When a correctable error is detected in the system by an IP, it sends the error signal along with the error address to EPC module. The EPC module will scan this error address in the 16-entry CAM. If there is a match then the address is discard and no error is generated to ESM by the ECP. It takes one cycle to scan one address at a time through the CAM. The idea is to allow the system to tolerate a correctable error occurring on the same address because this error has been handled before by the CPU. This error scenario is particularly frequent when the software is in a for loop fetching the same address. Because there are multiple IPs which can simultaneously detect correctable errors in the system, the EPC employs a 4-entry FIFO per IP interface so that error addresses are not lost.

If an address is not matched in the CAM then it depends if there is empty entry in the CAM. If there is an empty entry then the new address is stored into the empty entry. For each entry there is a 4-bit valid key. When a new address is stored the 4-bit key is updated with "1010". It is programmable to generate a correctable error to the ESM if the address is not matched and there is an empty CAM entry. Once CPU is interrupted, it can choose to evaluate the error address and handle it accordingly. The software can also invalidate the entry by writing "0101".

If an dress is not matched and there is no empty entry in the CAM then the correctable error is immediately sent to the ESM. The new error address is lost if there is no empty entry left in the CAM.