CN104142889A - Test point setting and optimization method of single-particle soft errors at assembly level - Google Patents
Test point setting and optimization method of single-particle soft errors at assembly level Download PDFInfo
- Publication number
- CN104142889A CN104142889A CN201410353315.4A CN201410353315A CN104142889A CN 104142889 A CN104142889 A CN 104142889A CN 201410353315 A CN201410353315 A CN 201410353315A CN 104142889 A CN104142889 A CN 104142889A
- Authority
- CN
- China
- Prior art keywords
- node
- fundamental block
- information
- function
- optimization
- 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
Abstract
The invention discloses a test point setting and optimization method of single-particle soft errors at an assembly level. The method comprises the following steps that basic blocks are divided to generate a block table, function return optimization is carried out, function calling optimization is carried out, and loop optimization is carried out. Through the function return optimization method, the function calling optimization method and the loop optimization method, on the premise that it is ensured that the coverage rate of the control flow soft error detection is basically changeless, the number of test points is decreased, the actual detection frequency is reduced, the detection expenses are reduced greatly, and the number of optimized test points is decreased by more than 30% than that of test points not optimized.
Description
Technical field
The present invention relates to a kind of test point setting and optimization method of single-particle soft error of assembly level, belong to system anti-single particle soft error fault-tolerant technique field.
Background technology
Along with the current Space Science and Technology of China and engineering are fast-developing, the basic law of research single particle effect, the measure of finding prevention and slowing down single particle effect have great urgency.
From application point analysis, to carry out state-detection to operating DSP, first analysis list particle effect forms soft error failure pattern and feature thereof, as disorderly in programmed control flow process, reading and writing data is abnormal, functional status mistake etc., the software code that DSP is built carries out the classifying rationally of module, and then determines rational test point between module and module.
DSP is comprised of program/data-carrier store, program/Data Cache, general-purpose register, arithmetic logical unit and peripheral hardware and controller thereof.The storage-type unit such as program/data-carrier store of single particle effect major effect DSP, Cache, register.And according to DSP internal circuit unit to the susceptibility of single particle effect and the difference of type of breaking down, DSP internal module can be divided into storage-type module and function pattern.
The fault that storage-type module SEU causes mainly contains, the bit flipping of program storage area, the bit flipping of program Cache, the bit flipping of data storage area and the bit flipping of register.The fault that functional module SET/SEFI causes mainly contains, the fault that the fault that the fault that the fault that command supervisor SET/SEFI causes, arithmetic element SET/SEFI cause, JTAG/PORSEFI cause, JTAG/PORSEFI cause etc.
In conjunction with dsp program feature, above fault can be divided into following two classes:
(1) single-particle soft error causes the mistake of result of calculation, but does not affect the realization of DSP allomeric function;
(2) single-particle soft error causes the disorder of DSP program execution flow, and the mistake of logic function mistake or results of intermediate calculations etc., cause the execution flow process of program to make a mistake.
The level of target program is different when dividing fundamental block, fundamental block partitioning technology can be divided into two classes: towards the fundamental block partitioning technology of C language with towards the fundamental block partitioning technology of assembly language.
C language is a door high level language, has platform-neutral, and program code is succinctly easy to read.For the fundamental block partitioning technology of C language, be easy to realize, the fundamental block marking off structurally with in function has very strong correspondence, substantially each fundamental block corresponding a specific function realize.But C programmer finally will could move after the stages such as compiling, link in machine.In higher level lanquage, in the transfer process of low-level language, " fundamental block " of C language one-level can not be mapped on low-level language well.On low-level language, be difficult to identify a C language fundamental block, and the fundamental block of a lot of C language one-levels is in low-level language one-level " intersection " together, this makes to divide fundamental block in C language one-level does not have great practical significance.
