CN101231584A - Processing system as well as correlation technique for processing instruction thereof - Google Patents
Processing system as well as correlation technique for processing instruction thereof Download PDFInfo
- Publication number
- CN101231584A CN101231584A CNA2007100040776A CN200710004077A CN101231584A CN 101231584 A CN101231584 A CN 101231584A CN A2007100040776 A CNA2007100040776 A CN A2007100040776A CN 200710004077 A CN200710004077 A CN 200710004077A CN 101231584 A CN101231584 A CN 101231584A
- Authority
- CN
- China
- Prior art keywords
- instruction
- flow process
- count value
- process control
- program counter
- 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
Images
Abstract
A preferable embodiment of the invention discloses a method of processing predetermined commands in a processing system. The processing system is provided with a plurality of processing units. The method comprises the following steps of: providing a global program counter; determining count values of the global program counter to be used as predetermined commands to be executed; assigning a local program counter to each processing unit and determining a count value of the local program counter according to a current command executed by the processing units; and enabling at least one processing unit to execute a specified command of the predetermined commands according to the count values stored in the local program counter and the global program counter.
Description
Technical field
The present invention relates to the method for predetermined instruction that a kind of processing comprises the disposal system of a plurality of processing units, be particularly related to a kind of disposal system especially, it has a plurality of processing units that the local program of each processing unit counter is provided.And the present invention is relevant for the correlation technique of the predetermined instruction in this processing system of processing.
Background technology
Generally speaking, single instruction multiple data (Single Instruction Multiple Data, SIMD) instruction execution flow of disposal system control (flow control, or be called branch control branchcontrol) all suitable difficulty, because a plurality of different pieces of informations for same branch instruction processing, may cause different branch outcome, but be subject to this processing system and only can interiorly at one time carry out single the instruction in the characteristic of a plurality of data, manage herein under the system architecture so need some special skills to solve, carry out the problem of different instruction flow process.The control of bank shell type flow process has more worsened this kind problem.There are some existing methods can solve partly problem, solve aforesaid bank shell type flow process control problem but wasted hardware resource.
Usually in the field of handling of drawing, all utilize the parallel processing architected features of SIMD to handle many data doing similar computing.Yet, usually suitable limited of SIMD.For instance, can handle many different data simultaneously though more common problem is most SIMD processing units of parallel processing system, each processing unit all must be carried out the specified instruction of same programmable counter (program counter).In fact, traditional SIMD processing unit only has a programmable counter.
Therefore, needing novel method and apparatus addresses the above problem.
Summary of the invention
Therefore, one of purpose of the present invention provides a disposal system with most processing units to solve foregoing problems, and wherein, processing unit has the local program counter.The present invention more provides the relevant method of disposal system therewith, in order to handle the predetermined instruction in disposal system.
Preferred embodiment of the present invention has disclosed the disposal route of the predetermined instruction in a kind of disposal system, and this processing system has most processing units, and the method comprises:
One overall situation (global) programmable counter is provided, and sets the count value of this global program's counter, with instruction as the described predetermined instruction that is performed; Specify a local program counter to give each processing unit, and a count value of this this locality (local) programmable counter is set in a performed instruction now according to this processing unit; And enable at least one processing unit to carry out a specific instruction of described predetermined instruction according to the count value of this local program counter that is stored in this processing unit and this global program's counter.
Preferred embodiment of the present invention has disclosed a kind of in order to handle the disposal system of predetermined instruction.This processing system comprises: an instruction buffer, in order to receive and to cushion this predetermined instruction; One global program's counter is coupled to this instruction buffer, and in order to storing a count value, and this count value is the instruction as this predetermined instruction that is performed; Most processing units, wherein each comprises: a performance element, in order to execution command; One this locality (local) programmable counter is in order to a now instruction setting one count value performed according to this performance element; And first-class process control unit, be coupled to this global program's counter and each processing unit, in order to according to being stored in this local program counter of this processing unit and the count value of this global program's counter enables at least one processing unit to carry out a specific instruction of described predetermined instruction.
