Running MORSE headless

Running MORSE “headless” means running MORSE without launching a graphical user interface (GUI) at all.

This is useful in several scenarios. For example, running MORSE on a remote server; integrating MORSE in a Continuous Integration (CI) pipeline; running automated tests, etc.

While “headless” really means “no GUI”, it is also related to: “how to run MORSE without 3D acceleration (i.e., without a GPU)”. We address both points below.

Important

Headless MORSE is currently only tesed on Linux.

Overview

It is important to understand that MORSE requires OpenGL. There is currently no way around this, and, as a 3D simulator, we are likely to keep this requirement for the foreseeable future. So, running a “headless” MORSE still requires that your OS provides an OpenGL implementation.

OpenGL does not mandate a GPU, though. So, it is perfectly possible to run MORSE on a CPU-only computer (on a server, on a cluster in the cloud without GPUs, etc). Obviously, no GPU means no 3D hardware acceleration, but depending on your application, performance may still be perfectly okay. Some non-scientific benchmarks are provided below, for reference.

To run “headless”, we also need to prevent MORSE’s main window from appearing. This can easily be achieved by using Xvfb a special graphic server that renders 3D applications to memory instead of to the screen.

Important

The Xvfb X server does not support hardware acceleration, and will not make use of your GPU, even if you have one. Alternatives like Xdummy exist, but have not been tested.

CPU-based OpenGL

In the Linux world, the best option for CPU-based 3D acceleration is LLVMpipe. It is fairly easy to compile and install.

• Install the system dependencies. For Debian/Ubuntu, apt-get install llvm-dev scons python-mako libedit-dev
• Grab the lastest version of Mesa here (tested with 9.2.2 and 11.0.7): mesa FTP.
• Compile with:

$ scons build=release llvm=yes libgl-xlib

This will result in a new libGL.so that uses the CPU instead of the GPU. Blender runs pretty well with it.

To run MORSE with this library:

LD_LIBRARY_PATH=<path to your Mesa>/build/linux-x86_64/gallium/targets/libgl-xlib morse run <env>

Some quick performance results with the default environment:

$ morse create test

Then, edit test/default.py and add env.show_framerate() at the end to display the FPS.

On a test computer with mesa-11.0.7 and LLVMpipe on an Intel Core i7-4790 @ 3.6GHz and 16GB RAM, we achieved 30 FPS vs 60 FPS with 3D acceleration.

By only rendering wireframes (i.e., using MORSE’s fast mode: edit test/default.py and switch fastmode to True), we reached 60 FPS.

Note

In fast mode, you only get the wireframe of the models: this is fine if you do not do any vision-based sensing.

Going headless

To prevent MORSE from opening a window, you can use Xvfb to create a ‘fake’ display to run MORSE. On debian/Ubuntu, apt-get install xvfb.

$ Xvfb -screen 0 1024x768x24 :1 &
$ LD_LIBRARY_PATH=<LLVMpipe path> DISPLAY=:1 morse <...>

Which such a configuration, we successfully streamed full 3D 800x600 images with ROS at 20Hz.

Note

While you need to create a display with a 24-bit color depth, the size can be made smaller, for further improved performances. For instance, 100x100px is big enough to run MORSE. However, if you use video cameras, note that you can’t stream images larger than the window size.

Note

Assuming imagemagik is installed, you can easily take a screenshot of your invisible window with DISPLAY=:1 import -window root screenshot.png