data of local internal memory line (Completion of local read line) when local node sends this locality, the data of representing local node only will read this local memory data bus do not make an amendment.At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the SHARE state by the WASH state.That is under the SHARE state, the data of local internal memory line are stored in local node and remote node simultaneously.
Under the WASH state, shown in path 10, read invalid and when finishing the reception (Completion of local read invalidate) of the data of local internal memory line when local node sends this locality, represent local internal memory will read the data of this local memory data bus and make an amendment.At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the HOME state by the WASH state.That is under the HOME state, the data of local internal memory line only can be stored in local node.
Above description is the flow process of sharing local internal memory (local memory) between the known DSM system node.So the DSM controller of each node must be safeguarded a memory coherency directory (memorycoherency directory), this table of comparisons is used for writing down the state of each local internal memory line (localmemory line) in the local internal memory, when remote node is desired to read the local internal memory of this local node, can be according to the state of the local internal memory line that writes down in the table of comparisons, decide and read action and how to carry out, to guarantee to read correct data.
Yet, in known distributed shared memory multicomputer system, in HOME state following time.Please refer to Fig. 1, when remote node will read the data of local internal memory line, at first, the DSM controller of local node 100 140 can send a system bus transaction (System BusTransaction) at system bus 160, in order to the place, position of the data reality of inquiring this local internal memory line.
(I) the data system of supposing this local internal memory line exists in the high-speed cache (cachememory) of processor 110 or 120, then processor 110 or 120 can send a cache hit (Hit) order, and the data of sending this local internal memory line are to system bus 160, and write back local internal memory 150 by internal memory control chip 130, simultaneously, DSM controller 140 is also received the data of this local internal memory line and is delivered to remote node by system bus 160.And the interior memory coherency directory of local node this moment can be the SHRAE state with the state of this local internal memory line.
(II) the data system of supposing this local internal memory line is stored in the local internal memory 150, and then processor 110 or 120 can not send cache hit (Hit) order, and this memory data bus of this interval scale is present in the local internal memory 150.Therefore, DSM controller 140 can utilize internal bus 135 to require internal memory control chip 130 that the data of local internal memory line are delivered to DSM controller 140, delivers to remote node afterwards.And the interior memory coherency directory of local node this moment can be the SHRAE state with the state of this local internal memory line.
Because the HOME state of record in the memory coherency directory in the local node, when remote node requires to read the data of local memory data bus, the correct storage location of having no idea to learn the data of this local memory data bus (may exist in the local internal memory, in the high-speed cache, perhaps the two).So DSM controller 140 must send the place, position of the data reality of this local internal memory line of system bus transaction inquiry.And when above-mentioned (II) situation, because the data of local memory data bus have existed in the local internal memory 150, and DSM controller 140 still needs to send the conclude the business storage location of the data of confirming this local memory data bus of a system bus.Therefore, the transaction of this system bus not only can increase the workload of system bus 160, and can postpone to read the time of the data of local internal memory line.
Summary of the invention
In view of this, the purpose of this utility model is to provide a kind of remote node to read a kind of distributed shared memory system node in the local internal memory maintaining method, be used for remote node and read a local internal memory line, comprise the following steps: at least to judge whether the latest data of this this locality internal memory line is positioned at the local internal memory of local node to the local node requirement; And when latest data is positioned at local internal memory, asks for latest data and deliver to remote node via internal bus.
For reaching above-mentioned and other purpose, the utility model provides a kind of distributed shared memory system remote node to read local internal memory attending device, comprising: at least one processor; One system bus is coupled to this at least one processor; One internal memory control chip is coupled to this system bus; One local internal memory is coupled to this internal memory control chip, wherein should be divided into a plurality of local internal memory lines in memory field, this locality; One DSM controller is coupled to this system bus; One internal bus is coupled to this DSM controller and this internal memory control chip;
Wherein, being provided with a remote node in the described processor reads signal and accepts module, this reads signal and accepts module and link to each other with described DSM controller, in the described DSM controller judge module is arranged, this judge module this specific local internal memory line in this this locality internal memory sends the judgement signal, be positioned at this latest data of this this locality internal memory, can deliver to this remote node via this internal bus, that is: a remote node sends a specific local internal memory line that reads in this this locality internal memory to processor and reads signal, whether this DSM controller sends this specific local internal memory line in this this locality internal memory a latest data is positioned at this this locality internal memory is judged signal, in other words, when a remote node requires to read a specific local internal memory line in this this locality internal memory, this DSM controller judges whether a latest data of this specific local internal memory line is positioned at this this locality internal memory, and when this latest data is positioned at this this locality internal memory, asks for this latest data and deliver to this remote node via this internal bus.