Assembly language is low-level language, has good corresponding relation between assembly instruction and machine instruction, and assembly instruction is translated and got machine instruction by memonic symbol in essence.This can be mapped on the executable code consisting of machine instruction the fundamental block that assembly language one-level marks well.Like this, object and the demand (as added signature etc.) at program language one-level division fundamental block can embody well in program process.In addition, in compilation one-level, realize fundamental block and divide and can also obtain about basic block structured information, as the address information of key instruction etc., to support the control stream detection scheme based on fundamental block integrality.
But the DSP of take is object, from carrying out fundamental block division based on assembly language angle, the optimization method to set up with certain compatibility of obtaining test point has not yet to see report.
(1) search " soft error detection ", " DSP soft error detection ", " software detection " or " soft error detection module " etc. in database, find following relevant open source information:
A, " Spaceborne signal processing platform single particle effect detect with reinforcement technique research ", proposed the interval monitoring method of redirect of program execution flow.The method is based on higher level lanquage, as C, C++ etc., the basic module of define program code.This patent be towards compiling after assembly code, carry out the division of basic module.
B, " a kind of research and implementation of spaceborne computer software error detecting technology ", proposed the fundamental block division methods for the intermediate code level of ARM7TDMI.The architecture of ARM7TDMI belongs to ARMV4T architecture, different from DSP, the relatively more simple a kind of instruction set of its instruction set.Therefore, institute's extracting method does not have upward compatibility.
(2) search " Soft error detection ", " DSP soft error detection " or " DSP soft error test block " etc. in database, find following relevant open source information:
A, O.Goloubeva, M.Rebaudengo, M.SonzaReorda, et al.Soft-error Detection Using Control Flow Assertions, in Proceedings of the18th IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems (DFT.03), 2003:58-588., for the fault detection problem of computer software, has proposed the Research Thinking of basic module execution flow process from the angle of higher level lanquage.This patent be towards compiling after assembly code, carry out the division of basic module.
B, N.Oh, P.P.Shirvaniand, E.J.MeCluskey.Error Detection by Duplicated Instructions in Super-Scalar Processors, IEEE Transaetions on Reliability, 2002,51 (l): 63-75. are the fault detection problem for computer software equally, have proposed the Research Thinking of basic module execution flow process from the angle of higher level lanquage.This patent be towards compiling after assembly code, carry out the division of basic module.
Summary of the invention
Technology of the present invention is dealt with problems and is: for the deficiencies in the prior art, a kind of test point setting and optimization method of DSP single-particle soft error of assembly level are provided, by adopting function to return to the method for optimization, function call optimization and loop optimization, guaranteeing to control under the substantially constant prerequisite of stream soft error detection coverage rate, reduce the quantity of test point and reduce the actual number of times detecting, significantly having reduced detection overhead.
Technical solution of the present invention is:
A kind of test point setting and optimization method of single-particle soft error of assembly level comprise following steps:
The first step: divide fundamental block and generate piecemeal table
According to the key word of fundamental block, assembly routine is divided into the set of several fundamental blocks and the structural information of each fundamental block is write in piecemeal table; Described fundamental block refers to the program code that one section of order is carried out; Described fundamental block key word refers to jump instruction, function call instruction and the function link order in assembly language; Described piecemeal table is a chained list, for preserving the structural information of fundamental block, comprises fundamental block entrance, fundamental block outlet, fundamental block length, attribute field, next hop address and next node address of piecemeal table;
Second step: carry out function and return to optimization
According to the identification information of each fundamental block, travel through original piecemeal table, find out node corresponding to function link order place fundamental block, in the corresponding node of fundamental block after information in this node " merging " is returned to function call, form new fundamental block and node, formation function returns to piecemeal table and the chained list thereof after optimization;
The 3rd step: carry out function call optimization
According to function, return to the identification information of the fundamental block after optimization, traversal function returns to the piecemeal table after optimization, find out node corresponding to function call instruction place fundamental block, by the information in this node " merging " in the node at first fundamental block place of subfunction, form new fundamental block and node, formation function returns to piecemeal table and the chained list thereof after optimization and function call optimization;
The 4th step: carry out loop optimization
The identification information of the fundamental block after optimizing according to function call, piecemeal table after traversal function call is optimized, find out the node corresponding to loop body place fundamental block of substance for circulation, information in this node " merging ", in the node at fundamental block place after loop body, is formed to new piecemeal table and node.
