FastCarPlay

This is C++ implementation of carplay receiver for "Autobox" dongles. The purpose of the project was to make application lightweight to run on Raspberry PI Zero 2W and use hardware decoding.

Dongles

The dongles are readily available from Amazon or Aliexpress labeled by . Devices might have different vendor and product id's. Check your with lsusb and update settings if necessary.

Setup

Dependencies

The project is based on SDL2, FFMPEG, LIBUSB. It use XXD for resource embedding.

sudo apt install build-essential xxd libsdl2-dev libsdl2-ttf-dev libavformat-dev libavcodec-dev libavutil-dev libswscale-dev libusb-1.0-0-dev

To run the application you also need to install runtime

sudo apt install ffmpeg libsdl2-2.0-0 libsdl2-ttf-2.0-0 libavformat59 libavcodec61 libavutil57 libswscale7 libusb-1.0-0

Build and run

The application can be started with settings file. Sample of the settings file can be found in settings.txt The project is using make. From the repository root run following

make clean
make release
./out/app ./settings.txt

Customisation

You can change font and background images by replacing files in ./src/resource

background.bmp for background image. Use BMP format only.
font.ttf for font. Use TTF fdrmat only

The names of the file need to be exactly same. Resources are embedded in executable, remake the project to regenerate resources

make clean
make release

Keys

The following keys have been mapped:

Left - navigate left
Right - navigate right
Enter - select active item
Backspace - Go back
f - toggle fullscreen mode

Status

What is working:

Video
Audio (multiple channels)
Key navigation
Simple touch

What is not working:

Multi touch - i have no means to test it
Microphone - that's next step for me to figure out how to feed sound
Telephone - the listening part will work, but because there is no mic implementation you can't speak

Notes

Regardless the resolution there are 2 types of settings

source-width source-height source-fps - defines what video parameters will be requested from device
width height fps - defines video drawing resolution and fps

The SDL will anyway scale the image to your screen/window size using internal HW scaling, so ideally you should use same values for width and height. If the source parameters will be different the scaling will be applied, however that scaling is not hardware accelerated and can consume a lot of CPU. You can set the 'scaler' in setting to define scaling algorithm. On my Raspbery Pi Zero 2W even easiest algorithm loads CPU to 100% and cause fram drop.

Increasing FPS above Source-FPS will cause app to run UI loop with less delays and do more event polling. This can increase responsivenes of the system, but also will make X11 to use more resources.

Next plans

Implement direct buffer transfer from video decoder to renderer (should reduce amount of memory copies and CPU load)
Control audio buffers better (now system use 3 decoding threads but in reality only 2 required)
Reduce music volume when there is navigation messages
Add abilities to run script on device connect and device disconnect