EDAC FOR MEMORY PROTECTION IN ARM PROCESSOR

Transcription

1 EDAC FOR MEMORY PROTECTION IN ARM PROCESSOR Mrs. A. Ruhan Bevi ECE department, SRM, Chennai, India. Abstract: The ARM processor core is a key component of many successful 32-bit embedded systems. Embedded systems often use multiple memory devices. The probability of having multiple errors is increased only when the size of memory increased. In order to maintain good level of reliability, it is necessary to protect memory cells from errors by protection codes. Error Detection And Correction (EDAC) codes plays a main role in memory protection and for reliability improvement. So, in this paper, an error detection and correction method to protect the memory against the errors is proposed. HVD is high level error detection and correction method to protect against errors based on 4D parities checking for each row, column and diagonal in slash and black slash directions. It can correct up to 3 upsets in a data block but it can detect huge number of multiple faults in memories. It can improve the fault detection and correction coverage using parity codes which having low computation latency and implementation complexity. 1.Introduction: In many computer systems, the contents of memory are protected by an error detection. Ms.Srija Nadendla ECE department, SRM, Chennai, India and correction (EDAC) code. In order to maintain good level of reliability, it is necessary to protect memory cells, for this purpose, various error detection and correction methods are being used. The goal of any Error Detection And Correction code is to provide protection against transient errors (soft errors) that manifest themselves as bit-flips in the memory. These codes are usually implemented in hardware, but require extended memory bus architecture to accommodate parity bits and additional encoding/decoding circuitry. The reliability issue can be solved using other forms of redundancy than hardware redundancy because hardware redundancy schemes like duplication or triple modular redundancies are expensive. In low-cost satellite projects, the reliability of a system can be improved by cost effective software error detection and correction code schemes. An important advantage of using software-based code scheme is that, one can change it dynamically to meet the observed response of the memory devices. Various types of error detection and correction codes are used in computers, communication systems based on constrains exposed by applications. In this paper, a high-level error detection and correction method to protect

2 against soft errors is proposed. This method is based on parities for each row, column and diagonal in slash and backslash directions. This method, which is named HVD, provides very high detection coverage rate that can correct up to three upsets in a data block. To show tradeoffs between parameters, a large data block is partitioned into some smaller data segments. Then some experiments have been done to investigate the effect of data segments granularity in a data block. The results show that HVD method is able to detect 100% faults injected in the protected block and we prove that it can correct up to 3 upsets. Although the proposed method, includes memory overhead more than previous methods, however, it improves the fault detection and correction coverage using parity codes which have low computing latency and implementation complexity. 2. Proposed Method The proposed error detection and correction method is called HVD code as the parity bits are applied on the row, column and two diagonals on a data part. In addition to horizontal (H) and vertical (V) parity bits, the diagonal (D) parity bits in two directions can be used as shown in Fig 1. (a) horizontal (b) Vertical (c)slash direction (d)backslash direction Fig 1: parity calculation for data bits in each direction To increase the ability of detection, an additional parity bit is computed on the basis of calculated parity bits of each dimension. In our proposed HVD code implementation, h, v, d and d represent the number of errors in the horizontal, vertical and slash and backslash lines respectively.

3 Whereas h1, v1, f1 and b1 represent the position of first error in horizontal, vertical and the both diagonal parity lines, respectively. The proposed method HVD (Horizontal-Vertical-Diagonal) is based on 2-d parities. Parities are generated in the 4 directions that are horizontal, vertical, forward slash and back slash diagonals. Fig 2: parity bits for a given memory block The 8 x 8 array is given in Fig 2 where the symbols h, v, d and d denote the parity bits in horizontal, vertical, forward slash and backward slash respectively and the subscripts indicate the position of parity. 2.1: detection algorithm At the receiver end the parities are calculated in all the directions for whole array (for example, from h1 to h8 in horizontal direction in the above figure).after the calculation of parities at the receiver the comparison of received parities and the calculated parities is performed. If the result of comparison shows any difference, it means the received data at the receiver is incorrect and correction is required. So the erroneous parity lines are identified and then the correction process starts but if the comparison shows no difference, it means data is correct and no correction is needed. In order to detect bit upsets in the codeword, at receiver end, all mentioned types of parity bits on the data part of the codeword are required to be calculated again. As the different types of parity bits can be calculated independently in parallel i.e., while computing the horizontal parity bits, vertical parity bits are calculated simultaneously; encoding and consequently detection procedure can be done by means of parallel computation. In real time and high speed applications the above mentioned property becomes very useful. On the other hand, when the first difference between the received parity bits and the calculated ones is detected, the detection algorithm is then stopped and there is no need of further comparison between other calculated parity bits with the received ones. 2.2: correction algorithm In correction algorithm, we introduce how the data array can be corrected after detecting a probable error. First, it is supposed that three upsets are present only on data part that we need to correct. After receiving a codeword in the receiver side, the parity bits must be computed again. When the differences in calculated and received parity in the vertical, horizontal and both of the diameters dimensions is found and it sets the corresponding syndrome bit to one. As a result, four arrays are produced such as example; the elements of the

