How can I reduce memory allocation and execution time while ray tracing using VoxelRayTracers package?

General Usage Performance
1

Hi,

I am trying to use VoxelRayTracers.jl package to trace a ray through cubic voxels.

I want to improve the performance of ray tracing by reducing execution time because in my application I have to use many rays in three dimensions.

I wrote a simple example to show my case. I only used one ray, but variable itr represents number of rays. Then I have to sum traced voxel values in each ray.

using VoxelRayTracers
using BenchmarkTools
edges = (-2.0:0.1:2.0, -2.0:0.1:2.0)
ray = (position=[4.946152422706632, 3.433012701892219], velocity=[-0.8660254037844393,-0.5])
all_val = rand(41,41)
itr = 400000

function exit_data(ray, edges, val, itr)
      for i in 1:itr
            out = 0.0
            for hit in eachtraversal(ray, edges)
              out += val[hit.voxelindex]
            end
      end
end

I have checked execution time and memory allocation by using @btime

@btime exit_data($ray, $edges, $all_val, $itr)

and the result is as below:
710.617 ms (2949120 allocations: 145.00 MiB)

In my real case, I have more than 1.5M rays and points will be 3 dimensions… so, execution time and memory allocation will be very high… Moreover, I have to iterate this whole process more than 1000 times using another algorithm. So, can anyone suggest some ideas to improve performance in terms of execution time and memory allocation?

2

Try using static vectors here. This should eliminate many of the allocations inside eachtraversal if I understand correctly.

2 Likes
3

Its a bug. Inference could not handle this line, not sure why. I found a workaround anyway, fixed on master and tagged a release. There should be no need to use a static vector.

3 Likes
4

@mohamed82008 : Thanks for your reply. I tried with static vectors like ray = (position=@SVector[4.946152422706632, 3.433012701892219], velocity=@SVector[-0.8660254037844393,-0.5]) , but it give similar results with @btime exit_data($ray, $edges, $all_val, $itr)

5

@jw3126 : Thanks for your fix. Now, I am getting timing values as 506.445 ms (0 allocations: 0 bytes) and time of execution also improved by 200ms.

Thanks Again,
Manu

1 Like