This latest data is not positioned at this this locality internal memory, and this judge module in this DSM controller sends the judgement signal to described processor, is positioned at this latest data of described processor, can deliver to this remote node via this system bus; In other words, when this DSM controller is judged this latest data and is not positioned at this this locality internal memory, can judge again whether this latest data is positioned at this at least one processor, and when this latest data is positioned at this at least one processor, sends a system bus via this system bus and conclude the business and ask for this latest data and deliver to this remote node.
One judge module of safeguarding that can safeguard a memory coherency directory is arranged, in order to judge the position of this latest data in this DSM controller.
For reaching above-mentioned and other purpose, the utility model provides a kind of remote node to read local internal memory attending device, be used for first remote node and read local internal memory line to the local node requirement, at least comprise: the HOME state of this this locality internal memory line can be divided into two states of HOME-N, HOME-M, when this this locality internal memory line during, directly receive the data of local internal memory line and deliver to first remote node to local internal memory at HOME-N or SHARED state; When this this locality internal memory line during, directly read the data of local internal memory line and deliver to first remote node to second remote node at the GONE state; And, when this this locality internal memory linear system during, then on system bus, initiate a transaction in order to reading the data of this this locality internal memory line, and receive the data of local internal memory line and deliver to this first remote node by system bus at the HOME-M state.
By in the above-mentioned explanation as can be known, use distributed shared memory provided by the utility model system remote node and read local internal memory maintaining method and application apparatus thereof, when when request of reading that remote node sends local internal memory line, the DSM controller can be learnt the storage location of latest data, and within the shortest time, latest data is responded to remote node, and need not send the system bus transaction again, to confirm its actual storage location, so the data read that can reduce between shared drive postpones.
For above-mentioned and other purpose, feature and advantage of the present utility model can be become apparent, a preferred embodiment cited below particularly, and conjunction with figs. are described in detail below.
Description of drawings
Fig. 1 is a kind of distributed shared memory multicomputer system node block schematic diagram according to the utility model preferred embodiment;
Fig. 2 is at the constitutional diagram of local internal memory line in the known memory coherency directory;
Fig. 3 be according in the utility model memory coherency directory at the constitutional diagram of local internal memory line; And
Fig. 4 reads local internal memory maintaining method process flow diagram for a kind of remote node according to the utility model preferred embodiment.Label declaration:
100 distributed shared memory multicomputer system nodes
110 processors
120 processors
130 internal memory control chips
135 internal buss
The 140DSM controller
150 local internal memories
160 system buss
Embodiment
Please refer to shown in Figure 3ly, it is at the constitutional diagram of local internal memory in the utility model memory coherency directory.For the usefulness of elevator system bus, the utility model provides HOME-M state and HOME-N to replace HOME state in the known memory coherency directory.Wherein:
HOME-N: representative is read this local internal memory line without any remote node, and the data of up-to-date local internal memory line have been stored in the local internal memory.
HOME-M: representative is read this local internal memory line without any remote node, and the data of up-to-date local internal memory line are in local internal memory, but in the high-speed cache of processor.
Next at the variation of local internal memory line data and cooperate the conversion of these states to do an explanation.When original state, the state corresponding to this local internal memory line in the memory coherency directory in the local node (local node) is the HOME-N state.Under the HOME-N state, the local internal memory in this local node has the data of up-to-date local internal memory line.
Under the HOME-N state, shown in path 1,, represent remote node will read the data of local internal memory line when remote node sends long-range reading (Remoteread line).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the SHARED state by the HOME-N state.That is under the HOME-N state, the DSM controller can directly require the data of local internal memory line and deliver to remote node to the internal memory control chip via internal bus.Therefore, the DSM controller also need not send the system bus transaction.Moreover under the SHARED state, the data of local internal memory line are stored in local node and remote node simultaneously.
Under the HOME-N state, shown in path 2, when local node sends this locality when reading invalid (Localread invalidate) or local invalid (Local invalidate), represent the processor of local node will change the data of local internal memory line.At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the HOME-M state by the HOME-N state.Under the HOME-M state, the processor in this local node can be changed the data of this local internal memory line, and is stored in the high-speed cache.