Description of drawings
Fig. 1 shows the calcspar in order to the disposal system of handling predetermined instruction according to one embodiment of the invention.
Fig. 2 shows according to an embodiment of the invention in order to handle the method for predetermined instruction.
The reference numeral explanation
100 disposal systems
105 most processing units
106 performance elements
107 local program counters
108 write back the unit
109 register file
110 instruction buffers
120 global program's counters
130 instructions obtain/decoding unit
140 flow process control modules
Embodiment
In the middle of instructions and follow-up claim, used some vocabulary to censure specific assembly.The person with usual knowledge in their respective areas should understand, and hardware manufacturer may be called same assembly with different nouns.This instructions and follow-up claim are not used as distinguishing the mode of assembly with the difference of title, but the criterion that is used as distinguishing with the difference of assembly on function.Be an open term mentioned " comprising " in the middle of instructions and the follow-up request item in the whole text, so should be construed to " comprise but be not limited to ".In addition, " coupling " speech is to comprise any indirect means that are electrically connected that directly reach at this.Therefore, be coupled to one second device, then represent this first device can directly be electrically connected in this second device, or be electrically connected to this second device indirectly by other device or connection means if describe one first device in the literary composition.
Please refer to Fig. 1, Fig. 1 shows the calcspar in order to the disposal system 100 of handling predetermined instruction according to one embodiment of the invention.In Fig. 1, small arrow symbology one control path, on behalf of which kind of action, it desire the execution result that is performed and desires to be written into specific register.And big arrow is represented a data routing, and data routing has comprised instruction and data.Disposal system 100 comprises an instruction buffer 110.Instruction buffer 110 is in order to receive and the processed predetermined instruction of buffering desire.Global program's counter 120 is coupled to instruction buffer 110, in order to store a count value with the instruction as this predetermined instruction that is performed.Disposal system 100 also comprises most processing units 105, and wherein each all has a performance element 106 and a local program counter 107.Performance element 106 is stored in predetermined instruction in the instruction buffer 110 in order to execution.Local program counter 107 in order to according to performance element 106 a performed instruction now set a count value.
In addition, disposal system 100 comprises first-class process control unit 140.Flow process control module 140 is coupled to global program's counter 120 and each most processing unit 105.Flow process control module 140 is in order to enable at least one processing unit 105 to carry out a specific instruction of predetermined instruction according to the count value that is stored in local program counter 107 in the processing unit 105 and global program's counter 120.
When the flow process control module 140 of disposal system 100 specific local program counter 107 of 105 in being stored in particular processor unit has the count value that equates with global program counter 120, enable a particular processor unit 105 to carry out by the specific instruction of global program's counter 120 (point to) pointed.
In addition, when the specific local program counter 107 in the particular processor unit 105 is not enabled with the execution specific instruction by flow process control module 140 in the particular processor unit 105 of processing unit 105, keep a specific count value that is stored in specific local program counter 107.
If specific instruction be first-class process control instruction, the flow process control module 140 of disposal system 100 is controlled specific local program counter 107 and is stored in specific count value in the specific local program counter 107 with increase.
If specific instruction is first-class process control instruction, the performance element 106 calculation process steering orders of particular processor unit 105, and flow process control module 140 updates stored in specific count value in the specific local program counter 107 according to the result of calculation of performance element 106.
Flow process control module 106 is not when flow process steering order (or branch instruction) is judged into branch operation and will carries out (branch not taken), increase the specific count value that is stored in the specific local program counter 107, and flow process control module 106 is assigned corresponding to a predicted target address of another instruction and is given the specific count value that is stored in the specific local program counter 107 when flow process steering order (or branch instruction) is judged into branch operation and will carries out (branch taken).
