阅读说明：本技术 一种sram型fpga故障在线容错方法 (SRAM type FPGA fault online fault tolerance method ) 是由 毛志明 张颖 姚嘉祺 华屹峰 杨济中 陈鑫 于 2021-07-22 设计创作，主要内容包括：本发明公开一种SRAM型FPGA故障在线容错方法,属于基本电子电路的技术领域。在已经知晓电路故障所在位置的前提下,将电路中故障导致的错误映射到后续连接的逻辑资源上,然后对受影响的逻辑资源进行基于slice的细粒度双模冗余复制。在完成冗余复制之后,对冗余复制资源进行纠错容错处理。将复制后的资源输出和故障连接到的原始逻辑资源输出进行第一次异或,用以判断电路所存在的故障导致原始逻辑资源的输出发生改变,从而实现了故障的在线监控。再将第一次异或后的输出和故障连接到的原始逻辑资源的输出进行第二次异或,使输出发生错误的原始逻辑资源的输出恢复正常,从而实现电路在线容错,以较少的资源消耗实现故障的在线监测和直接纠正。(The invention discloses an SRAM type FPGA fault online fault tolerance method, and belongs to the technical field of basic electronic circuits. Under the premise that the position of the circuit fault is known, mapping the error caused by the fault in the circuit to the logic resource connected subsequently, and then carrying out fine-grained dual-mode redundancy replication on the affected logic resource based on slice. And after the redundant copying is finished, carrying out error correction and fault tolerance processing on the redundant copying resources. And carrying out first exclusive OR on the copied resource output and the original logic resource output connected with the fault to judge that the output of the original logic resource is changed due to the fault of the circuit, thereby realizing the online monitoring of the fault. And performing second XOR on the output after the first XOR and the output of the original logic resource connected with the fault, so that the output of the original logic resource with the output error is recovered to be normal, thereby realizing online fault tolerance of the circuit and realizing online monitoring and direct correction of the fault with less resource consumption.)

1. An SRAM type FPGA fault on-line fault-tolerant method is characterized in that,

for the circuit which completes the fault test and diagnosis positioning, dual-mode redundancy copying of Slice fine granularity is carried out on the original logic resource which is affected by the fault according to the fault type;

performing first exclusive-or operation on the output of the original logic resource and the output of the copied Slice logic resource to obtain a monitoring result of whether the fault causes the output of the original logic resource to generate error change;

and performing second exclusive-or operation on the monitoring result obtained by the first exclusive-or operation and the output of the original logic resource, correcting the output error of the original logic resource caused by the fault, and transmitting the corrected output signal or the original logic resource to a subsequent circuit.

2. The SRAM type FPGA fault online fault tolerance method of claim 1, wherein the specific method for performing Slice fine-grained dual-mode redundancy replication on an original logic resource affected by a fault according to the fault type is as follows:

when the fault is a fixed fault or an open-circuit fault only existing on one interconnection line, mapping an error caused by the fixed fault or the open-circuit fault to a first logic resource connected with the interconnection line behind a fault node, and only performing Slice fine-grained copy redundancy on the first logic resource connected with the interconnection line behind the fault node;

when the fault is a leading bridging interconnection fault existing on a pair of interconnection lines, mapping an error influenced by the fault on a leading interconnection line to a first logic resource connected behind the leading interconnection line, and performing Slice-based fine-grained copy redundancy on the first logic resource connected behind the leading interconnection line;

and when the fault is a logic fault, mapping error output caused by the logic fault on the original logic resource acted by the fault, and performing Slice fine-grained copy redundancy on the original logic resource acted by the logic fault.

3. The SRAM type FPGA fault online fault tolerance method of claim 1, wherein when the first XOR result is 1, it indicates that a fault causes an erroneous change in the output of the original logic resource; when the result of the first exclusive-or is 0, it indicates that the fault does not cause an error change in the output of the original logic resource.

4. The SRAM type FPGA fault online fault tolerance method of claim 3, wherein when a fault causes an error change in the output of the original logic resource, a second XOR operation corrects the error in the output of the original logic resource caused by the fault, and transmits a corrected output signal to a subsequent circuit of the original logic resource; and when the fault does not cause the error change of the output of the original logic resource, the second exclusive-or operation maintains the output of the original logic resource unchanged, and the original logic resource is transmitted to a subsequent circuit.