The specific implementation of information in described second step " merging " is as follows:
(2a) in original piecemeal table, find out the node corresponding to entrance place fundamental block of principal function, from then on node place starts to read piecemeal table node information;
(2b) order reads piecemeal table node and calls according to the information discriminant function of node attribute field, if attribute field is 1, existence function calls, and the address of preserving the next node of current node with a storehouse S forwards (2c) to and carries out; If attribute field is not 1, read the next node information of chained list, repeating step (2b);
(2c) node information of the object subfunction of order read step (2b) function call judging, if the node attribute field reading is 1, show to have in subfunction nested function call, the node information of the nested subfunction of this subfunction is sequentially read in continuation, judge whether whether this nested subfunction also exists nested function call, until each node attribute field of certain nested subfunction is 0, show to find last fundamental block of current subfunction, the fundamental block of current fundamental block and the sensing of heap stack top is merged, form new fundamental block, realize and optimizing, and enter step (2d), if the node attribute field reading is 2, gets back to step (2c) and continue to carry out, if the node attribute field reading is 0, show to find last fundamental block of current subfunction, the fundamental block of current fundamental block and the sensing of heap stack top is merged, form new fundamental block, realize and optimizing, and enter step (2d),
(2d) the entrance information of current node correspondence fundamental block is write in the entrance field of the fundamental block of piling stack top sensing, and two fundamental block length informations are added in the block length field of the fundamental block node that writes the sensing of heap stack top; From chained list, discharge current node, and safeguard the integrality of chained list;
(2e) read the node that stack top is pointed to, then eject stack top information, according to step (2b), continue to carry out last node to chained list.
In described the 3rd step, the specific implementation of information " merging " is as follows:
(3a) in the piecemeal table from function returns to optimization, find out the node corresponding to entrance of principal function, from this node, start to read this piecemeal table node information;
(3b) order function reading is returned and is optimized rear each node, and judgement node attribute field information, if node attribute field is 1, shows to find a function call, forwards (3c) to and carries out; If node attribute field is not 1, re-execute step (3b);
(3c) next-hop address information in the node of the function call place fundamental block finding according to step (3b) is found the node at first fundamental block place of subfunction, and the information of function call place fundamental block is merged in first fundamental block of subfunction.
(3d) the entrance information of the corresponding fundamental block of current node is write in the entrance field of first fundamental block of subfunction, and two fundamental block length informations are added and are write in the block length field of first fundamental block of subfunction; From chained list, discharge function call place node, and safeguard the integrality of chained list;
(3e) return to step (3b) and continue to carry out, until be scanned up to function, return to last node to chained list of carrying out after optimization.
In described the 4th step, the specific implementation of information " merging " is as follows:
(4a) in the piecemeal table from function call optimization, find out the node corresponding to entrance of principal function, from this node, start to read each node information.
(4b) order reads each node of piecemeal table, and judges whether current node next hop address equals fundamental block entry address in current node, if so, shows to find a substance circulation for, starts to carry out loop optimization; Otherwise, re-execute step (4b);
(4c) the entrance information of the corresponding fundamental block of current node is inserted to obtaining in entrance information of the next corresponding fundamental block of node of chained list;
(4d) from chained list, discharge current node, carry out chained list completeness maintaining;
(4e) returning to step (4b) continues until scanned last node of the chained list after all function calls are optimized.
The present invention's beneficial effect is compared with prior art:
(1) the present invention merges fundamental block by return to place, function call place and circulation place in function call, merges test point, at least declines 30% in theory before making test point quantity after optimizing compare optimization.
