CN102567130A - Multi-sampling judgment method for fault-tolerant filtering based on remainder check - Google Patents

Abstract

The present application provides a multi-sampling judgment method in the error-tolerant filter processing based on the remainder check, so as to solve the problem of missed fault detection in the error-tolerant filter processing method based on the remainder check in the prior art. In this application, when the filter output of the first branch and the filter output of the second branch are not equal, but the data after the remainder of the filter output of the two branches are equal, and the faulty branch cannot be judged by a single sampling, the data can be temporarily Save in the cache, continue to process other data, and pass multiple judgments of the filtering results. It is a multi-sampling judgment method until one of the data after the remainder of the filter output of the first branch and the second branch is not equal to the reference data. At this time, it can be judged which branch has a fault and which branch The processing result of the branch is correct, and there will be no problem of missed fault detection.

Description

A multi-sampling decision method in error-tolerant filtering processing based on remainder check

technical field

The present application relates to error-tolerant signal processing technology, in particular to a multi-sampling decision method in error-tolerant filtering processing based on remainder checking.

Background technique

Radiation will affect the performance of the device, which may cause errors in the output data of the device. For example, the signal processing equipment carried on some satellites is often affected by space radiation, resulting in signal processing errors. In order to solve this problem, fault-tolerant signal processing technology has emerged.

Triple-mode redundancy (TMR) technology is widely used in fault-tolerant signal processing in radiation environments. TMR uses three identical modules to perform the same filtering operation on the input data, and determines the final output through the majority selector at the output end, completely eliminating the influence of a single branch failure on the signal processing output. However, in the triple-mode redundancy technology, the consumption of all computing and storage resources is tripled, which makes the triple-mode redundancy technology unusable in many resource-constrained applications, such as fault tolerance on the on-board computing platform signal processing, etc.

For this reason, a fault-tolerant signal processing technology based on remainder checking is produced, which further reduces resource requirements in the fault-tolerant signal processing technology. Existing error-tolerant signal processing technologies based on remainder checking usually use a single-sampling decision method, that is, one branch in the triple-mode redundancy becomes a remainder filtering branch, and when the processing results of the other two branches are inconsistent, the two After taking the remainder of the processing result of the branch, compare it with the result of the residue filtering branch, and select the correct processing result after one judgment, and judge the branch that has a fault.

But this method has a defect, that is, when the wrong output of the faulty branch is exactly congruent with the correct output, the remainder filtering branch cannot judge which branch is faulty, that is, the problem of missed fault detection occurs.

Contents of the invention

The present application provides a multi-sampling judgment method in the error-tolerant filter processing based on the remainder check, so as to solve the problem of missed fault detection in the error-tolerant filter processing method based on the remainder check in the prior art.

In order to solve the above problems, the present application discloses a multi-sampling decision method in error-tolerant filtering processing based on remainder check, including:

S11, respectively inputting the sampling data into the first branch, the second branch and the third branch;

S12, the processing steps of the first branch and the second branch include:

performing filtering processing on the sampling data respectively to obtain processed first filtering data and second filtering data;

Taking the remainder of the first filtering data and the second filtering data respectively to m to obtain the corresponding first data and the second data, wherein m is the modulus of taking the remainder, and m is a positive integer;

S13, for the third branch, input the sampling data into a remainder-based filter for filtering processing to obtain processed reference data, where the modulus of the remainder in the remainder-based filter is m;

S14. If the first filtered data and the second filtered data are not equal, respectively compare the first data and the second data with the reference data;

The results of the comparison include:

If the first data is equal to the benchmark data and the second data is not equal to the benchmark data, execute S15;

If the second data is equal to the benchmark data and the first data is not equal to the benchmark data, execute S16;

If the first data is equal to the second data and is equal to the reference data, execute S17;

S15, outputting the first filtered data of the first branch as result data;

S16, outputting the second filtered data of the second branch as result data;

S17, store the first filtering data and the second filtering data in the first cache and the second cache respectively, and continue to input the sampling data and execute S11 to S14 until one of the first data and the second data is not equal to the reference data .

Preferably, if the first filtered data of the first branch is output as result data, the method further includes:

output the first filtering data in the first buffer, and clear the second filtering data in the second buffer.