Technical Field

The invention relates to an integrated circuit fault tolerance technology, in particular to an SRAM type FPGA fault online fault tolerance method, and belongs to the technical field of basic electronic circuits.

Background

The SRAM Field Programmable Gate Array (FPGA) has the advantages of multiple times of programming, flexible structure, short development period, etc. and is applied more and more widely, and the reliability requirement of the device itself is also continuously improved along with the wide application, so the research on the test of the SRAM FPGA is also more and more emphasized. In order to ensure the correctness of the user design on the FPGA, a series of processes such as application-related testing, fault location, fault tolerance and the like need to be performed on the FPGA resources used by the user design. After the fault testing and positioning are completed, how to carry out fault tolerance on the fault is the best way to determine the success or failure of the whole scheme.

After the fault test and the diagnosis and positioning are completed, namely on the premise that the position of the circuit fault is known, the common SRAM type FPGA fault-tolerant method replaces the fault resource or fault output by using redundant resources. The hardware-level redundancy fault tolerance keeps the function unchanged when a fault occurs, and mostly, a fault part of a circuit is remapped to another part based on row/column shift, so that the fault tolerance is realized; when the configuration-level redundancy fault tolerance fails, the circuit is re-laid and wired through a complex routing algorithm so that the circuit avoids fault resources to work; the redundancy fault tolerance of the circuit level is realized by performing dual-mode or triple-mode redundancy on key modules in the circuit and combining a majority voter to select correct output during circuit design, wherein the redundancy granularity is mainly based on the modules and consumes more resources, and the redundancy fault tolerance strategy is to combine the majority voter to vote, ignore wrong output and select correct output, thereby realizing online fault tolerance. Although the fault tolerance method using the redundant resources for replacement can realize fault tolerance on the fault and enable the circuit to recover normal operation, the methods have large resource area consumption, and only one avoidance measure can be continuously carried out on the fault when the fault occurs, and the fault which has occurred is not corrected.

Disclosure of Invention

The invention aims to provide an SRAM type FPGA fault online fault-tolerant method aiming at the defects of the background technology, which is characterized in that the logic resources for fault monitoring and error correction are copied in a fine granularity mode according to the fault type and the logic resources directly influenced by the fault, and the XOR result is transmitted to a subsequent circuit after the fault is monitored and corrected online by adopting two times of XOR operation, so that the fault resources can be processed, the reliability of the circuit is improved, the consumption of the resources is reduced, the influence on the performance of the circuit is reduced, the aim of monitoring and correcting the fault online by consuming less resources is fulfilled, and the technical problems that the SRAM type FPGA fault-tolerant method consumes the resource area and can only realize fault tolerance but can not effectively correct the fault are solved.

The invention adopts the following technical scheme for realizing the aim of the invention: the invention is different from the traditional large-area redundant resource replacement method based on row/column shift and the method for re-laying and wiring a circuit to avoid faults, provides the inventive concept of carrying out online fault tolerance on the circuit faults based on the fine-grained redundancy of a single slice (slice), maps the errors caused by the faults in the circuit to the original logic resources in the circuit, then carries out fine-grained copy based on the slice on the original logic resources according to different faults and fault types existing in the circuit, and carries out XOR processing on the original logic resources and the copied logic resources twice to realize fault tolerance on the faults. The first exclusive-or processing is carried out, the copied logic resource output and the original logic resource output connected with the fault are subjected to exclusive-or, and the fault can be monitored on line; and the second exclusive-OR processing is carried out, the output after the first exclusive-OR processing is subjected to exclusive-OR with the output of the original logic resource connected with the fault, so that the fault can be corrected on line, and the corrected output is connected with a subsequent circuit, so that the circuit can recover to work normally.

By adopting the technical scheme, the invention has the following beneficial effects:

(1) the fault online fault-tolerant method provided by the invention maps errors caused by faults to logic resources connected subsequently, and fine-grained copy is carried out on the original logic resources based on Slice and a plurality of interconnection line segments, so as to realize a redundancy strategy.