(2) with respect to the conventional control stream soft error detection scheme based on signature, the present invention is on TI6700 series DSP program example, and the test point quantity after optimization drops to 50% left and right before optimization.
(3) in the present invention, test point arranges optimization method and has versatility, can promote the use of in the platform of other frameworks, as ARM, X86 etc.
Accompanying drawing explanation
Fig. 1 is the inventive method process flow diagram;
Fig. 2 is piecemeal table node structure figure of the present invention;
Fig. 3 is piecemeal list structure figure of the present invention;
Fig. 4 is not for applying the original assembly code of FFT of the present invention;
Fig. 5 is that the present invention is applied in the fundamental block division result in FFT;
Fig. 6 is that the present invention is applied in the test point optimum results in FFT.
Embodiment
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is further described in detail.
As shown in Figure 1, test point setting and the optimization method of the DSP single-particle soft error of a kind of assembly level of the present invention, comprise step as follows:
The first step: divide fundamental block and generate piecemeal table
According to the key word of fundamental block, assembly routine is divided into the set of several fundamental blocks and the structural information of each fundamental block is write in piecemeal table; Described fundamental block refers to the program code that one section of order is carried out; Described fundamental block key word refers to jump instruction, function call instruction and the function link order in assembly language; Described piecemeal table is a chained list, for preserving the structural information of fundamental block, comprises fundamental block entrance, fundamental block outlet, fundamental block length, attribute field, next hop address and next node address of piecemeal table.
According to the key word of fundamental block, assembly routine is divided into several fundamental blocks:
(a) skip instruction, function call and function call are returned to the outlet as a basic module;
(b) a rear entrance as basic module of current redirect (call, call and return);
(c) the object statement of current redirect (call, call and return) is as the entrance of basic module;
(d) the last sentence of the object statement of current redirect (call, call and return) is as the outlet of basic module.
As shown in Figure 2,3, piecemeal table is a chained list, and wherein each node is preserved the structural information of a fundamental block; Node is the basic structure of chained list, and chained list is exactly that several nodes get up to form by " next node address " link.Node mainly comprises fundamental block entrance, fundamental block outlet, fundamental block length, attribute field, next hop address and next node address of piecemeal table.Fundamental block entrance, fundamental block outlet refer to respectively basic module first statement and the logical address of basic module the last item statement in assembly routine; Fundamental block length refers to the assembly instruction number in fundamental block; Attribute field only stores 0,1 and 2 three kind of state, if fundamental block is to using a skip instruction as outlet, attribute field puts 2; If a fundamental block be take function call instruction as outlet, attribute field puts 1; In other situation, attribute field sets to 0.Attribute field is the important evidence of carrying out test point optimization.Next hop address refers to the destination address of jump instruction.Next node address of piecemeal table is for the maintenance of chained list.
Second step: carry out function and return to optimization
According to the identification information of each fundamental block, travel through original piecemeal table, find out node corresponding to function link order place fundamental block, in the corresponding node of fundamental block after information in this node " merging " is returned to function call, form new fundamental block and node, formation function returns to piecemeal table and the chained list thereof after optimization;
The specific implementation of information " merging " is as follows:
The specific implementation of information in second step " merging " is as follows:
(2a) in original piecemeal table, find out the node corresponding to entrance place fundamental block of principal function, from then on node place starts to read piecemeal table node information;
(2b) order reads piecemeal table node and calls according to the information discriminant function of node attribute field, if attribute field is 1, existence function calls, and the address of preserving the next node of current node with a storehouse S forwards (2c) to and carries out; If attribute field is not 1, read the next node information of chained list, repeating step (2b);
(2c) node information of the object subfunction of order read step (2b) function call judging, if the node attribute field reading is 1, show to have in subfunction nested function call, the node information of the nested subfunction of this subfunction is sequentially read in continuation, judge whether whether this nested subfunction also exists nested function call, until each node attribute field of certain nested subfunction is 0, show to find last fundamental block of current subfunction, the fundamental block of current fundamental block and the sensing of heap stack top is merged, form new fundamental block, realize and optimizing, and enter step (2d), if the node attribute field reading is 2, gets back to step (2c) and continue to carry out, if the node attribute field reading is 0, show to find last fundamental block of current subfunction, the fundamental block of current fundamental block and the sensing of heap stack top is merged, form new fundamental block, realize and optimizing, and enter step (2d),
(2d) the entrance information of current node correspondence fundamental block is write in the entrance field of the fundamental block of piling stack top sensing, and two fundamental block length informations are added in the block length field of the fundamental block node that writes the sensing of heap stack top; From chained list, discharge current node, and safeguard the integrality of chained list;
(2e) read the node that stack top is pointed to, then eject stack top information, according to step (2b), continue to carry out last node to chained list.
