Open philzook58 opened 4 years ago
In general any example that requires deep exploration of memory states is probably promising, we might want to look at one of the crackme examples.
Haven't tested too thoroughly, but nqueens should work for n=9 to 16 now. Also, the debruijn example (from memory) used to choke at 16 bits. We could try scaling that one as well (16, 32, and 64 bit versions are in theory implemented). I'm not sure how much different this is, scaling-wise, to the nqueens example, but it is another data point.
+1 for getting a crackme running. I can take a stab at this after the grammatech stuff.
I had a trick that turns out not to work for n=9 in the nqueens example. I fixed that and tried running it with the phil/eval_speedup branch. It took maybe 30 minutes for me but seems to output a correct answer now.
We have a good number of simplistic examples and we have difficult real world examples, but it seems like we could use some controlled examples that push the boundary of what our tool can scale to. N-queens for different n is good but it is currently bounded to 8 due to it's implementation. There may be other aspects we may want to check scaling with respect to, such as number of function calls or total memory usage of arrays, number of dereferences and indirections, branches, etc.