4 ms·
Due to how round-off is specified, cumulative errors in floating point calculations should average to 0 for typical workloads.
by batterseapower 15y ago
Due to how round-off is specified, cumulative errors in floating point calculations should average to 0 for typical workloads.
- maaku 15y agoNo. If you were to repeat an experiment a million times, the standard deviation should (quickly) limit to zero. That does not mean that the value compued is correct, however. Due to how round-off is specified, the estimated error will increase with every floating point operation.