The 3rd step: carry out function call optimization
According to function, return to the identification information of the fundamental block after optimization, traversal function returns to the piecemeal table after optimization, find out node corresponding to function call instruction place fundamental block, by the information in this node " merging " in the node at first fundamental block place of subfunction, form new fundamental block and node, formation function returns to piecemeal table and the chained list thereof after optimization and function call optimization;
The specific implementation of information " merging " is as follows:
(3a) in the piecemeal table from function returns to optimization, find out the node corresponding to entrance of principal function, from this node, start to read this piecemeal table node information;
(3b) order function reading is returned and is optimized rear each node, and judgement node attribute field information, if node attribute field is 1, shows to find a function call, forwards (3c) to and carries out; If node attribute field is not 1, re-execute step (3b);
(3c) next-hop address information in the node of the function call place fundamental block finding according to step (3b) is found the node at first fundamental block place of subfunction, and the information of function call place fundamental block is merged in first fundamental block of subfunction.
(3d) the entrance information of the corresponding fundamental block of current node is write in the entrance field of first fundamental block of subfunction, and two fundamental block length informations are added and are write in the block length field of first fundamental block of subfunction; From chained list, discharge function call place node, and safeguard the integrality of chained list;
(3e) return to step (3b) and continue to carry out, until be scanned up to function, return to last node to chained list of carrying out after optimization.
The 4th step: carry out loop optimization
The identification information of the fundamental block after optimizing according to function call, piecemeal table after traversal function call is optimized, find out the node corresponding to loop body place fundamental block of substance for circulation, information in this node " merging ", in the node at fundamental block place after loop body, is formed to new piecemeal table and node;
The specific implementation of information " merging " is as follows:
(4a) in the piecemeal table from function call optimization, find out the node corresponding to entrance of principal function, from this node, start to read each node information.
(4b) order reads each node of piecemeal table, and judges whether current node next hop address equals fundamental block entry address in current node, if so, shows to find a substance circulation for, starts to carry out loop optimization; Otherwise, re-execute step (4b);
(4c) the entrance information of the corresponding fundamental block of current node is inserted to obtaining in entrance information of the next corresponding fundamental block of node of chained list;
(4d) from chained list, discharge current node, carry out chained list completeness maintaining;
(4e) returning to step (4b) continues until scanned last node of the chained list after all function calls are optimized.
As shown in Figure 4,5, 6, the present invention is applied in the program code of FFT conversion, the original assembly code of FFT has 2,000 row left and right, after simplification, be 650 row, according to fundamental block division principle, can be divided into 31 fundamental blocks, through returning to optimization, calling and optimize and loop optimization, be reduced to 16 fundamental blocks, obtain 16 test points.
As can be seen here, adopt the test point after prioritization scheme obviously to reduce, therefore, detection overhead also can produce significant decline.
The content not being described in detail in instructions of the present invention belongs to those skilled in the art's known technology.