Under the HOME-N state, shown in path 6,, represent remote node will read and change the data of this local internal memory line when remote node sends long-rangely when reading invalid (Remote read invalidate).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the GONE state by the HOME-N state.Under the GONE state, representing the data of the local internal memory line in the local node is old incorrect data, and the data of up-to-date local internal memory line are stored in remote node now.And under the HOME-N state, the DSM controller can directly require the data of local internal memory line and deliver to the modification that remote node supplies to the internal memory control chip via internal bus.Therefore, the DSM controller also need not send the system bus transaction.
Under the HOME-M state, shown in path 1,, represent remote node will read the data of local internal memory line when remote node sends long-range reading (Remoteread line).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the SHARED state by the HOME-M state.That is under the SHARED state, the data of local internal memory line are stored in local node and remote node simultaneously.And under the HOME-M state, the DSM controller must send system bus transaction make processor send that cache hit (Hit) is ordered and data that up-to-date local internal memory line is provided to system bus, in order to be stored to local internal memory, the DSM controller also can receive and deliver to remote node simultaneously.
Under the HOME-M state, shown in path 6,, represent remote node will read and change the data of this local internal memory line when remote node sends long-rangely when reading invalid (Remote read invalidate).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the GONE state by the HOME-M state.Under the GONE state, representing the data of the local internal memory line in the local node is old incorrect data, and the data of up-to-date local internal memory line are stored in remote node now.And under the HOME-M state, the DSM controller must send system bus transaction make processor send to go hit (Hit) order and data that up-to-date local internal memory line is provided to the DSM controller and deliver to the change that remote node supplies.
Under the HOME-M state, shown in path 3, when local node sends this locality and reads (Local readline), represent the data of local internal memory line to write back local internal memory by high-speed cache.Memory coherency directory in the local node can change the state of this local internal memory line into the HOME-N state by the HOME-M state.Under the HOME-N state, at this moment, the local internal memory in the local node has the data of up-to-date local internal memory line.
Under the SHARED state, shown in path 2,, represent local node will change the data of local internal memory line when local node sends this locality when reading invalid (Localread invalidate) or local invalid (Local invalidate).At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the HOME-M state by the SHARED state.Under the HOME-M state, the processor in this local node can be changed the data of this local internal memory line, and is stored in the high-speed cache.
Under the SHARED state, shown in path 4, when remote node sends long-range data (Remote rollout of shared copy) that discharge this local internal memory line, represent remote node to abandon having the data of this local internal memory line.At this moment, the memory coherency directory in the local node can also must change the state of this local internal memory line into the HOME-N state by the SHARED state.Under this state, the data of local internal memory line system is stored in the local internal memory of local node.
Under the GONE state, shown in path 8, when remote node sends the data (remote rollout of modified copy) of disengaging this local internal memory line, represent remote node the data of revising local internal memory line later will be released and be back to local node.Memory coherency directory in the local node can change the state of this local internal memory line into the HOME-N state by the GONE state.Revising the data of local internal memory line later gets back in the local internal memory of local node again.
Under the WASH state, shown in path 10, read invalid and when finishing the reception (Completion of local read invalidate) of the data of local internal memory line, represent local internal memory will read the data of this local memory data bus and make an amendment when local node sends this locality.At this moment, the memory coherency directory in the local node can change the state of this local internal memory line into the HOME-M state by the WASH state.That is under the HOME-M state, the data of local internal memory line can be changed and be stored in the high-speed cache by processor.
Above-mentioned for the utility model changes known HOME state after HOME-M state and the HOME-N state into, the relation between these two new states and SHARED state, GONE state and the WASH state.When the HOME-N state, no matter skip to SHARED state or GONE state, the DSM controller can directly require the internal memory control chip that the data of local internal memory line directly are provided via internal bus, do not need again extra the conclude the business storage location of the data of inquiring this local internal memory line of system bus that sends, therefore can reduce the utilization rate of system bus, promote the task performance of local node.As for the transformational relation between SHARED state, GONE state and the WASH state, then with known identical repeating no more.
Above-mentioned is memory coherency directory state variation maintenance instruction in the node, cooperates with reference to this memory coherency directory, and then the DSM controller can develop the remote node that as shown in Figure 4 and reads local internal memory maintaining method.At first, receive the request of reading (S300) of the local internal memory line that sends from remote node.Judge then whether the local internal memory line states in the table of comparisons is HOME-N state or SHARED state (S305).
When at HOME-N state or SHARED state, represent the data of local internal memory line just being stored in the local internal memory, therefore, directly via internal bus to the data (S310) of this local internal memory line of local request memory and respond the data (S315) of this local internal memory line of remote node.
