search

deltapi / abc-feedback

Collection of bugs and Feedback regarding the abc triple search on https://abc.tngtech.com
7 stars 0 forks source link

Runs that find 0 triplets return instantly; Ones with solutions take minutes #19

Open zhilvinas opened 2 years ago

zhilvinas commented 2 years ago

General observation about which I am quite confused. Runs that successfully find single or multiple triplets take around few minutes to run. The runs that do not find triplets return instantly.

Why is it so? If there is some instant early-out condition that a range will not contain any solutions, why is it even being served as a job? Otherwise I would still expect some number crunching to, well, check the range for solutions?

Likta commented 2 years ago

@zhilvinas see #12

zhilvinas commented 2 years ago

Thank you for the reply, the passing comment about 0s being quick to find eluded me. That was silly of me, given I read all previous feedback before posting this... From a curious bystander point of view this feels odd, even suspicious. Perhaps this disparity would be worth highlighting on the website?

Likta commented 2 years ago

It also confusing to me, in #12 is only part answer to your questions, about why it's not filtered before we get it to crunch on.

SP-au commented 2 years ago

I have been running the code for a few days. There are periods where the result is returned very quickly, but there are also periods where it takes quite a long time to return even a negative result. My guess is that's just the natural effect the calculation has.

deltapi commented 2 years ago

Hey all, just for you to know, we ported an algorithm that does no rigorous prefiltering of intervalls. So, there is a quite large chance that you get unlucky with the intervalls. I am currently redesigning the algorithm in order to be more "rewarding", but that is not an easy task at all. Please stay tuned! And thank you all for your participation.