Preferably, if the second filtered data of the second branch is output as result data, the method further includes:

Outputting the second filtering data in the second buffer, and clearing the first filtering data in the first buffer.

Preferably, the filters include FIR filters and IIR filters.

Preferably, for the filter based on the remainder in the third branch, the operands that participate in the multiplication operation each time take a remainder to m, the operands after the remainder are multiplied, and the multiplied result is taken to m Get the corresponding modular multiplication result.

Preferably, for the filter based on the remainder in the third branch, the operands participating in the addition operation each time take a remainder to m, the operands after the addition are added, and the result after the addition is taken to m Get the corresponding modular addition result.

Preferably, if the first filtering data is equal to the second filtering data, the processing of the first branch and the second branch are both correct, and the processing result of one branch is selected as output.

Preferably, if the first data, the second data and the reference data are not equal to each other, both the first branch and the second branch are faulty.

Compared with the prior art, the present application includes the following advantages:

First of all, in this application, when the filter output of the first branch and the filter output of the second branch are not equal, but the data after the remainder of the two branches are equal, and the faulty branch cannot be judged by a single sampling, the data can be temporarily Save in the cache, continue to process other data, and pass multiple judgments of the filtering results. It is a multi-sampling judgment method until one of the data after the remainder of the filter output of the first branch and the second branch is not equal to the reference data. At this time, it can be judged which branch has a fault and which branch The processing result of the branch is correct, and there will be no problem of missed fault detection.

Secondly, when the application cannot judge the faulty branch temporarily, the data can be stored in the cache, and the data processing process can be continued until the faulty branch is judged, and then the data in the corresponding cache of the correct branch is output, and the The faulty branch corresponds to clearing the data in the cache, and the whole process does not affect the overall data processing speed of the system.

Thirdly, the probability of missed detection and the resource consumption of fault-tolerant technology in the prior art are contradictory to each other: the smaller the modulus, the simpler the branch operation and the smaller the resource consumption, but the greater the missed detection rate of faults at this time ; Conversely, the larger the modulus, the smaller the failure detection rate, but the greater the resource consumption. However, in the method described in this application, the multi-sampling judgment method adopted can not only maintain the low resource consumption caused by the small modulus, but also greatly reduce the failure detection rate.

Description of drawings

Fig. 1 is the structural diagram of the fault-tolerant filter processing circuit based on remainder check described in the application;

FIG. 2 is a flow chart of a multi-sampling decision method in a remainder-check-based error-tolerant filtering process according to an embodiment of the present application;

Fig. 3 is a flow chart of a multi-sampling decision method in a remainder-check-based error-tolerant filtering process according to a preferred embodiment of the present application.

Detailed ways

In order to make the above objects, features and advantages of the present application more obvious and comprehensible, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

Referring to FIG. 1 , it shows the structural diagram of the error-tolerant filtering processing circuit based on remainder check in this application.

The error-tolerant filter processing circuit based on the remainder check includes an input sampling module, a first branch, a second branch, a third branch and a comparison and judgment module. Wherein, the first branch includes a filter 1 and a remainder sub-module 1, the second branch includes a filter 2 and a remainder sub-module 2, and the third branch includes a remainder-based filter.

The input sampling module inputs the sampled data into the first branch, the second branch and the third branch respectively, and the three branches respectively process the sampled data, when the filtered data processed by the first branch and the second branch When they are equal, the data processed by the two branches are correct, and there is no faulty branch. When the filtered data processed by the first branch and the second branch are not equal, there is a faulty branch, and the data processed by the first branch and the second branch need to be input into the comparison judgment module and the third branch. The processed data is compared to determine the correct processing result.

When making a specific judgment, in the prior art, the sampled data is filtered through the first branch and the second branch to take the remainder, and the sampled data is passed through the filter based on the remainder in the third branch (i.e., the remainder filtering branch). For filter processing, the processing results of the three branches are input into the comparison and judgment module, and the data after the remainder of the first branch and the second branch are compared with the data of the third branch, and the correct one is selected after one judgment. Process the results and judge the branch where the fault occurred. However, if the data after the remainder of the first branch and the second branch are equal, it is impossible to judge which branch has a fault at this time, and which branch has the correct processing result, and the problem of missed fault detection has occurred.