(2) The fault on-line fault-tolerant method provided by the invention realizes the error correction and fault-tolerant strategy through the subsequent two times of XOR operation, the two times of XOR operation are realized on an idle lookup table (LUT), the resource area consumption is relatively less, the output of the first exclusive-or processing can reflect whether the fault causes the error change of the original logic output, the second exclusive-or processing can directly carry out online correction on the error output of the original logic caused by the fault, and the corrected output is transmitted to the subsequent circuit to restore the circuit to normal work, compared with a configuration level fault tolerance method for re-laying and wiring the circuit to avoid faults based on a complex routing algorithm, the invention directly processes the faults through two exclusive OR logic gates on a fault tolerance strategy, has simple and convenient operation, and when the fault does not cause the logic output to have error change, the normal operation of the circuit can not be influenced.

(3) The fault online fault-tolerant method provided by the invention has the advantages that the resource consumption is low in a redundancy strategy, most of the fault online fault-tolerant method is based on Slice and redundancy copying of a plurality of interconnected line segments, in the fault-tolerant strategy, the fault output is directly subjected to error correction, the output after error correction is transmitted to a subsequent circuit, the error output is prevented from influencing the normal work of other circuits, and compared with a dual-mode/triple-mode redundancy fault-tolerant method for realizing correct output by switching a redundancy module after the error output is ignored, the fault online fault-tolerant method realizes online monitoring and correction of faults with less resource consumption.

(4) The error correction fault-tolerant strategy of the invention only inserts two exclusive-OR gates on the designed key path, has little influence on the designed time delay, and the two exclusive-OR gates can be realized by LUT, so the increased resource consumption is less compared with the traditional module-level error detection and correction resource consumption.

Drawings

FIG. 1 is a general block diagram of an SRAM type FPGA fault online fault tolerance method disclosed by the present invention.

Fig. 2 is a simplified diagram of a correct example circuit without faults.

Fig. 3 is a schematic diagram of a circuit structure for fault tolerance of logic faults.

Fig. 4 is a schematic diagram of a circuit configuration for fault tolerance of an interconnect failure.

Detailed Description

The technical scheme of the invention is explained in detail in the following with reference to the attached drawings.

As shown in fig. 1, the method for fault online fault tolerance of an SRAM type FPGA according to the present invention performs fine-grained dual-mode redundancy replication on a circuit that has completed fault testing and diagnosis positioning, that is, on the premise that the location of a circuit fault is known, on a logic resource directly affected by the fault, and then performs xor processing twice to implement fault online fault tolerance on the fault.

Firstly, mapping errors caused by faults existing in a circuit to logics connected subsequently, wherein the faults in the circuit are divided into logic faults and interconnection faults, aiming at the logic faults, no matter what types, the errors caused by the faults are mapped to slice logic resources where the faults are located, and slice-based fine-grained copy redundancy is directly carried out on the logic resources where the logic faults are located; for interconnection faults, different fault types need to be considered, a fixed fault or an open fault only exists on one interconnection line, errors caused by the fixed fault or the open fault are mapped to subsequent logic resources, namely, a first slice logic resource connected behind a fault node of the interconnection line only needs to be copied in a fine-grained manner to the logic resources, a leading bridging interconnection fault exists on a pair of interconnection lines, any one of the two interconnection lines is possibly affected by the fault, so that a leading party and a trailing party need to be known, and the logic resources are copied in a fine-grained manner based on the slice by mapping the errors affected by the fault on the leading interconnection line to the subsequent logic resources. It should be noted that, unlike a logic failure, the duplicated logic resources in the interconnect failure need to avoid the interconnect failure point for routing, so as to ensure that the duplicated logic resources are not affected by the failure.

Secondly, performing XOR processing twice on the resources after dual-mode redundancy, performing XOR processing for the first time, and performing XOR processing on the output of the copied logic resources and the output of the original logic resources connected with the fault; and performing XOR for the output after the XOR processing for the first time and the output of the original logic resource connected with the fault for the first time. And finally, taking the result after the second exclusive-or as the correct output of the logic resource circuit directly influenced by the fault, namely, transmitting the result of the second exclusive-or to a subsequent circuit to restore the normal operation of the whole circuit, thereby realizing the online fault tolerance of the whole circuit.

The error correction fault-tolerant strategy proposed by the invention is realized by two times of exclusive-or operation, and the truth table of the fault correction fault-tolerant strategy is shown in table 1. The third column in table 1 is the result of the first xor processing, which is the xor between the outputs of the dual-mode redundancy resource, because the copied resource has the same function as the original logic resource and the output is not affected by the fault, the xor between the two outputs can implement online monitoring of the fault, when the first xor result is "1", it indicates that the fault existing in the circuit causes an error change in the output of the original logic resource, and when the first xor result is "0", it indicates that the fault existing in the circuit does not affect the output of the original logic resource.

