Pipelined error checking and correction for cache memories.

摘要

A scheme for the implementation of single error correction, double error detection function is provided for cache memories wherein the normal cache access time is not affected by the addition of the ECC function. Check bits are provided for multiple bytes of data, thereby lowering the overhead of the error detecting and correcting technique. When a single error is detected, a cycle is inserted by the control circuitry of the cache chip. At the same time, the clocks for the CPU are held high until released by the cache chip on the next cycle. Error correction on multi-byte data is performed using, for example, the 72/64 Hamming code. The technique requires a 2-port cache array (one write port, and one read port). However, the density of a true 2-port array is too low; therefore, the technique is implemented with a 1-port array using a time multiplexing technique, providing an effective 2-port array but with the density of a single port array.