Skip to content

combine

Benchmarks of ordito.combine against every library that computes the same thing. Each chart shows the time of every library relative to ordito on CUDA (1x, in green): a longer bar took longer. The table under each chart lists every input the function was measured on. See Benchmarks for how to read and reproduce these numbers.

Measured on GPU: NVIDIA GeForce RTX 5090 · CPU: Intel Xeon w7-2495X

Summary

0.1x1x10x100x1000xordito (CUDA): the reference, 1xordito (CUDA)1xPyTorch3D (CUDA): 13x ordito's time, geometric mean over 1 functions; slower than ordito in 1PyTorch3D (CUDA)13x (1)MeshLib: 15x ordito's time, geometric mean over 2 functions; slower than ordito in 2MeshLib15x (2)trimesh: 163x ordito's time, geometric mean over 2 functions; slower than ordito in 2trimesh163x (2)Open3D: 181x ordito's time, geometric mean over 1 functions; slower than ordito in 1Open3D181x (1)MeshLab: 346x ordito's time, geometric mean over 1 functions; slower than ordito in 1MeshLab346x (1)
Time relative to ordito (CUDA), geometric mean over the combine functions both implement; in parentheses, the number of functions. Log scale. Runs stopped at the time limit are left out.
Values
Library Time relative to ordito Functions compared ordito faster in
PyTorch3D (CUDA) 13x 1 1
MeshLib 15x 2 2
trimesh 163x 2 2
Open3D 181x 1 1
MeshLab 346x 1 1

split

0.1x1x10x100x1000xordito (CUDA): 278.8 µs, 1x ordito's timeordito (CUDA)1x · 278.8 µsMeshLib: 2.19 ms, 7.9x ordito's timeMeshLib7.9x · 2.19 mstrimesh: 37.93 ms, 136x ordito's timetrimesh136x · 37.93 msMeshLab: 42.79 ms, 153x ordito's timeMeshLab153x · 42.79 msOpen3D: 66.73 ms, 239x ordito's timeOpen3D239x · 66.73 ms
sphere_med (81 920 faces) · time relative to ordito, log scale
Values
Library Median Relative to ordito
ordito (CUDA) 278.8 µs 1x
MeshLib 2.19 ms 7.9x
trimesh 37.93 ms 136x
MeshLab 42.79 ms 153x
Open3D 66.73 ms 239x
All inputs
Input ordito (CUDA) MeshLab MeshLib Open3D trimesh
sphere_med 278.8 µs 42.79 ms 2.19 ms 66.73 ms 37.93 ms
parts_64 566.0 µs 135.20 ms 2.08 ms 72.81 ms 37.08 ms
parts_1024 1.45 ms 1.631 s 2.36 ms 278.63 ms 172.07 ms

concatenate

0.1x1x10x100xordito (CUDA): 151.1 µs, 1x ordito's timeordito (CUDA)1x · 151.1 µsPyTorch3D (CUDA): 952.1 µs, 6.3x ordito's timePyTorch3D (CUDA)6.3x · 952.1 µsMeshLib: 3.53 ms, 23x ordito's timeMeshLib23x · 3.53 mstrimesh: 11.77 ms, 78x ordito's timetrimesh78x · 11.77 ms
sphere_med (81 920 faces) · copies=8 · time relative to ordito, log scale
Values
Library Median Relative to ordito
ordito (CUDA) 151.1 µs 1x
PyTorch3D (CUDA) 952.1 µs 6.3x
MeshLib 3.53 ms 23x
trimesh 11.77 ms 78x
All inputs
Input ordito (CUDA) MeshLib PyTorch3D (CUDA) trimesh
sphere_med copies=8 151.1 µs 3.53 ms 952.1 µs 11.77 ms
sphere_med copies=512 670.3 µs 108.02 ms 18.47 ms 582.84 ms

Without a reference

No other library computes these, so they have no comparison. ordito (CUDA)'s median time on the largest input it was measured on:

Function Input ordito (CUDA)
split_with_offsets parts_1024 431.7 µs