I've been fascinated by CPs for some time. In particular, efficient ways in which to generate them. So I placed my source code on github that demonstrates a few ways I've done it in the past. My hope is that others can offer suggestions and insight into different (especially faster) approaches to generate them.
The approach of hardcoding a set of nested loops per each possible count of input sets is probably okay for most common use-cases, but wouldn't a more general algorithm be better? You'd either need to generate N nested loops via template metaprogramming, or else use an algorithm that didn't explicitly have the nested loops.
I've thought of doing something more general (see my while loops), and I've looked for other bits of source code that use a general approach, but haven't found any in C++. Maybe I've just over-looked it
You're right, most people hard code the for loops they need. I call that the "intuitive approach". It works and it makes sense I think and performs OK. In fact, whenever I come up with new ways to do CPs I always compare their performance to the hard coded loops.
3 comments
[ 3.7 ms ] story [ 21.7 ms ] threadI've been fascinated by CPs for some time. In particular, efficient ways in which to generate them. So I placed my source code on github that demonstrates a few ways I've done it in the past. My hope is that others can offer suggestions and insight into different (especially faster) approaches to generate them.
You're right, most people hard code the for loops they need. I call that the "intuitive approach". It works and it makes sense I think and performs OK. In fact, whenever I come up with new ways to do CPs I always compare their performance to the hard coded loops.