Fault tolerant computer
View Patent ↗A new method for the detection and correction of errors or faults induced in a computer or microprocessor caused by external sources of single event upsets (SEU). This method is named Time-Triple Modular Redundancy (TTMR) and is based upon the idea that very long instruction word (VLIW) style microprocessors provide externally controllable parallel computing elements which can be used to combine time redundant and spatially redundant fault error detection and correction techniques. This method is completed in a single microprocessor, which substitute for the traditional multi-processor redundancy techniques, such as Triple Modular Redundancy (TMR).
1. A computer system with improved fault tolerance from microprocessor data errors, comprising: a microprocessor; a fault tolerant software routine configured to send a first instruction and at least a second instruction to the microprocessor, the first and at least the second instructions being identical and inserted into spatially separated functional computational units of a VLIW microprocessor at different clock cycles; a first and at least a second memory device in communication with the microprocessor, the first memory device configured to store the first instruction, the second memory device configured to store at least the second instruction; a software instruction to compare the first instruction to at least the second instruction; and a comparator to compare the first instruction to the second instruction.
2. The system of claim 1 further comprising a third instruction sent by the fault tolerant software routine to the microprocessor, the third instruction stored in a third memory device in communication with the microprocessor.
3. The system of claim 2 wherein the software instruction directs the comparator to compare the first, second, and third instruction.
4. The system of claim 3 wherein a match of any of the first, second, and third instructions is accepted by the microprocessor.
5. The system of claim 1 wherein the VLIW microprocessor comprises a first, a second and a third spatially separated functional computational unit.
6. A software and hardware computer system with improved fault tolerance from microprocessor data errors, comprising: a very long instruction word (VLIW) microprocessor; a fault tolerant software routine comprising a first instruction and a second instruction, each inserted into two spatially separate functional computational units in the VLIW microprocessor at two different clock cycles and stored in a memory device in communication with the microprocessor, the first and second instructions being identical; a software instruction to compare the first and second instruction in the memory device in communication with the VLIW microprocessor compare or branch units, and configured to perform an action if the first and second instruction match, the fault tolerant software routine comprising a third instruction inserted into a third spatially separate functional computational unit in the VLIW microprocessor at a third different clock cycle and stored in a third memory device in communication with the microprocessor, the first, second, and third instructions being identical; and the software instruction to compare the first, second, and third instructions in the memory devices in communication with a VLIW microprocessor compare or branch units, and configured to perform an action if any of the first, second and third instructions match.
7. A method of processing data in a fault tolerant computer system, comprising: generating a first instruction at a first time interval; generating a second instruction identical to the first instruction at a second time interval; comparing the first and second instructions to each other; performing an action based on the matched first and second instructions; generating a third instruction identical to the first and second instructions at a third time interval if the first and second instructions do not match; matching the first, second, and third instructions to each other; and performing an action based on a match between any one of the first, second, and third instructions.