In another embodiment of the present invention, each processing unit 105 comprises a call state position (callstatus bit), when carries out an instruction of calling out piece (call block) or bank cover calling piece (nested call block) in order to expression processing unit 105.Suitable important of this characteristic because if local program counter 107 has minimum local program count value, will make corresponding processing unit 105 execution commands.Each processing unit 105 can have a call state position, and when processing unit 105 was carried out a call instruction (Call) that enters a calling piece or bank cover calling piece, the call state position was set.And when processing unit 105 execution were present in non-bank cover calling piece or be positioned at an answer instruction (Return) of top bank cover calling piece, the call state position was eliminated.In this example, the processing unit 105 with call state position of setting at first is regarded as a group, and in this group, if processing unit 105 has minimum local program count value, will make corresponding processing unit 105 execute instruction.When not having the call state position to be set, then get back to previous rule, by this, flow process control module 140 can be from processing unit 105, choosing the corresponding local program count value with minimum sets global program's counter 120 and grasps instruction, that is processing unit 105 has minimum corresponding local program count value person and can execute instruction, and the call state position can't be influential in the case.
Easily it, the processing unit 105 that has the call state position of setting in this embodiment has the priority that obtains and execute instruction.Processing unit 105 the processing units 105 than the call state position that do not have setting of method with call state position of setting preferentially are enabled by this.For example, one first processing unit has M local program count value, and this local program count value group has the call state position of setting, and one second processing unit has M-1 local program count value, and this local program count value group has the call state position of removing.In the case, because have the call state position of setting, the instruction pointed by the M local program count value of higher value obtained and carried out on the contrary.
At last, more comprise one in order to the processing unit 105 of the disposal system 100 of handling predetermined instruction and write back (write-back) unit 108 and a register file 109.In an embodiment of the present invention, flow process control module 140 is used to control execution unit 106 to carry out the instruction of following the flow process steering order according to local program counter 107.Flow process control module 140 also is used to control and writes back unit 108 execution result is write register file 109.
Above-mentioned disposal system 100 can be a parallel processing system, and processing unit 105 can be a parallel processing unit, but only in order to for example, is not in order to limit the present invention.Know this skill person when knowing that the present invention and correlation technique are not to need this type of parallel processing system, the system relevant with the parallel processing system also can use at system and method for the present invention, make the present invention be able to integrate with other existing apparatus and method, it also within the scope of the present invention.
Please refer to Fig. 2, Fig. 2 shows according to an embodiment of the invention in order to handle the method for predetermined instruction.The method is corresponding to the disposal system described in Fig. 1 100, and its step is as described below:
Step 200
Beginning
Step 210
Set the count value of global program's counter 120, make its minimum value identical with the local program counter of each processing unit 105 of call state position that has setting respectively.If there is not the call state position of setting, then set the count value of global program's counter 120, make its minimum value identical with the local program counter of each processing unit 105.
Step 220
Obtain global program's counter 120 next instruction pointed.
Step 230
Judgement is for each processing unit, and whether does global program's counter 120 equate with local program counter 107? if, to step 280, if not, to step 240.
Step 240
Do you judge that instruction now is first-class process control instruction? if, to step 250.If not, to step 290.
Step 250
The computational fields that calculates instruction now is with decision flow process control result.To step 260.
Step 260
Whether the calculating of processing unit 105 determines to obtain flow process control result, just, carry out branch instruction? if, to step 295.If not, to step 270.
Step 270
Set local program counter 107, make it equal local program counter 107 and add one.To step 210.
Step 280
Keep local program counter 107 for not changing.Covering register file 109 is enabled to prevent to write.To step 210.
Step 290
Carry out instruction now, and set local program counter 107, make it equal local program counter 107 and add one.To step 210.
Step 295
Set local program counter 107, it is equaled and the relevant destination address of carrying out now of flow process steering order.If the execution of flow process instruction now enters one and calls out piece, then set the call state position.If the execution of flow process instruction is now left one and called out piece, then remove the call state position.To step 210.
