[Question] Evaluating gradient at particle vs at point about fluid-engine-dev

[giordi91]

Apologies, after reasoning more about it, the derivative would be zero, for i==j, meaning we would end up adding nothing to the gradient. Avoiding the i==j case is basically a small optimization. Please correct me if wrong.
In my code I have organized the grid to hold the particles in buckets, so I don't have a list where for each particle I have all the neighbors, but from a position, I find a cell, from that cell(and adjacent cells) I get the contained particles, kinda tempted to leave the extra computation in to avoid more complex logic/branching. Of course I will only know by profiling. I will try to get back to you on the matter. Sorry for the bother.

[doyubkim]

Hi @giordi91 ! Yes, I think you got it right. i == j will give you zero gradient contribution from j. All the operators including gradient and laplacian can be computed from any random points in the space. Some non-SPH methods have 1/r factor in the equation which introduces singularities, but for SPH, it's all good :)