4 horizontal array include the number of the rows that has bit upsets. After that, all ordered pairs of them are produced by choosing two different arrays. Through every ordered pair, there may be maximum one candidate bit present. Note that no candidate bit may be determined from some case of the ordered pairs. Each parity bit has a corresponding line of codeword bits that generate it. Consider lines of each mistaken parity bit. When we intersect two lines, if there is a conjunction then it is a candidate bit. To find all candidate bits, we should intersect each two lines. Parity bits required for block read and write operations just in one cycle can be calculated by the proposed hardware. Therefore, the performance overhead of the method is significantly low. As this hardware is shared for all of the memory blocks, its area overhead is negligible compared to the other hardware structure used for error detection and correction code. All parity bits in horizontal, vertical slash and backslash directions are checked in read operation. We assume that the granularity of the memory access is a block. So the numbers of the parity bits to be calculated in the block read and write operations are the same. The number of XOR gates needed is shown in Table 1. Table 1 The number of the XOR gates used for a given block memory Parity symbol Number of XOR gates direction Horizontal h 7x8+8=64 Vertical v 7x8+8=64 Slash d (7x6)/2+(6x5)/2+14=50 backslash d (7x6)/2+(6x5)/2+14=50 It is assumed that a dedicated hardware is used to calculate Horizontal, Vertical, Slash and Backslash diagonal parities simultaneously in just one cycle. Meanwhile, in the cases that parity calculation delay can be tolerated, we can use a simpler hardware (i.e. use a hardware having a few number of XOR gates, but with delay penalty) to reduce the leakage power overhead. 3.Results In this paper an easy method with fewer computations is presented. This method can detect and correct the errors in data bits as well as in the parity bits without any extra calculations. As it is a software EDAC method so no extra hardware circuitry is required and also the transient faults and bit flips in the memory will be immediately corrected. 3.1 Percentage of Redundancy The simulation results shows that a small increase in the number of parity bits can cause an increase in number of data bits that can be transmitted as a whole codeword as shown in Table 2 and variation in data and code word shown in fig 3. Table 2 No. of data bits according to the no. of parity bits No.of parity bits No.of data bits

5 code size data size Fig 3:data vs code size 3.2 Comparison Between Triple Bit Error Correcting Codes HVD code is compared with other triple error correcting codes such as Golay code, BCH code and Reed-Solomon code. This comparison is based on bit overhead and code rate. It is presented in table 3 Table 3 Comparison of triple bit 8, 16, 32,64 and 128bit. The CPU time for different length of code is given in the Fig 4. Fig 4: cpu time analysis The analysis shows as the length of code is increases, the time required to correct the error increases non-linearly. 4. Conclusion In this paper, an easy method of error detection and correction with lesser computations is presented. It is a more simple method to find the errors by removal error correcting codes: of candidate bits, so it reduces the complexity. This method can detect and correct the errors in data bits as well as in Error No. of data No. of parity the No. of parity Bit bits overhead without Code any rate extra correcting bits bits codeword calculations. As (%) it is a software (%) EDAC method method bits so no extra hardware circuitry is Golay(23,12,7) required 23 and also the transient faults and bit Bch (31, 16, 7) flips in the memory will be immediately corrected. A large combination of multiple RS (63, 57, 7) 7 57 faults 64 can be corrected with this method. HVD(64) The 124 number of errors that can be corrected varies with the length of the coded word array. The number increases with the length As seen, that the proposed HVD of the array. In applications for which high code has the moderate bit overhead and code error detection coverage is very important, rate as compared to the other code schemes. HVD code can be useful. This method can detect and correct a single error in a 3.3 Time Analysis particular line of code of the coded word The CPU time are analysed on the array as it uses hamming codes for error same platform for the various word length of execution time(sec) data size

6 detection and correction and it is an only limitation of this method. Another advantage of this code is that it can be used in combination with the other code schemes that has high correction coverage and low detection coverage in comparison with HVD code, so that the number of errors can be increased that can be corrected in a single line. References [1] Mostafa Kishani, Hamid R. Zarandi, Hossein Pedram, Alireza Tajary, Mohsen Raji and Behnam Ghavami, HVD: horizontal-vertical-diagonal error detecting and correcting code to protect against with soft errors international conference conducted by Des Autom Embed Syst 15: (2011). [2] Argyrides C, Zarandi HR, Pradhan DK (2007) Multiple upsets tolerance in SRAM memory In: International symposium on circuits and system, New Orleans, LA, (May 2007). [3] Thirunavukkarasu U, Babu Anne N, Latifi S (2004) Three and four-dimensional parity-check codes for correction and detection of multiple errors In: International conference on information technology:coding and computing, p 480(2004) [4] Berrou, C., Glavieux A. and Thitimajshima P., Near Shannon limit error-correcting coding and decoding Turbocodes International Conference on Communications, (1993). [5] Fernanda Lima, Luigi Carro, Ricardo Reis Designing Fault Tolerant Systems into SRAM-based FPGAs Anaheim, California, USA, DAC 03, June 2-6, 2003.