| 摘要 |
Read errors in data transferred from a remote memory to a buffer memory are detected and corrected by a series of error detection and correction techniques. In the present invention, the transferred data includes user data, a checksum that detects read errors in the user data, row and column syndromes that identifies read errors in the user data, and a Hamming code that identifies read errors in the row and column syndromes. To minimize any performance degradation, a checksum is initially calculated from the user data and compared with the stored checksum. If an error is detected, a Hamming code is calculated from the stored row and column syndromes and compared with the stored Hamming code. Corrections are made, as needed, and then row and column syndromes are calculated from the user data. The calculated row and column syndromes are then compared with the stored row and column syndromes the identify and correct single-bit read errors, and report multiple-bit read errors. |