Claims (4)
1. test point setting and the optimization method of the single-particle soft error of assembly level, is characterized in that comprising following steps:
The first step: divide fundamental block and generate piecemeal table
According to the key word of fundamental block, assembly routine is divided into the set of several fundamental blocks and the structural information of each fundamental block is write in piecemeal table; Described fundamental block refers to the program code that one section of order is carried out; Described fundamental block key word refers to jump instruction, function call instruction and the function link order in assembly language; Described piecemeal table is a chained list, for preserving the structural information of fundamental block, comprises fundamental block entrance, fundamental block outlet, fundamental block length, attribute field, next hop address and next node address of piecemeal table;
Second step: carry out function and return to optimization
According to the identification information of each fundamental block, travel through original piecemeal table, find out node corresponding to function link order place fundamental block, in the corresponding node of fundamental block after information in this node " merging " is returned to function call, form new fundamental block and node, formation function returns to piecemeal table and the chained list thereof after optimization;
The 3rd step: carry out function call optimization
According to function, return to the identification information of the fundamental block after optimization, traversal function returns to the piecemeal table after optimization, find out node corresponding to function call instruction place fundamental block, by the information in this node " merging " in the node at first fundamental block place of subfunction, form new fundamental block and node, formation function returns to piecemeal table and the chained list thereof after optimization and function call optimization;
The 4th step: carry out loop optimization
The identification information of the fundamental block after optimizing according to function call, piecemeal table after traversal function call is optimized, find out the node corresponding to loop body place fundamental block of substance for circulation, information in this node " merging ", in the node at fundamental block place after loop body, is formed to new piecemeal table and node.
2. test point setting and the optimization method of the single-particle soft error of a kind of assembly level according to claim 1, is characterized in that: the specific implementation of information in described second step " merging " is as follows:
(2a) in original piecemeal table, find out the node corresponding to entrance place fundamental block of principal function, from then on node place starts to read piecemeal table node information;
(2b) order reads piecemeal table node and calls according to the information discriminant function of node attribute field, if attribute field is 1, existence function calls, and the address of preserving the next node of current node with a storehouse S forwards (2c) to and carries out; If attribute field is not 1, read the next node information of chained list, repeating step (2b);
(2c) node information of the object subfunction of order read step (2b) function call judging, if the node attribute field reading is 1, show to have in subfunction nested function call, the node information of the nested subfunction of this subfunction is sequentially read in continuation, judge whether whether this nested subfunction also exists nested function call, until each node attribute field of certain nested subfunction is 0, show to find last fundamental block of current subfunction, the fundamental block of current fundamental block and the sensing of heap stack top is merged, form new fundamental block, realize and optimizing, and enter step (2d), if the node attribute field reading is 2, gets back to step (2c) and continue to carry out, if the node attribute field reading is 0, show to find last fundamental block of current subfunction, the fundamental block of current fundamental block and the sensing of heap stack top is merged, form new fundamental block, realize and optimizing, and enter step (2d),
(2d) the entrance information of current node correspondence fundamental block is write in the entrance field of the fundamental block of piling stack top sensing, and two fundamental block length informations are added in the block length field of the fundamental block node that writes the sensing of heap stack top; From chained list, discharge current node, and safeguard the integrality of chained list;
(2e) read the node that stack top is pointed to, then eject stack top information, according to step (2b), continue to carry out last node to chained list.
3. test point setting and the optimization method of the single-particle soft error of a kind of assembly level according to claim 1, is characterized in that: in described the 3rd step, the specific implementation of information " merging " is as follows:
(3a) in the piecemeal table from function returns to optimization, find out the node corresponding to entrance of principal function, from this node, start to read this piecemeal table node information;
(3b) order function reading is returned and is optimized rear each node, and judgement node attribute field information, if node attribute field is 1, shows to find a function call, forwards (3c) to and carries out; If node attribute field is not 1, re-execute step (3b);
(3c) next-hop address information in the node of the function call place fundamental block finding according to step (3b) is found the node at first fundamental block place of subfunction, and the information of function call place fundamental block is merged in first fundamental block of subfunction.