This application provides a multi-sampling decision method in error-tolerant filtering processing based on remainder checking. When the first branch filter output and the second branch filter output are not equal, but the two branch filter outputs take the remainder data When they are equal, the data can be temporarily stored in the cache and continue to process other data, after multiple judgments of the filtering results. It is a multi-sampling judgment method until one of the data after the remainder of the filter output of the first branch and the second branch is not equal to the reference data. At this time, it can be judged which branch has a fault and which branch The processing result of the branch is correct, and there will be no problem of missed fault detection.

Referring to FIG. 2 , it shows a flow chart of a multi-sampling decision method in the error-tolerant filtering process based on remainder checking according to the embodiment of the present application.

S11, respectively inputting the sampling data into the first branch, the second branch and the third branch;

S12, the processing steps of the first branch and the second branch include:

performing filtering processing on the sampling data respectively to obtain processed first filtering data and second filtering data;

Taking the remainder of the first filtering data and the second filtering data respectively to m to obtain the corresponding first data and the second data, wherein m is the modulus of taking the remainder, and m is a positive integer;

Assume that the sampling data is x, the first filtered data is y ₁ , the second filtered data is y ₂ , the first data r ₁ , and the second data r ₂ . Then the filtering result of the first branch is the first filtering data y ₁ , and the filtering result of the second branch is the second filtering data y ₂ .

The remainder of the first filtered data to m is (y ₁ ) _m = r ₁

The remainder of the second filtering data to m is (y ₂ ) _m = r ₂

Wherein, m is the modulus of the modulus, and the modulus of m means dividing the data by m to obtain the remainder, for example, the modulus of 16 to 5, that is, (16) ₅ =1.

S13, for the third branch, input the sampling data into a remainder-based filter for filtering processing to obtain processed reference data, where the modulus of the remainder in the remainder-based filter is m;

Suppose the benchmark data is r.

For the filter based on the remainder in the third branch, the operands participating in the multiplication operation are multiplied by m each time, and the operands after the multiplication are multiplied, and the multiplied result is obtained by taking the remainder of m to obtain the corresponding modulus Multiply the result.

Assume that the operands participating in the multiplication operation are p and q respectively.

Then the corresponding modular multiplication result is ((p) _m ×(q) _m ) _m .

For the filter based on the remainder in the third branch, the operands participating in the addition operation are modulused to m each time, the operands after the modulus are added, and the result after the addition is modulused to m to obtain the corresponding modulus Add results.

Assume that the operands participating in the addition operation are a and b respectively.

Then the corresponding modular addition result is ((a) _m +(b) _m ) _m .

The first filtered data y ₁ is compared with the second filtered data y ₂ to determine whether the first filtered data y ₁ is equal to the second filtered data y ₂ .

If the first filtered data and the second filtered data are not equal, that is, y ₁ ≠y ₂ , go to S14.

If the first filtering data is equal to the second filtering data, that is, y ₁ =y ₂ , and the data of both branches are correct, then the processing result of one branch is selected from the first branch and the second branch as output.

S14, respectively comparing the first data and the second data with the reference data;

The results of the comparison include:

If the first data is equal to the reference data, and the second data is not equal to the reference data, that is, r ₁ =r and r ₂ ≠r, it means that the processing result of the first branch is correct, and the second branch fails, and execute S15;

If the second data is equal to the reference data and the first data is not equal to the reference data, that is, r ₂ =r and r ₁ ≠r, it means that the processing result of the second branch is correct, and the first branch has a fault, and execute S16;

If the first data is equal to the second data is equal to the reference data, that is, r ₁ =r ₂ =r, execute S17;

If the first data, the second data and the reference data are not equal to each other, that is r ₁ ≠r ₂ ≠r, both the first branch and the second branch are faulty.

S15, outputting the first filtered data of the first branch as result data;

S16, outputting the second filtered data of the second branch as result data;

S17, store the first filter data and the second filter data in the first cache and the second cache respectively, continue to input other sampling data and execute S11 to S14 until one of the first data and the second data is not equal to the reference data until.

After step S17, if it is judged that the processing result of the first branch is correct and the second branch fails, the first filtered data of the first branch is output as the result data, and the first filtered data in the first cache is output, And clear the second filtering data in the second cache.

