Developing and deploying apps

Client runtime environment

The BOINC client runs apps in a particular way. Each job runs in its own directory; its input and output files either reside in that directory, or are linked to a 'project directory' where jobs can share files. The app communicates with the BOINC client by sending messages via shared memory.

The runtime environment is described here. You don't need to know the details but you should understand the basic ideas.

Tools

• Adding applications and platforms
• Adding app versions

Native apps

A native app is one that uses the runtime system directly. You need to make minor modifications to the source code, compile it with BOINC header files, and link it with the BOINC runtime library.

• API for native apps
• Cross-platform utility functions
• Example applications

Wrappers

'Wrappers' let you run existing executables under BOINC.

• The BOINC wrapper
• Apps that run in Docker containers
• Apps that run in VirtualBox virtual machines

Multicore applications

• Multicore applications
• OpenCL CPU applications
• MPI applications

GPU applications

• Applications that use GPUs
• CUDA applications
• OpenCL applications

Adding graphics

• Graphics apps
• API for graphics apps
• Graphics apps using HTML

Specialized application types

• Non-CPU-intensive applications
• Sporadic applications
• Apps that do network communication
• Intermediate upload
• Trickle messages

Languages other than C/C++

• FORTRAN
• Java
• Python

Low-level APIs

• Startup and status data
• Diagnostics
• Trickle messages API
• API for wrappers

Building applications

• Intro
• Windows
• Mac OS X
• Linux
• Android
• CUDA and OpenCL
• Dynamic library naming issues

Debugging applications

• Debugging applications
• Windows
• Android

Beta-testing applications

• Beta-testing applications

Customizing the scheduler

• Job level scheduler customization