(3d) the entrance information of the corresponding fundamental block of current node is write in the entrance field of first fundamental block of subfunction, and two fundamental block length informations are added and are write in the block length field of first fundamental block of subfunction; From chained list, discharge function call place node, and safeguard the integrality of chained list;
(3e) return to step (3b) and continue to carry out, until be scanned up to function, return to last node to chained list of carrying out after optimization.
4. test point setting and the optimization method of the single-particle soft error of a kind of assembly level according to claim 1, is characterized in that: in described the 4th step, the specific implementation of information " merging " is as follows:
(4a) in the piecemeal table from function call optimization, find out the node corresponding to entrance of principal function, from this node, start to read each node information.
(4b) order reads each node of piecemeal table, and judges whether current node next hop address equals fundamental block entry address in current node, if so, shows to find a substance circulation for, starts to carry out loop optimization; Otherwise, re-execute step (4b);
(4c) the entrance information of the corresponding fundamental block of current node is inserted to obtaining in entrance information of the next corresponding fundamental block of node of chained list;
(4d) from chained list, discharge current node, carry out chained list completeness maintaining;
(4e) returning to step (4b) continues until scanned last node of the chained list after all function calls are optimized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410353315.4A CN104142889B (en) | 2014-07-23 | 2014-07-23 | A kind of test point of the single-particle soft error of assembly level is arranged and optimization method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410353315.4A CN104142889B (en) | 2014-07-23 | 2014-07-23 | A kind of test point of the single-particle soft error of assembly level is arranged and optimization method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104142889A true CN104142889A (en) | 2014-11-12 |
| CN104142889B CN104142889B (en) | 2017-03-15 |
Family
ID=51852069
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410353315.4A Active CN104142889B (en) | 2014-07-23 | 2014-07-23 | A kind of test point of the single-particle soft error of assembly level is arranged and optimization method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104142889B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106652416A (en) * | 2017-01-10 | 2017-05-10 | 国电南瑞三能电力仪表(南京)有限公司 | Debugging output method suitable for centralized meter reading terminal |
| CN109344612A (en) * | 2018-09-25 | 2019-02-15 | 郑州昂视信息科技有限公司 | The active defense method and system inversely attacked for program code static analysis |
| CN109614274A (en) * | 2018-12-06 | 2019-04-12 | 上海航天计算机技术研究所 | The means of defence of processor instruction Cache single-particle inversion soft error |
| CN111077878A (en) * | 2019-12-12 | 2020-04-28 | 北京新能源汽车技术创新中心有限公司 | Method and system for simplifying test judgment table of automobile controller |
| CN112541188A (en) * | 2019-09-20 | 2021-03-23 | 武汉斗鱼网络科技有限公司 | Method and device for preventing application program code from being statically analyzed |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6668341B1 (en) * | 1999-11-13 | 2003-12-23 | International Business Machines Corporation | Storage cell with integrated soft error detection and correction |
| CN101763291A (en) * | 2009-12-30 | 2010-06-30 | 中国人民解放军国防科学技术大学 | Method for detecting error of program control flow |
| CN101944064A (en) * | 2010-10-12 | 2011-01-12 | 中国人民解放军国防科学技术大学 | Control flow error detection optimizing method based on reconstructed control flow graph |
-
2014
- 2014-07-23 CN CN201410353315.4A patent/CN104142889B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6668341B1 (en) * | 1999-11-13 | 2003-12-23 | International Business Machines Corporation | Storage cell with integrated soft error detection and correction |
| CN101763291A (en) * | 2009-12-30 | 2010-06-30 | 中国人民解放军国防科学技术大学 | Method for detecting error of program control flow |
