v0.3 Benchmarks

Introduction

This page shows performance comparisons against other libraries : Assimp, OpenCascade and Lib3MF.

Each benchmark comparison consists of four consecutive performance tests:

1. Read the input STL file into some mesh object M (eg. aiScene for Assimp)
2. Write mesh M to STL ascii file
3. Write mesh M to STL binary little-endian file
4. Write mesh M to STL binary big-endian file

In case some functionality is not supported by the opponent then N/A is displayed.
The column speedup is the execution time of the opponent divided by the execution time of gmio, it tells how much gmio is faster(ie. > 1) or slower than its opponent.

To be relevant, the benchmark code for gmio produces the same resulting objects as its opponent.
benchmark_assimp "read" produces an aiScene object the same way the Assimp importer does.
benchmark_opencascade "read" builds an StlMesh_Mesh object.
benchmark_lib3mf "read" builds a NMR::CMesh object.

Each performance test is run 5 times, only the best execution time is kept in the final result.

Running configuration

OS: Windows 7 Professional SP1 x64
CPU: Intel i5-4670 @ 3.4GHz
RAM: 16GB

Library	Version	Arch/compiler	Linkage	Note
gmio	v0.3.1	x64/vs2013	Static	See build options below (*)
Assimp	v3.1.1	x64/?	DLL	Official "release" binaries from sourceforge
OpenCascade	v7.0.0	x64/vs2013	DLL	Official "release" binaries
Lib3MF	#a466df4	x64/vs2013	Static	Rebuilt as a static library

(*) Build options of gmio:
GMIO_BUILD_STRICT_C90=OFF
GMIO_STR2FLOAT_LIB=irrlicht_fast_atof
GMIO_FLOAT2STR_LIB=google_doubleconversion

Results for testing model 1

File: Final_Eiffel_Tower.stl
Format: Binary STL (little-endian)
Count of triangles: 174762
Size: 8.33MB

Test	gmio	Assimp	speedup
read	14ms	18ms	1.29x
write(ascii)	158ms	1698ms	10.75x
write(binary/le)	7ms	129ms	18.43x
write(binary/be)	8ms	N/A	N/A

Test	gmio	OpenCascade	speedup
read	572ms	2221ms	3.88x
write(ascii)	259ms	3807ms	14.70x
write(binary/le)	9ms	4816ms	535.11x
write(binary/be)	10ms	N/A	N/A

Test	gmio	Lib3MF	speedup
read	71ms	79ms	1.11x
write(ascii)	166ms	N/A	N/A
write(binary/le)	10ms	30ms	3.00x
write(binary/be)	12ms	N/A	N/A

Results for testing model 2

File: Buddha.stl
Format: ASCII STL
Count of triangles: 540372
Size: 143.78MB

Test	gmio	Assimp	speedup
read	482ms	476ms	0.99x
write(ascii)	511ms	5576ms	10.91x
write(binary/le)	21ms	382ms	18.19x
write(binary/be)	23ms	N/A	N/A

Test	gmio	OpenCascade	speedup
read	2234ms	10734ms	4.80x
write(ascii)	731ms	9580ms	13.11x
write(binary/le)	28ms	12485ms	445.89x
write(binary/be)	31ms	N/A	N/A

Test	gmio	Lib3MF	speedup
read	749ms	N/A	N/A
write(ascii)	519ms	N/A	N/A
write(binary/le)	31ms	N/A	N/A
write(binary/be)	34ms	N/A	N/A

Results for testing model 3

File: Behotec-J66-HP.step
Format: STEP
Size: 21.78MB

This benchmark measures performance of writing TopoDS_Shape internal triangulation. It's using gmio_stl_mesh_occshape and OpenCascade's StlAPI_Writer.

Test	gmio	OpenCascade	speedup
brepwrite(ascii)	757ms	3326ms	4.39x
brepwrite(binary/le)	44ms	74ms	1.68x
brepwrite(binary/be)	46ms	N/A	N/A