X-ABFT
A class of online fault-tolerance techniques for matrix operations. The basic approach applies checksums to the input matrix along rows or columns and detects and locates computational faults by comparing output checksums against expected values. It belongs to the category of lightweight checksum-based schemes.
Timeline (last 90 days)
The paper reports that the sparse-recovery decoder uses at least 2 times fewer reads than the tuned X-ABFT baseline, but the authors note that X-ABFT in this scenario performs more reads than reading everything, making it a stress test rather than a like-for-like baseline.
X-ABFT is a class of online fault-tolerance techniques for matrix operations that detect and locate faults in computation by applying checksums to the rows/columns of the input matrix.
All Facts (2)
The paper reports that the sparse-recovery decoder uses at least 2 times fewer reads than the tuned X-ABFT baseline, but the authors note that X-ABFT in this scenario performs more reads than reading everything, making it a stress test rather than a like-for-like baseline.
50%UnverifiedX-ABFT is a class of online fault-tolerance techniques for matrix operations that detect and locate faults in computation by applying checksums to the rows/columns of the input matrix.
50%