Should BLOCKY here be m2 instead?

orzzzjq / Parallel-Banding-Algorithm-plus

Compute the exact Euclidean Distance Transform and Voronoi Diagram for 2D and 3D binary images using the GPU.

https://www.comp.nus.edu.sg/~tants/pba.html

MIT License

73 stars 8 forks source link

Should BLOCKY here be m2 instead? #6

Closed aaronma37 closed 3 years ago

aaronma37 commented 3 years ago

https://github.com/orzzzjq/Parallel-Banding-Algorithm-plus/blob/2f58720302c3478d4367b21268dfbb263b46c4d2/pba-plus-3D/pba/pba3DHost.cu#L105

Same for BLOCKY in m1 in the function above. In the 2D version, this is the case.

Does the current code imply phase 1 band = 4, phase 2 band =4, and phase 3 band =2?

orzzzjq commented 3 years ago

For 3D distance transform, the banding method is not useful as there is already enough computation. Currently m1=m2=1, m3=2. You can see https://www.comp.nus.edu.sg/~tants/pba.html for more info.