After step S17, if it is judged that the processing result of the second branch is correct and the first branch fails, the second filtered data of the second branch is output as the result data, and the second filtered data in the second cache is output, And empty the first filtering data in the first cache.

In summary, in this application, when the filter output of the first branch and the filter output of the second branch are not equal, but the data after the remainder of the filter output of the two branches are equal, the data can be temporarily saved in the cache and continue processing Other data, after multiple judgments of the filtering results. It is a multi-sampling judgment method until one of the data after the remainder of the filter output of the first branch and the second branch is not equal to the reference data. At this time, it can be judged which branch has a fault and which branch The processing result of the branch is correct, and there will be no problem of missed fault detection.

Referring to FIG. 3 , it shows a flow chart of a multi-sampling decision method in a residual-check-based error-tolerant filtering process according to a preferred embodiment of the present application.

Wherein, y ₁ is the first filtered data, and y ₁ [i] is the first filtered data at time i. y ₂ is the second filtered data, and y ₂ [i] is the second filtered data at time i. r ₁ is the first data, then r ₁ [i] is the first data at time i. r ₂ is the first data, then r ₂ [i] is the second data at time i. r is the reference data, then r[i] is the reference data at time i.

First, input the data y ₁ [i], y ₂ [i], r ₁ [i], r ₂ [i] and r[i] for comparison and judgment, if the judgment y ₁ [i]=y ₂ [i] , then choose a branch as the result output, if it is judged that y ₁ [i]≠y ₂ [i], then proceed to the next step of judgment.

If it is judged that r ₁ [i]≠r[i], and r ₂ [i]=r[i], then output the second branch as the result.

If it is judged that r ₁ [i]≠r[i], and r ₂ [i]≠r[i], errors occur in both branches.

If it is judged that r ₁ [i]=r[i], and r ₂ [i]≠r[i], then output the first branch as the result.

If it is judged that r ₁ [i]=r[i], and r ₂ [i]=r[i], then store y ₁ [i] and y ₂ [i] in the first cache b ₁ and the second cache respectively _b2 .

At this time, i=i+1, N++, that is, the data at the next moment (i+1 moment) is input, and the above judgment process is continued until it is judged that one of r ₁ [i] and r ₂ [i] is the same as r [i] Until they are not equal, output the result of the correct branch, and complete the judgment of N+1 outputs.

The method described in this application is discussed below in a manner of specific implementation.

For example, assume that the remainder modulus m=7, and the filter is a 16-order FIR filter (filter coefficients are h _l , l=0, 2, . . . , 15).

Step (1): Input the sampling data x _i (i=0, 1, ..., ∞) into the first branch, the second branch and the third branch respectively;

Wherein, the first branch and the second branch include a 16-order FIR filter, and the third branch includes a 16-order FIR filter based on modulo 7 processing.

而第二支路的16阶FIR滤波器由于发生故障，滤波系数从h_l，l＝0，2，...，15变成了

l＝0，2，...，15，从而导致n时刻的输出变为 $y_2\left[n\right]=\underset{i=n}{\overset{n+15}{\mathrm{\Σ}}}{x}_i{\tilde{h}}_{15-i+n}=22.$ Step (2): The 16-order FIR filter of the first branch works normally, and the output at time n is

However, the 16-order FIR filter of the second branch has a fault, and the filter coefficient changes from h _l , l=0, 2, ..., 15 to

l=0, 2, ..., 15, resulting in the output at time n becoming ${\mathrm{the\; y}}_2\left[\mathrm{no}\right]=\underset{i=\mathrm{no}}{\overset{\mathrm{no}+15}{\mathrm{\Σ}}}{x}_i{\tilde{h}}_{15-i+\mathrm{no}}=\mathrm{twenty\; two}.$

The two filtered data after filtering are respectively processed by modulo 7 to obtain r ₁ [n]=(y ₁ [n]) ₇ =1 and r ₂ [n]=(y ₂ [n]) ₇ =1.