Flow process of the present invention starts from step 200.In step 210, if there is not the call state position of setting, then set the count value of global program's counter 120, make its minimum value identical with the local program counter 107 of each processing unit 105.Yet as discussed previously, the local program counter 107 with call state position of setting is assigned the higher-order that obtains and execute instruction.Therefore, if at least one call state position is set, the flow process of this method will preferentially enable to have the processing unit 105 of the call state position of setting, rather than have the processing unit 105 of the call state position that is eliminated.As shown in Figure 2, global program's counter 120 minimum value of local program counter of being had each processing unit 105 of call state position of setting is respectively set.
Then in step 220, next instruction is obtained according to global program's counter 120.In step 230, to each processing unit 105, the present invention confirms global program's counter 120 (PC
Globa1) whether with local program counter 107 (PC
Loca1) identical.If global program's counter 120 is identical with specific local program counter 107, then flow process is to step 240, otherwise flow process is to step 280.In the step 240, because two programmable counters, global program's counter 120 and specific local program counter 107 have been confirmed to be equal, and therefore the present invention now must confirm whether instruction is the flow process steering order now.If instruction is now arrived step 250 for first-class process control instruction, otherwise to step 290.In step 250, it is first-class process control instruction that this moment of instruction now has been identified as, and therefore must calculating now, flow process steering order palpus is calculated with decision flow process control result.Then, to step 260.In step 260, if particular processor unit 105 calculates the result that will carry out branch operation (branch taken), easily it represents that promptly processing unit 105 calculation process steering orders and its will obtain flow process control, and the present invention is to step 295 then.If particular processor unit 105 calculates the result who does not carry out branch operation (branch not taken), then the present invention is to step 270.In step 270, local program counter 107 is configured to equal local program counter 107 and adds (a PC just,
Loca1=PC
Loca1+ 1), the present invention arrives step 210 then.
Get back to step 230, if global program's counter 120 equates that with specific local program counter 107 flow process proceeds to step 280, wherein the present invention makes local program counter 107 keep now value and does not change.In addition, register file 109 cresteds (mask) are enabled to prevent to write.Then, flow process is got back to step 210.
In the step 240, because two programmable counters, global program's counter 120 and specific local program counter 107 have been confirmed to be equal, and therefore the present invention now must confirm whether instruction is the flow process steering order now.If instruction now is not first-class process control instruction, to step 290.In step 290, performance element 106 is carried out instruction now.In addition, local program counter 107 is configured to equal local program counter 107 and adds (a PC just,
Loca1=PC
Loca1+ 1), the present invention arrives step 210 then.
Get back to step 260, in step 260, if particular processor unit 105 calculates the result that will carry out branch operation (branch taken), easily it, promptly branch operation is performed, the present invention is to step 295 then.In step 295, local program counter 107 is configured to the destination address relevant with the flow process steering order of carrying out now and equates that this flow process steering order of carrying out now is performed by performance element 106, and flow process is got back to step 210.In addition, call out piece if the execution of flow process instruction now enters one, the call state position that then has been eliminated can be set.Yet, if leaving one, the execution of flow process instruction now calls out piece, removing the call state position can be eliminated.
Above-mentioned flow process steering order can be IF flow process steering order, but only in order to for example, is not in order to qualification the present invention, that is it also can be other kind instruction such as LOOP, REP, BREAK, ELSE, ENDIF, ENDLOOP, ENDREP, BREAKC....Therefore, scope of the present invention should comprise following method: use the flow process control structure with local program counter that has disclosed, this structure makes predetermined instruction comprise an IF flow process steering order, and this IF flow process steering order does not need to have (but also can have) corresponding terminal flow process steering order: EndIf flow process steering order.
