Modules

cacao is split into modules. Each module is a directory at the top of the repository, and each one builds a shared library that milk loads at runtime. Within the milk CLI, a module’s commands are prefixed by its shortname.

Directory	Library	CLI shortname	Description
AOloopControl	cacaoAOloopControl	cacao	AO loop control engine: modal and zonal filtering, modal statistics. Also installs the cacao setup, task manager, FPS and cacao-aorun-* scripts.
AOloopControl_DM	cacaoAOloopControlDM	cacaodm	DM operation: DM channel combination, simulated turbulence, astrogrid
AOloopControl_IOtools	cacaoAOloopControlIOtools	cacaoio	I/O tools: WFS image acquisition, WFS camera simulation, spot finding
AOloopControl_acquireCalib	cacaoAOloopControlacquireCalib	cacaoac	Acquire calibration: linear response measurement
computeCalib	cacaocomputeCalib	cacaocc	Compute calibration: control modes, control matrices, Hadamard decoding, masks
AOloopControl_PredictiveControl	cacaoAOloopControlPredictiveControl	cacaopc	Predictive control filters
AOloopControl_compTools	cacaoAOloopControlcompTools	cacaoct	Miscellaneous computation tools
AOloopControl_perfTest	cacaoAOloopControlperfTest	cacaopt	Performance monitoring and testing, including latency measurement
pyramidWFStools	cacaopyramidWFStools	cacaopyrwfs	Pyramid WFS tools
pycacao	Python package		Python AO tools (calibration, DM tools)

Dependencies

All modules link milk’s CLIcore, and all modules except AOloopControl link cacaoAOloopControl. Linear algebra comes from OpenBLAS with LAPACKE, or from Intel MKL when found. computeCalib uses CUDA and MAGMA when milk is configured with -DUSE_CUDA=ON / -DUSE_MAGMA=ON.

Modules also use headers from milk modules they don’t link, and AOloopControl_acquireCalib calls functions in computeCalib without linking it. These symbols are resolved at runtime because milk loads all modules into one process. So a module can’t be used on its own without milk.

Loading a module

milk > mload cacaoAOloopControl
milk > m?