| CN101944064A (en) * | 2010-10-12 | 2011-01-12 | 中国人民解放军国防科学技术大学 | Control flow error detection optimizing method based on reconstructed control flow graph |
Non-Patent Citations (3)
| Title |
|---|
| O.GOLOUBEVA: "Soft-error Detection Using Control Flow Assertions", 《PROCEEDINGS OF THE 18TH IEEE INTERNATIONAL SYMPOSIUM ON DEFECT AND FAULT TOLERANCE IN VLSI SYSTEMS》 * |
| 吴艳霞: "一种基于控制流检测的低功耗基本块划分方法", 《计算机工程与应用》 * |
| 康晓军等: "基于扩展块的星载软件控制流容错评价方法", 《航天返回与遥感》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106652416A (en) * | 2017-01-10 | 2017-05-10 | 国电南瑞三能电力仪表(南京)有限公司 | Debugging output method suitable for centralized meter reading terminal |
| CN109344612A (en) * | 2018-09-25 | 2019-02-15 | 郑州昂视信息科技有限公司 | The active defense method and system inversely attacked for program code static analysis |
| CN109614274A (en) * | 2018-12-06 | 2019-04-12 | 上海航天计算机技术研究所 | The means of defence of processor instruction Cache single-particle inversion soft error |
| CN109614274B (en) * | 2018-12-06 | 2022-05-27 | 上海航天计算机技术研究所 | Protection method for single event upset soft error of processor instruction Cache |
| CN112541188A (en) * | 2019-09-20 | 2021-03-23 | 武汉斗鱼网络科技有限公司 | Method and device for preventing application program code from being statically analyzed |
| CN112541188B (en) * | 2019-09-20 | 2022-05-13 | 武汉斗鱼网络科技有限公司 | Method and device for preventing application program code from being statically analyzed |
| CN111077878A (en) * | 2019-12-12 | 2020-04-28 | 北京新能源汽车技术创新中心有限公司 | Method and system for simplifying test judgment table of automobile controller |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104142889B (en) | 2017-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104142889A (en) | Test point setting and optimization method of single-particle soft errors at assembly level | |
| CN104077144B (en) | Data race detection and evidence generation method based on multithreaded program constraint building | |
| US7836435B2 (en) | Checking for memory access collisions in a multi-processor architecture | |
| US5845064A (en) | Method for testing and verification of a CPU using a reference model | |
| CN103218272A (en) | Spaceborne digital signal processor turning reinforcing method | |
| CN103440457B (en) | Based on the binary program analytic system of process simulation | |
| CN103488563A (en) | Data race detection method and device for parallel programs and multi-core processing system | |
| CN103365776A (en) | Parallel system weak consistency verifying method and system based on deterministic replay | |
| CN104461871A (en) | Deadlock detection method based on petri net | |
| CN106022107A (en) | Method and system for protecting program execution integrity | |
| US20150161030A1 (en) | Detecting method and system for concurrency bugs | |
| CN102270166A (en) | Simulator and method for injecting and tracking processor faults based on simulator | |
| CN106354575A (en) | Troubleshooting device and method based on stack tracing | |
| CN105320579A (en) | Self-repairing dual-redundancy assembly line oriented to SPARC V8 processor and fault-tolerant method | |
| Xu et al. | Melton: a practical and precise memory leak detection tool for C programs | |
| CN107003897A (en) | Monitor the utilization rate of issued transaction resource | |
| CN106802861A (en) | A kind of method and apparatus for detecting RAM leakage | |
| Wang et al. | More anti-chain based refinement checking | |
| CN108614772A (en) | The framework and method of a kind of concurrently detection stacker buffer spilling | |
| CN109885489B (en) | Data race detection method and device in driver | |
| CN104951276B (en) | A kind of detection method and system of the failure of chip instruction cache | |
| CN105260256B (en) | A kind of fault detect of duplication redundancy streamline and backing method | |
| US20140089769A1 (en) | Content addressable memory continuous error detection with interleave parity | |
| CN102708054A (en) | Detection method for security flaws in loop write-only memory of binary program | |
| CN105630479A (en) | Processing method and apparatus for exception in program running process |
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 |