It is noted that, relatively the count value of global program's counter 120 and the action of the count value in the local program counter 107 in the processing unit 105 are the processing units that has minimum local program count value in order to differentiation, and this minimum local program count value is indicated by the count value of being stored in global program's counter 120.Yet, can also adopt other embodiment.For instance, flow process control module 140 relatively is stored in the count value in the local program counter 107 in the processing unit 105 has minimum local program count value with differentiation processing unit.Then, the comparative result that instruction buffer 110 is produced according to flow process control module 140 is providing the instruction of desiring to be performed, and the performance element 106 that is judged in other processing unit 105 is enabled to execute instruction according to comparative result by flow process control module 140.In brief, the action of flow process control module 140 is equal to the count value of comparison one reference value and local program counter 107 to produce a comparative result.Reference value can be the count value that is stored in global program's counter, the count value in another local program counter 107, or other electronic package according to the value that must designing institute gives.Therefore, instruction buffer 110 provides instruction according to comparative result, and performance element 106 is enabled to execute instruction according to comparative result by the flow process control module.
In sum, the invention provides most processing units, and each processing unit has the local program counter.And only there is performance element to carry out instruction now with minimum local program counter.In the situation of bank cover or the control of non-bank shell type call flow, the processing unit of carrying out bank cover or non-bank shell type call instruction is classified into same a group, and its local program counter is preferentially confirmed whether be minimum value, will carry out instruction now to learn that processing unit.Whether processing unit can be easy to be differentiated in the situation of bank shell type flow process control, and for example, an available mode bit or a particular variables are judged it.The present invention can be at all processing units during with the calculation process steering order, use the destination address relevant with by leaving (early-out in early days with the flow process steering order, that is to say, all be judged when taking same branch operation in all data, can leave bank cover or non-bank shell type flow process controll block ahead of time, rather than rest in bank cover or the non-bank shell type flow process controll block always, grasp the instruction that need not carry out and carry out, again its execution result is shielded at last) to obtain better efficient.
The above only is preferred embodiment of the present invention, and all equalizations of being done according to the present patent application claim change and modify, and all should belong to covering scope of the present invention.
Claims (19)
1. the disposal route of the predetermined instruction in the disposal system, this disposal system has most processing units, and this disposal route comprises:
(a) provide global program's counter, and set the count value of this global program's counter, with instruction as the described predetermined instruction that is performed;
(b) specify a local program counter to give each processing unit, and a count value of this local program counter is set in a performed instruction now according to this processing unit; And
(c) according to the count value that is stored in local program counter and this global program's counter of this processing unit, enable at least one processing unit to carry out the specific instruction in the described predetermined instruction.
2. one kind in order to handle the disposal system of predetermined instruction, and this disposal system comprises:
One instruction buffer is in order to receive and to cushion this predetermined instruction;
One global program's counter is coupled to this instruction buffer, and in order to storing a count value, and this count value is the instruction as this predetermined instruction that is performed;
Most processing units, wherein each comprises:
One performance element is in order to execution command; And
One local program counter is in order to a now instruction setting one count value performed according to this performance element; And
First-class process control unit, be coupled to this global program's counter and each processing unit, in order to according to being stored in this local program counter of this processing unit and the count value of this global program's counter enables at least one processing unit to carry out a specific instruction of described predetermined instruction.
3. disposal system as claimed in claim 2, wherein, during this flow process control module this local program counter in being stored in this particular processor unit has and this global program's counter equates count value, enable a particular processor unit to carry out by global program's counter specific instruction pointed.
4. disposal system as claimed in claim 3, wherein, this specific instruction is other instruction beyond the flow process steering order, and this flow process control module is controlled this specific local program counter and is stored in this specific count value in this specific local program counter with increase.
5. disposal system as claimed in claim 3, wherein, this specific instruction is first-class process control instruction, and this performance element of this particular processor unit is carried out this flow process steering order producing a result of calculation, and this flow process control module updates stored in this specific count value in this specific local program counter according to this result of calculation.
6. disposal system as claimed in claim 5, wherein, when this flow process control module makes that in this result of calculation the branch operation of this flow process control or branch instruction appointment is not carried out, increase this specific count value that is stored in this specific local program counter, and this flow process control module is assigned corresponding to a predicted target address of another instruction and is given this specific count value that is stored in this specific local program counter when this result of calculation makes that the branch operation of this flow process control or branch instruction appointment will be carried out.
