GPGPU

CPU & GPU implementations of Iterative Closest Point using a Vantage-Point Tree.

Examples

Horse Transformation 1

Horse Transformation 2

Build

To build the project, run the following commands:

mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make

This generates an executable at build/bin/main.

Usage

To run the project, run the following command from inside the build directory:

./bin/main <path/to/reference.txt> <path/to/transformed.txt> [options]

Options can be any of the following:

• --device cpu|gpu: the device to run on [default: gpu]
• --capacity <capacity>: the vp tree capacity [default: 32 on cpu, 256 on gpu]
• --iterations <iterations>: the maximum number of iterations [default: 200]
• --error <error>: the error threshold [default: 1e-5]

This should output on stdout:

• the transformation matrix.

This should output on stderr:

• the parameters used,
• the progress,
• the first ten points of the reference, and their closest point of the transformed with the transformation applied.

Tests

To build the tests, run the following commands if you haven't ran them earlier:

mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release

Then run:

make libcpu_tests

This generates an executable at build/bin/libcpu_tests. You can run this executable to test the lib cpu.

Benchmarks

To build the benchmarks, run the following commands if you haven't ran them earlier:

mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release

Then run:

make bench

Note: the benchmarks only compile in RELEASE build type.

This generates an executable at build/bin/bench. You can run this executable to launch the benchmarks. We recommand redirecting stderr to a file or null: ./bin/bench 2>/dev/null.