slow execution when generating an all line map

[akv-comp-arch]

Hi,

I am currently using the package to analyse the street network of Old Delhi. My code runs fine when I am executing it for a small pilot area. However, when running for the entire area, it takes days to generate an all line map, and then another day (or more) to reduce it to a fewest line map.

I was wondering if there is any way I can speed things up?

[pklampros]

Hi @akv-comp-arch, I'm afraid there isn't... These algorithms are quite old and unoptimised, so unless you plan to look at the algorithm itself in C++ there's no other way to speed things up.

[akv-comp-arch]

I was wondering if I can use a parallel processing package and run the all line map function within it. Will it work? I am not really sure if it will result in a speed up as creating the map is inherently "serial", the outputs are depended on previous outputs.

[pklampros]

You might as well try, but I seriously doubt it's going to work, as this is a single process through and through. One suggestion would be to split up the city into smaller parts and then have those parts given to the parallel processes. Then you can create axial lines manually to join the parts up, however that might be a lot more work than it's worth.

You might also try reducing the number of vertices in your map, as the process goes through the vertices of the street canyon.

[akv-comp-arch]

I tried to run the all line function using the parallel package using all the cores of my system. It generated the map in two and a half hours (for a map which previously took 24 hours for generating 8% of the all line map). However, I noticed that some parts of the map was missing in the all line map. That same map gives complete results in the depthmap software. I am not sure what is the issue here. Do you have any suggestions?