7. disposal system as claimed in claim 2, wherein, specific local program counter in one particular processor unit is not enabled when carrying out this specific instruction in this particular processor unit of described processing unit, keeps a specific count value that is stored in described specific local program counter.
8. the disposal route of the predetermined instruction in the disposal system, this disposal system has most processing units, and this method comprises:
(a) relatively be stored in most count values in most local program counters to produce a comparative result, wherein said count value is assigned to this processing unit respectively;
(b) provide an instruction of this predetermined instruction according to this comparative result; And
(c) enable a particular processor unit in the described processing unit to carry out this instruction according to this comparative result.
9. one kind in order to handle the disposal system of predetermined instruction, comprises:
Most local program counters are coupled to most processing units, and wherein, described local program counter is stored most count values respectively;
First-class process control unit is coupled to described local program counter, in order to more described count value;
One instruction buffer receives and cushions an instruction of described predetermined instruction, and a comparative result that is produced according to this flow process control module provides this instruction; And
One performance element is coupled to this instruction buffer and is enabled to carry out this instruction according to this comparative result by this flow process control module.
10. disposal system as claimed in claim 9, wherein, if this performance element is not enabled and can be kept with a specific count value of carrying out this instruction, then be stored in a local program counter that is assigned to this processing unit.
11. disposal system as claimed in claim 9, wherein, if this instruction is other instruction beyond the flow process steering order, this flow process control module more increases a specific count value that is stored in should a specific local program counter of performance element.
12. disposal system as claimed in claim 9, wherein, if this instruction is first-class process control instruction, this performance element is carried out this flow process steering order to produce a result of calculation and this flow process control module and to update stored in the specific count value in should a specific local program counter of performance element according to this result of calculation.
13. disposal system as claimed in claim 12, wherein, if this result of calculation makes the branch operation of this flow process control or branch instruction appointment not carry out, this flow process control module increase is stored in this specific count value in this specific local program counter, and if this result of calculation makes the branch operation of this flow process control or branch instruction appointment to carry out, this flow process control module is assigned corresponding to a predicted target address of another instruction and is given this specific count value that is stored in this specific local program counter.
14. the disposal system in order to the processing predetermined instruction comprises:
One local program counter is coupled to a processing unit, in order to store a count value;
First-class process control unit is coupled to this local program counter, in order to relatively this count value and a reference value;
One instruction buffer receives and cushions an instruction of described predetermined instruction, and provides this instruction according to the comparative result that this comparing unit produced; And
One performance element is coupled to this instruction buffer and is enabled to carry out this instruction according to this comparative result by this flow process control module.
15. disposal system as claimed in claim 14, wherein, this flow process control module enables this performance element to carry out this instruction when this count value is equal with this reference value.
16. disposal system as claimed in claim 14, wherein, if this instruction is other instruction beyond the flow process steering order, this flow process control module more increases this count value that is stored in should this local program counter of performance element.
17. disposal system as claimed in claim 14, wherein, if this instruction is first-class process control instruction, this performance element is carried out this flow process steering order and is updated stored in this specific count value in should this local program counter of performance element according to this result of calculation to produce a result of calculation and this flow process control module.
18. disposal system as claimed in claim 17, wherein, if this result of calculation makes the branch operation of this flow process control or branch instruction appointment not carry out, this flow process control module increase is stored in this count value in this local program counter, and if this result of calculation makes the branch operation of this flow process control or branch instruction appointment to carry out, this flow process control module assigns a predicted target address to give this specific count value that is stored in this specific local program counter, and this predicted target address instructs corresponding to another.
