초록 ▼

When a fault is detected in a processor, program execution in this processor is interrupted and taken up again by a standby processor from an earlier uncorrupted state, a recovery point. Such recovery points are specially provided in the program. A save copy of the state at the recovery points is cr

When a fault is detected in a processor, program execution in this processor is interrupted and taken up again by a standby processor from an earlier uncorrupted state, a recovery point. Such recovery points are specially provided in the program. A save copy of the state at the recovery points is created in each case in a state save unit by recording changes compared with the respective previous state. The data memory existing in the state save unit has pairs of memory words, in which arrangement the state at the recovery point last reached is in each case saved in one of the memory words and the current changes are in each case recorded in the respective other memory word. The memory words are accessed via pointers which are formed by a control logic from two check bits allocated to each memory word pair. Processing of the check bits and of the pointers is very fast. It is not necessary to copy data within the state save unit. The standby processor has direct access to the saved data.

대표청구항 ▼

In a fault-tolerant multiprocessor arrangement comprising at least a first and a second processor, a first data memory for storing memory words, each of which being stored under an address different from the address of another of said memory words, a second data memory for storing a pair of memory w

In a fault-tolerant multiprocessor arrangement comprising at least a first and a second processor, a first data memory for storing memory words, each of which being stored under an address different from the address of another of said memory words, a second data memory for storing a pair of memory words under each of said addresses, and a control logic with at least a check bit memory having for each pair of memory words stored under the same address in said second data memory at least two check bits for completing the address of each pair of memory words and a one-bit storage cell containing a control bit, an operating method comprising the steps of: said first processing during execution of a program containing recovery points writes a first of said memory words under a first of said addresses into said first data memory, transfers said memory word including said first address from said first to said second data memory, and writes said transferred first memory word and its respective address into said second data memory under control of said control logic, said first processor during execution of said program further transfers a recovery point signal indicating in each case the arrival of said program at a recovery point, thereby causing the control logic to invert said control bit of said one-bit storage cell at each recovery point in response to said transferred recovery signal, said control logic during the process of writing of said first memory word into said second data memory reads out of said check bit memory a first and a second check bit discriminated by said first address, forms a first pointer bit from said first, said second and said control bit in accordance with the rule: OLD(1):=BIT1(1).A+BIT2(1).A and a second pointer bit having a binary state which is the inverse of said first pointer bit, feeds said second pointer bit to an address input of said second data memory to complete the pair address for said first address of said first memory word of said second data memory, first newly determines the binary states of said and second control bits in accordance with the rules: BIT(1):=NEW(1).A+OLD(1).A, and DIT2(1):=NEW(1).A+OLD(1).A, writes said newly determined first and second bits into said check bit memory, said control logic in said interval between said two current recovery points further reads out of said check bit memory a third and a fourth check bit discriminated by a second address of a pair of secondary memory words into which not data have yet been entered within said interval, forms a third and a fourth pointer bit from said third and said fourth check and said control bit in accordance with the rules: OLD(2):=BIT(2).A+BIT2(2).A, and NEW(2):=BIT1(2).A+BIT2(2).A newly determines the binary states of said third and said fourth check bit from said third and said fourth pointer and said control bit in accordance with the rules: BIT1(2):=NEW(2).A+OLD(2).A, and BIT2(2):=NEW(2).A+OLD(2).A, and writes said newly determined third and fourth check bit into said check bit memory, said second processor in response to a fault signal of said first processor takes over and continues the execution of said program at a recovery point last reached by said first processor during execution of said program by use of the data states at this recovery point transferred into said second data memory, said control logic during the process of each reading of said second processor from said second data memory forms in the case of said first or said second address said first or said third pointer bit, and feeds said formed first or third pointer bit to said address input of said second data memory for completing the pair addresses for said first or second address of said first or second memory word of said second data memory, wherein A means said control bit, BIT(1), BIT2(1), BIT1(2) and BIT2(2) mean said first, second, third and fourth check bit, OLD(1), NEW(1), OLD(2) and NEW(2) mean said first, second, third and fourth pointer bit, the dot means a logical AND operation, the plus sign means a logical OR operation, and the bar above the symbols means negation.