The fourth column in table 1 is the result of the second xor processing, which is the xor between the result of the first xor and the output of the original logic resource, if the fault in the circuit does not change the output of the original logic resource, that is, the result of the first xor is "0", the xor between the result of the first xor and the output of the original logic resource is performed again, because the xor result between "0" and any value is the characteristic of the value itself, the output result of the second xor is the same as the correct output of the original logic resource, and at this time, the result of the second xor is connected to the subsequent circuit, which does not affect the normal operation of the circuit; if the output of the original logic resource is changed in error due to a fault in the circuit, namely the first XOR result is '1', the first XOR result is XOR-ed with the output of the original logic resource, and the XOR result of '1' and any value is the opposite value of the value by utilizing the characteristic that the XOR result of '1' and any value is the opposite value of the value, so the output result after the second XOR is opposite to the error output value of the original logic resource, thereby correcting the output error of the original logic resource caused by the fault, ensuring the output to be recovered to be correct, connecting the result of the second XOR to a subsequent circuit, recovering the normal operation of the circuit, and realizing the online fault tolerance of the fault.

TABLE 1

Considering that the granularity of application-related test and fault diagnosis positioning of FPGA logic resources is mostly accurate to CLB, the CLB of the SRAM type FPGA contains 2 slices, the 2 slices are independent, and are respectively connected with a switch matrix and a universal wiring array, the invention considers that the granularity of logic fault tolerance is accurate to slice, and reduces resource consumption. A correct example circuit without fault is given in fig. 2, and now assuming that there is a logic fault in the circuit and it is located in Slice X1Y1 in CLB2, the proposed method of the present invention is used to perform fine-grained redundancy and fault-tolerant processing on the logic fault. A circuit diagram after fault tolerance for the fault is shown in fig. 3. Since the fault exists in Slice X1Y1 in CLB2, and the error caused by the fault is directly mapped on the logic resource where the fault exists, the spare resource Slice X1Y0 is directly used in CLB2 to copy the logic function in Slice X1Y1, and LUT resources are used in CLB4 to implement twice exclusive-or fault-tolerant processing, the twice exclusive-or circuit is described in the lower right corner of fig. 3, and the result of the second exclusive-or is transmitted as a correct output to Slice X2Y1 in the subsequent CLB3, so that the circuit recovers normal operation, and the logic fault online fault tolerance is implemented.

Considering that the granularity of application-related testing and fault diagnosis positioning of FPGA interconnection resources is mostly accurate to a programmable switch or an interconnection line segment, most of logic resources connected with faults after a fault node are LUTs, but a Slice of the SRAM type FPGA contains a plurality of resources such as LUTs, MUXs, DFFs, carry chains and the like, in a circuit, the LUT connected with the interconnection fault may be connected with other resources such as the MUXs, DFFs and the like in the same Slice, if redundancy copy is only performed on the LUT in the Slice, in a subsequent error correction fault-tolerant strategy, under the condition that the function of a designed circuit is not changed, more resources may be consumed in order to recover the original connection of the LUTs in the circuit, and therefore the invention considers that the interconnection fault-tolerant granularity is also accurate to the Slice. Also taking the example circuit in fig. 2 as an example, assuming that there is an open interconnection fault in the example circuit and the open interconnection fault is located in the AB segment, the interconnection fault is fault-tolerant by using the method proposed by the present invention, and fig. 4 is a schematic circuit diagram after fault tolerance. The fault error on the AB segment is mapped to the subsequent logic resource, i.e. Slice X1Y1 in CLB2, so that the spare resource Slice X1Y0 is also used in CLB2 to perform fine-grained copy on Slice X1Y1, but different from the logic fault tolerance, the copied logic resource needs to avoid a fault line segment or a fault programmable point for routing, i.e. the copied logic resource is guaranteed not to be affected by the interconnection fault, so that in fig. 4, to avoid the AB fault segment, the copied logic resource is connected by using a CD segment. And then, carrying out twice exclusive-or fault tolerance processing on the CLB4, wherein an exclusive-or circuit for twice is described in the lower right corner of FIG. 4, connecting the result obtained after the second exclusive-or as a correct output to Slice X2Y1 in the CLB3, recovering the normal operation of the circuit, and realizing the online fault tolerance of the interconnection faults.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.