19. the disposal route of the predetermined instruction in the disposal system, this method comprises:
(a) reference value and be stored in a count value in the local program counter that is assigned to a processing unit relatively to produce a comparative result;
(b) provide an instruction of this predetermined instruction according to this comparative result;
(c) enable in this processing unit to carry out this instruction according to this comparative result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007100040776A CN101231584A (en) | 2007-01-23 | 2007-01-23 | Processing system as well as correlation technique for processing instruction thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007100040776A CN101231584A (en) | 2007-01-23 | 2007-01-23 | Processing system as well as correlation technique for processing instruction thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101231584A true CN101231584A (en) | 2008-07-30 |
Family
ID=39898090
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007100040776A Pending CN101231584A (en) | 2007-01-23 | 2007-01-23 | Processing system as well as correlation technique for processing instruction thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101231584A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101819523A (en) * | 2009-03-04 | 2010-09-01 | 威盛电子股份有限公司 | The method that microprocessor and dependent instruction are carried out |
| CN103718158A (en) * | 2011-07-19 | 2014-04-09 | 松下电器产业株式会社 | Multiprocessor system |
-
2007
- 2007-01-23 CN CNA2007100040776A patent/CN101231584A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101819523A (en) * | 2009-03-04 | 2010-09-01 | 威盛电子股份有限公司 | The method that microprocessor and dependent instruction are carried out |
| CN101819523B (en) * | 2009-03-04 | 2014-04-02 | 威盛电子股份有限公司 | Microprocessor and related instruction execution method |
| CN103718158A (en) * | 2011-07-19 | 2014-04-09 | 松下电器产业株式会社 | Multiprocessor system |
| US9317287B2 (en) | 2011-07-19 | 2016-04-19 | Panasonic Intellectual Property Management Co., Ltd. | Multiprocessor system |
| CN103718158B (en) * | 2011-07-19 | 2016-10-19 | 松下知识产权经营株式会社 | Multicomputer system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107341542B (en) | Apparatus and method for performing recurrent neural networks and LSTM operations | |
| CN110689126A (en) | Device for executing neural network operation | |
| US8397234B2 (en) | Determining a priority value for a thread for execution on a multithreading processor system | |
| JPH10171652A (en) | Method and device for updating of branch history table | |
| US9164769B2 (en) | Analyzing data flow graph to detect data for copying from central register file to local register file used in different execution modes in reconfigurable processing array | |
| EP2131278A1 (en) | Scheduling of multiple tasks in a system including multiple computing elements | |
| US20120082167A1 (en) | Method and Apparatus for Predicting Characteristics of Incoming Data Packets to Enable Speculative Processing to Reduce Processor Latency | |
| CN111008040B (en) | Cache device and cache method, computing device and computing method | |
| US20130268700A1 (en) | Prefetching for a shared direct memory access (dma) engine | |
| US6880063B2 (en) | Memory cache bank prediction | |
| CN111753994B (en) | Data processing method and device of AI chip and computer equipment | |
| US8214617B2 (en) | Apparatus and method of avoiding bank conflict in single-port multi-bank memory system | |
| JP2005518578A5 (en) | ||
| US20210311742A1 (en) | An apparatus and method for predicting source operand values and optimized processing of instructions | |
| CN101231584A (en) | Processing system as well as correlation technique for processing instruction thereof | |
| JP2580396B2 (en) | Branch instruction control method in pipeline | |
| JPH096612A (en) | Method and apparatus for processing of multiple branch instruction for execution of write to count register and link register | |
| JP2636088B2 (en) | Information processing device | |
| CN108874548A (en) | Data processing dispatching method, device, computer equipment and data processing system | |
| CN109948785B (en) | High-efficiency neural network circuit system and method | |
| US11609764B2 (en) | Inserting a proxy read instruction in an instruction pipeline in a processor | |
| CN101256547A (en) | Method for controlling nest-shaped process flow and processing system | |
| JP2003525493A (en) | Data processor with multi-command instructions | |
| US8294721B2 (en) | Processor for three dimensional computer graphics | |
| JP2765058B2 (en) | Stream operation completion detection method and apparatus |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20080730 |