Step (3): the output of the 16-order FIR filter based on the modulus 7 in the third branch at time n is $r\left[\mathrm{no}\right]=\underset{i=\mathrm{no}}{\overset{\mathrm{no}+15}{\mathrm{\Σ}}}{\left({\left(x_i\right)}_7{\left(h_{15-i+\mathrm{no}}\right)}_7\right)}_7=1;$

At this time, y ₁ [n]≠y ₂ [n], it means that a branch has a fault, and step (4) is performed.

Step (4): Compare r ₁ [n], r ₂ [n] and r[n].

At this time, r ₁ [n]=r ₂ [n]=r[n]=1, it is impossible to tell which branch has failed, so temporarily set y ₁ [n]=15 and y ₂ [n]=22 respectively Cache into two caches: _b1 [0]=15, _b2 [0]=22;

Step (5): Continue to execute the process from step (1) to step (4).

第二支路的16阶FIR滤波器仍旧由于错误的滤波系数产生错误的输出

将两个滤波处理后的滤波数据分别经过模7处理，得到r₁[n+1]＝(y₁[n+1])₇＝2和r₂[n+1]＝(y₂[n+1])₇＝4。The next moment is n+1 moment, the 16th-order FIR filter of the first branch is still working normally, and the output is

The 16th order FIR filter in the second branch still produces wrong output due to wrong filter coefficients

The filtered data after the two filtering processes are respectively processed by modulo 7, and r ₁ [n+1]=(y ₁ [n+1]) ₇ =2 and r ₂ [n+1]=(y ₂ [n +1]) ₇ =4.

In the third branch, the output of the 16-order FIR filter based on modulo 7 at time n+1 is $r[\mathrm{no}+1]=\underset{i=\mathrm{no}+1}{\overset{\mathrm{no}+16}{\mathrm{\Σ}}}{\left({\left(x_i\right)}_7{\left(h_{16-i+\mathrm{no}}\right)}_7\right)}_7=2.$

At this time, y ₁ [n]≠y ₂ [n], r ₁ [n], r ₂ [n] and r[n] still need to be compared.

r ₁ [n]=r[n]=2, r ₂ [n]≠r[n], it can be judged that the filter of the second branch has failed, and the first filtered data y ₁ of the first branch [n+1]=37 as result data.

Therefore, the correct outputs at time n and n+1 are y[n]=b ₁ [0]=15 and y[n+1]=y ₁ [n+1]=37, respectively. Outputting the result data means outputting the data in the first cache and clearing the data in the second cache.

As can be seen from the above, when the present application cannot judge the faulty branch temporarily, the data can be stored in the cache, and the data processing process can be continued until the faulty branch is judged, and then the data in the corresponding cache of the correct branch is output. And the data in the cache corresponding to the faulty branch is cleared, and the whole process does not affect the overall data processing speed of the system.

In the fault-tolerant signal processing technology based on remainder checking, the probability of missed detection and the resource consumption of fault-tolerant technology are contradictory: the smaller the modulus, the simpler the branch operation and the smaller the resource consumption, but at this time the fault missed The greater the detection rate is; on the contrary, the greater the modulus, the smaller the failure detection rate, but the greater the resource consumption.

The present application can simultaneously obtain low resource consumption brought by small modulus and low probability of missed detection brought by multi-sampling decision, without reducing the data processing throughput of the filter.

The method described in this application is carried out simulation test on Virtex-4FPGA, and test result is as follows:

For the 16-order FIR filter with 8 bits of input sample data and 8 bits of filter coefficient, when the modulus is 7, since the input sample data and filter coefficient processed by the third branch filter are both 3 bits, the third The consumption of branch processing resources is greatly reduced, so that the fault-tolerant signal processing technology based on the remainder check can save about 25% of FPGA resources compared with the triple-mode redundancy technology, but the single-sampling judgment method at this time has a failure detection rate as high as 14.3% and when the modulus is 63, although the fault missed detection rate of the single-sampling judgment method can be reduced to 1.6%, the input sampling data and filter coefficients processed by the third branch filter are only reduced to 6 bits from 8 bits , the resulting reduction in resource consumption is completely offset by the resource consumption introduced by the remainder operation of each step of operation, and finally makes the resource consumption of the filter fault-tolerant technology based on the remainder check and the resource consumption of the three-mode redundancy technology almost the same. quite.

