temujin

8 ms·

31.
by temujin 13y ago
Er, are you aware of the C++ standard library's std::nth_element() function? (Even my "pure C" programs tend to be compiled with g++ to gain access to that and std::sort().)
32.
by temujin 13y ago
This sort of thing is in fact exactly what I've been doing for the last year. (Took a widely used, but no longer maintained, program and sped/scaled up the important parts by 2-3+ orders of magnitude; this will soon become the of
33.
by temujin 13y ago
Yes, that point definitely should have been qualified better. If the new tool is much better than preexisting ones in important ways, it probably isn't a waste of time, and it could very well be a key asset. But he does have a point i
34.
by temujin 13y ago
Fair enough, and definitely agree with not making something you wouldn't use. (The "most [existing] stand-alone tools aren't flexible enough" problem is, however, one of the reasons why there's so much room to do b
35.
by temujin 13y ago
Yeah, libraries aren't a good way to start because there's not enough interest in using them. There are opportunities to build standalone tools which blow away their predecessors by multiple orders of magnitude, though; after gett
36.
by temujin 13y ago
Variation can be deliberately increased, not just decreased. And PGD creates an incentive for the former: the greater the difference between your own and your partner's DNA, the greater the expected benefit from informed selection.
37.
by temujin 13y ago
Floating-point multiplication generally isn't much slower than addition on modern machines. Logs are useful for avoiding floating point overflow/underflow, though.
38.
by temujin 14y ago
Yes, std::sort is better than qsort on simple types, but that's an isolated case. I'm currently working on a major "C++" project where literally the only C++ dependency is std::sort right now.