Should add option to have a --max-freq and --min-freq option where the probability of simulating a crash will have an additive increase when an error if found, and a multiplicative decrease when a verification is not found. Reasonable defaults should be provided, such as 1% minimum and 10% maximum, where the default is (min + max) / 2, with the additive constant being 0.05 * default and multiplicative being 0.5. The goal is that performance will be preserved for larger applications where errors occur very rarely, say, 1% of the time; by scaling back when no error is found, it can adapt to this, but during the 1% of time where the error occurs, its possible to ramp up and find errors a lot more often. Perhaps instead experiment with Multiplicative Increase/Additive Decrease as we technically want to ramp up fast when errors occur and slowly dwindle when no error is found.
Should add option to have a
--max-freqand--min-freqoption where the probability of simulating a crash will have an additive increase when an error if found, and a multiplicative decrease when a verification is not found. Reasonable defaults should be provided, such as 1% minimum and 10% maximum, where the default is (min + max) / 2, with the additive constant being 0.05 * default and multiplicative being 0.5. The goal is that performance will be preserved for larger applications where errors occur very rarely, say, 1% of the time; by scaling back when no error is found, it can adapt to this, but during the 1% of time where the error occurs, its possible to ramp up and find errors a lot more often. Perhaps instead experiment with Multiplicative Increase/Additive Decrease as we technically want to ramp up fast when errors occur and slowly dwindle when no error is found.