And the method described in the application adopts multi-sampling judgment method, if adopt 3 sampling judgments, and keep modulus to be 7, then under the situation that keeps 25% FPGA resource saving, fault missed detection probability is reduced to 0.3%, compared with The miss rate for a single-sample decision at modulo 63 is much lower.

It can be seen that, for the error-tolerant signal processing technology based on remainder checking, the multi-sampling decision method adopted in this application can not only maintain the low resource consumption brought by the small modulus, but also greatly reduce the failure detection rate.

When the researchers of this application were studying the error-tolerant filtering processing technology based on the remainder check, they did not stop at the advantages of low resource consumption produced by the single-sampling decision method, but went deeper to explore the cost of low resource consumption. What is it, and then creatively discovered the problem of missed fault detection in the single-sampling decision method, as well as the contradiction between the missed detection probability and resource consumption.

Aiming at the problem of missed detection, the researchers of this application creatively proposed a multi-sampling judgment method after in-depth research, which can effectively avoid the problem of missed detection, and maintain a relatively low resource consumption, so that the remainder-based verification Fault-tolerant filtering technology has taken another step forward.

Each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to each other.

Finally, it should also be noted that in this text, relational terms such as first and second etc. are only used to distinguish one entity or operation from another, and do not necessarily require or imply that these entities or operations, any such actual relationship or order exists. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of, or also include elements inherent in, such a process, method, commodity, or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The multi-sampling decision method in the error-tolerant filtering process based on the remainder check provided by the application has been introduced in detail above. In this paper, specific examples are used to illustrate the principle and implementation of the application. The above embodiments The description is only used to help understand the method of the present application and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation and application scope, in summary , the contents of this specification should not be construed as limiting the application.

Claims (8)

1. a multi-sampling decision method in the fault-tolerant filtering process based on remainder check, it is characterized in that, comprising: S11, respectively inputting the sampling data into the first branch, the second branch and the third branch; S12, the processing steps of the first branch and the second branch include: performing filtering processing on the sampling data respectively to obtain processed first filtering data and second filtering data; Taking the remainder of the first filtering data and the second filtering data respectively to m to obtain the corresponding first data and the second data, wherein m is the modulus of taking the remainder, and m is a positive integer; S13, for the third branch, input the sampling data into a remainder-based filter for filtering processing to obtain processed reference data, where the modulus of the remainder in the remainder-based filter is m; S14. If the first filtered data and the second filtered data are not equal, respectively compare the first data and the second data with the reference data; The results of the comparison include: If the first data is equal to the benchmark data and the second data is not equal to the benchmark data, execute S15; If the second data is equal to the benchmark data and the first data is not equal to the benchmark data, execute S16; If the first data is equal to the second data and is equal to the reference data, execute S17; S15, outputting the first filtered data of the first branch as result data; S16, outputting the second filtered data of the second branch as result data; S17, store the first filtering data and the second filtering data in the first cache and the second cache respectively, and continue to input the sampling data and execute S11 to S14 until one of the first data and the second data is not equal to the reference data . 2. The method according to claim 1, wherein, if the first filtered data of the first branch is output as result data, the method further comprises: output the first filtering data in the first buffer, and clear the second filtering data in the second buffer. 3. The method according to claim 1, wherein if the second filter data of the second branch is output as result data, the method further comprises: Outputting the second filtering data in the second buffer, and clearing the first filtering data in the first buffer. 4. The method according to claim 1, wherein the filter comprises a FIR filter and an IIR filter. 5. method according to claim 1, is characterized in that, for the filter based on remainder in the 3rd branch, the operand that participates in multiplication operation at every turn m gets remainder, and the operand after described taking remainder is equivalent Multiply, take the remainder of the multiplied result to m to get the corresponding modular multiplication result. 6. The method according to claim 1, characterized in that, for the filter based on the remainder in the third branch, the operands participating in the addition operation each time take a remainder to m, and the operands after the remainder are equal to Add, take the remainder of the added result to m to get the corresponding modular addition result. 7. The method of claim 1, further comprising: If the first filtering data is equal to the second filtering data, the processing of the first branch and the second branch are both correct, and the processing result of one branch is selected as output. 8. The method according to claim 1, wherein the comparison result further comprises: If the first data, the second data and the reference data are not equal to each other, both the first branch and the second branch are faulty.

Images