Algorithm based fault tolerance versus result-checking for matrix computations

Software-implemented fault detection for high-performance space applications

“…Both techniques have since been extended and refined by several researchers [4,7,10,13,14,2,6]; a comparison of RC and ABFT is in [12]. There are two designer-selectable choices controlling the numerical properties of this fault detection system: the checksum weights w and the comparison method indicated above by ?…”

Adaptable assertion checking for scientific software components

“…Due to the very precise, compute-intensive nature of science and engineering applications, they are more susceptible to overflow, underflow, and round-off errors than most IT applications [12,18,20]. The aggregation of round-off errors over the life of an iterative computation that can take days, weeks, or months to run can result in a tremendous waste of time and compute resources.…”

“…Like result checking techniques [18,20], postconditions depend upon the function being computed regardless of the underlying implementation algorithm.…”

Performance evaluation of checksum-based ABFT

“…The techniques were ranked based on MTTF and cost. Prata et al compared ABFT with result checking based on error coverage, overhead and ease of use [8]. In [2], ABFT for detection only was compared to detection and correction in terms of error coverage, memory size and execution time.…”

“…Hudak et al used FIAT to inject faults by changing a memory location (code or data) [7]. In [8], Xception was used for emulating transient faults in ALUs, data and address busses, general-purpose registers, condition registers and memory.…”