When now not at HOME-N state or SHARED state, judge then whether the local internal memory line states in the table of comparisons is GONE state (S315).When at the GONE state, represent the data of local internal memory line just being stored in other remote node, therefore, the data of direct this local internal memory line after the remote node of the data that have local internal memory line reads change are also deposited back local internal memory (S325), and the data (S315) of this local internal memory line of response remote node.
When now not at the GONE state, judge then whether the local internal memory line states in the table of comparisons is HOME-M state (S330).When at the HOME-M state, represent the data of local internal memory line just being stored in the high-speed cache of processor, therefore, directly send the system bus transaction, when the data that respond this local internal memory line when processor are returned local internal memory, capture the data (S335) of this local internal memory line, and the data (S315) of this local internal memory line of response remote node.
When now not at the HOME-M state, judge that then the local internal memory line states in the table of comparisons is WASH state (S340).Because the data of the local internal memory line that the representative of WASH state was modified are passed back to the transition state of local node by remote node.Therefore get back to the S305 step, rejudge the state of its change, read operation program according to what its state decision was carried out again.
In sum, the utility model has following advantage at least:
1 when remote node is desired to read local internal memory, need not repeat to send via system bus the system bus transaction of the Data Position of confirming local internal memory line under the state of MOME-N, can save the burden of system bus, and then promote its usefulness.
2 can reduce the delay of reading that remote node reads local internal memory.
Though the utility model with a preferred embodiment openly as above; right its is not in order to limit the utility model; any those of ordinary skills; in not breaking away from spirit and scope of the present utility model; when doing various equivalence changes and retouching, therefore protection domain of the present utility model is as the criterion with claim.
Claims (9)
1. a distributed shared memory system remote node reads local internal memory attending device, it is characterized in that, comprising:
At least one processor;
One system bus is coupled to this at least one processor;
One internal memory control chip is coupled to this system bus;
One local internal memory is coupled to this internal memory control chip, wherein should be divided into a plurality of local internal memory lines in memory field, this locality;
One can judge that whether a latest data is positioned at the DSM controller of this this locality internal memory, is coupled to this system bus;
One internal bus is coupled to this DSM controller and this internal memory control chip;
Wherein, described DSM controller comprises that also the specific local internal memory line of a judge module and this links to each other.
2. distributed shared memory as claimed in claim 1 system remote node reads local internal memory attending device, it is characterized in that, described DSM controller links to each other with system bus, and this system bus sends a system bus and concludes the business and ask for the latest data that is positioned at this at least one processor and deliver to this remote node.
3. distributed shared memory as claimed in claim 2 system remote node reads local internal memory attending device, it is characterized in that, has the conforming table of comparisons in the storage area of maintenance in the described DSM controller.
4. a distributed shared memory system remote node reads local internal memory attending device, it is characterized in that, comprising:
At least one processor;
One system bus is coupled to this at least one processor;
One internal memory control chip is coupled to this system bus;
One local internal memory is coupled to this internal memory control chip, wherein should be divided into a plurality of local internal memory lines in memory field, this locality;
One DSM controller is coupled to this system bus;
One internal bus is coupled to this DSM controller and this internal memory control chip;
Wherein, comprise in the described processor that a remote node reads signal and accepts module, this reads signal and accepts module and link to each other with described DSM controller, described DSM controller also comprises a judge module, this judge module this specific local internal memory line in this this locality internal memory sends the judgement signal, be positioned at this latest data of this this locality internal memory, can deliver to this remote node via this internal bus.
5. distributed shared memory as claimed in claim 4 system remote node reads local internal memory attending device, it is characterized in that, this judge module in described this DSM controller sends the judgement signal to described processor, be positioned at this latest data of described processor, can deliver to this remote node via this system bus.
6. distributed shared memory as claimed in claim 5 system remote node reads local internal memory attending device, it is characterized in that, a judge module of safeguarding of safeguarding a memory coherency directory is arranged in this DSM controller.
7. a remote node reads local internal memory attending device, is used for first remote node and reads local internal memory line to the local node requirement, it is characterized in that, this attending device directly receives the data of local internal memory line and delivers to first remote node to local internal memory; Also directly read the data of local internal memory line and deliver to first remote node to second remote node; Also on system bus, initiate a transaction in order to reading the data of this this locality internal memory line, and receive the data of local internal memory line and deliver to this first remote node by system bus.
8. remote node as claimed in claim 7 reads local internal memory attending device, it is characterized in that, this HAME-N of this this locality internal memory line, this HOME-M, this SHARED state are stored in a memory coherency directory.
