Simple Dockerfiles for both CPU and GPU

resources/docker/README.md

@@ -0,0 +1,75 @@
+# Docker files for Mitsuba 2
+
+Here are Docker files for running Mitsuba 2 either CPU only, or also with GPU support.
+
+**Attention!** For GPU and CUDA/OptiX to work, you need to have new Nvidia drivers with builtin OptiX libraries, and also updated Docker and Nvidia Docker that provide proper mounting of OptiX libraries. You do **not** need to install OptiX SDK, it is **not** necessary for new versions of Mitsuba 2! Only the updated Nvidia drivers and Docker are needed.
+
+## Building the images
+
+Building the CPU only version:
+
+```
+cd /path/to/mitsuba2/resources/docker
+docker build -t mitsuba2:cpu -f linux-cpu.Dockerfile .
+```
+
+Building the CPU/GPU version:
+
+```
+cd /path/to/mitsuba2/resources/docker
+docker build -t mitsuba2:gpu -f linux-gpu.Dockerfile .
+```
+
+The images will be tagged as `mitsuba2:cpu` or `mitsuba2:gpu`.
+
+## Starting the containers
+
+Running the CPU image:
+
+```
+cd /path/to/mitsuba2
+docker run -it -p 45678:8888 -v $(pwd):/mitsuba2 -u $(id -u):$(id -g) mitsuba2:cpu bash
+```
+
+This will mount the Mitsuba 2 root directory to `/mitsuba2` inside the container and map port `8888` (used for Jupyter) from the container to port `45678` in the host machine.
+
+Running the GPU image:
+
+```
+cd /path/to/mitsuba2
+docker run --runtime=nvidia --gpus all -it -p 45678:8888 -v $(pwd):/mitsuba2 -u $(id -u):$(id -g) mitsuba2:gpu bash
+```
+
+This will do the same as above, and also mount all available GPUs to the container. You can try running `nvidia-smi` inside the container to verify the GPUs have been mounted correctly.
+
+## Compiling Mitsuba 2
+
+Works the same as described in Mitsuba 2 documentation.
+
+Note: Do not forget to modify `mitsuba.conf` if you want to add GPU variants.
+
+```
+cd /mitsuba2
+mkdir build
+cd build
+cmake -GNinja ..
+ninja
+```
+
+## Running Mitsuba 2
+
+```
+cd /mitsuba2
+source setpath.sh
+mitsuba --help
+```
+
+## Running Jupyter
+
+```
+cd /mitsuba2
+source setpath.sh
+jupyter notebook --port 8888 --ip 0.0.0.0
+```
+
+And then you can connect to the Jupyter notebook via `localhost:45678` in your web browser.

resources/docker/linux-cpu.Dockerfile

@@ -0,0 +1,32 @@
+FROM ubuntu:18.04
+
+# Install build tools, including Clang and libc++ (Clang's C++ library)
+RUN apt-get update && apt-get install -y \
+    clang-9 \
+    cmake \
+    libc++-9-dev \
+    libc++abi-9-dev \
+    libjpeg-dev \
+    libpng-dev \
+    libxcursor-dev \
+    libxinerama-dev \
+    libxrandr-dev \
+    libz-dev \
+    ninja-build \ 
+    python3-dev \
+    python3-distutils \
+    python3-setuptools \
+    python3-pip
+
+# Install basic Python tools
+RUN pip3 install jupyterlab numpy matplotlib ipywidgets pytest
+
+# create /.local so that Python and Jupyter are happy
+RUN mkdir /.local && chmod a+rwx /.local
+
+# Set C/C++ compilers to Clang
+ENV CC=clang-9
+ENV CXX=clang++-9
+
+WORKDIR /mitsuba2
+CMD ["/bin/bash"]

resources/docker/linux-gpu.Dockerfile

@@ -0,0 +1,36 @@
+# Start from CUDA 10.2 development image
+FROM nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04
+
+ENV NVIDIA_VISIBLE_DEVICES all
+ENV NVIDIA_DRIVER_CAPABILITIES compute,utility,graphics
+
+# Install build tools, including Clang and libc++ (Clang's C++ library)
+RUN apt-get update && apt-get install -y \
+    clang-9 \
+    cmake \
+    libc++-9-dev \
+    libc++abi-9-dev \
+    libjpeg-dev \
+    libpng-dev \
+    libxcursor-dev \
+    libxinerama-dev \
+    libxrandr-dev \
+    libz-dev \
+    ninja-build \ 
+    python3-dev \
+    python3-distutils \
+    python3-setuptools \
+    python3-pip
+
+# Install basic Python tools
+RUN pip3 install jupyterlab numpy matplotlib ipywidgets pytest
+
+# create /.local so that Python and Jupyter are happy
+RUN mkdir /.local && chmod a+rwx /.local
+
+# Set C/C++ compilers to Clang
+ENV CC=clang-9
+ENV CXX=clang++-9
+
+WORKDIR /mitsuba2
+CMD ["/bin/bash"]