9. remote node as claimed in claim 8 reads local internal memory attending device, it is characterized in that, memory coherency directory also stores a GONE state.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 02240922 CN2569238Y (en) | 2002-07-01 | 2002-07-01 | Reading local internal memory maintenance device by remote distance node in distributive shared internal memory system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 02240922 CN2569238Y (en) | 2002-07-01 | 2002-07-01 | Reading local internal memory maintenance device by remote distance node in distributive shared internal memory system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN2569238Y true CN2569238Y (en) | 2003-08-27 |
Family
ID=33713890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 02240922 Expired - Fee Related CN2569238Y (en) | 2002-07-01 | 2002-07-01 | Reading local internal memory maintenance device by remote distance node in distributive shared internal memory system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN2569238Y (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100383763C (en) * | 2004-02-27 | 2008-04-23 | 中国人民解放军国防科学技术大学 | Page transport and copy method based on operation system reverse page table |
| CN100486178C (en) * | 2006-12-06 | 2009-05-06 | 中国科学院计算技术研究所 | A remote internal memory sharing system and its realization method |
| CN102147740A (en) * | 2010-02-08 | 2011-08-10 | 微软公司 | Fast machine booting through streaming storage |
| US10025509B2 (en) | 2010-02-08 | 2018-07-17 | Microsoft Technology Licensing, Llc | Background migration of virtual storage |
-
2002
- 2002-07-01 CN CN 02240922 patent/CN2569238Y/en not_active Expired - Fee Related
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100383763C (en) * | 2004-02-27 | 2008-04-23 | 中国人民解放军国防科学技术大学 | Page transport and copy method based on operation system reverse page table |
| CN100486178C (en) * | 2006-12-06 | 2009-05-06 | 中国科学院计算技术研究所 | A remote internal memory sharing system and its realization method |
| CN102147740A (en) * | 2010-02-08 | 2011-08-10 | 微软公司 | Fast machine booting through streaming storage |
| CN102147740B (en) * | 2010-02-08 | 2017-01-18 | 微软技术许可有限责任公司 | Fast machine booting through streaming storage |
| US10025509B2 (en) | 2010-02-08 | 2018-07-17 | Microsoft Technology Licensing, Llc | Background migration of virtual storage |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101042679A (en) | Method and system for maintenance memory consistency | |
| CN1157659C (en) | Non-uniform memory access (NUMA) data processing system that speculatively forwards read reguest to remote processing node | |
| CN1142503C (en) | Same stage-to-same stage cache memory moving in data processing system of multiple processors | |
| CN1276362C (en) | Method and system for data transmission in multi-processor system | |
| CN1116643C (en) | Non-uniform memory reference data processing system with shared intervention support | |
| CN1133933C (en) | Mixed HUMA/S-COMA system and method | |
| US10459661B2 (en) | Stream identifier based storage system for managing an array of SSDs | |
| CN1717645A (en) | Apparatus and method for multi-threaded processors performance control | |
| CN1447257A (en) | Data maintenance method for distributed type shared memory system | |
| CN1991795A (en) | System and method for information processing | |
| CN101055555A (en) | Non-volatile memory sharing system for multiple processors and method thereof | |
| CN1892605A (en) | Dynamic cluster code management method and system | |
| CN1932791A (en) | Method and apparatus for data processing | |
| CN101063957A (en) | System and method for managing replacement of sets in a locked cache | |
| CN1152330C (en) | Maintain method for remote node to read local memory and its application device | |
| CN1280732C (en) | Isomeric proxy cache memory consistency and method and apparatus for limiting transmission of data | |
| CN1851677A (en) | Embedded processor system and its data operating method | |
| CN1645352A (en) | Method for managing data in an array processor and array processor carrying out this method | |
| CN1898653A (en) | Method and apparatus for efficient ordered stores over an interconnection network | |
| CN1287361A (en) | Real-time processing method for flash storage | |
| CN1286016C (en) | Managing coherence via put/get windows | |
| CN1896972A (en) | Method and device for converting virtual address, reading and writing high-speed buffer memory | |
| CN2569238Y (en) | Reading local internal memory maintenance device by remote distance node in distributive shared internal memory system | |
| CN1464415A (en) | Multi-processor system | |
| RU2484520C2 (en) | Adaptive cache organisation for single-chip multiprocessors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CU01 | Correction of utility model patent |
Correction item: Priority Correct: 60/371,194 Number: 35 Volume: 19 |
|
| CU03 | Correction of utility model patent gazette |
Correction item: Priority Correct: 60/371,194 Number: 35 Volume: 19 |
|
| 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 |