HAL TWOPASS

LinuxCNC 2.5 supports TWOPASS processing of hal configuration files that can help in the modularization and readability of hal files. (Hal files are specified in an LinuxCNC ini file in the HAL stanza as [HAL]HALFILE=filename).

Normally, a set of one or more hal configuration files must use a single, unique loadrt line to load a kernel module that may handle multiple instances of a component. For example, if you use a two input AND gate component (and2) in three different places in your setup, you would need to have a single line somewhere to specify:

resulting in components and2.0, and2.1, and2.2.

Configurations are more readable if you specify with the names=option for components where it is supported, e.g.,:

resulting in components aa,ab,ac.

It can be a maintenance problem to keep track of the components and their names since when you add (or remove) a component, you must find and update the single loadrt directive applicable to the component.

TWOPASS processing is enabled by including an ini file parameter in the [HAL] section:

Where "anystring" can be any non-null string. With this setting, you can have multiple specifications like:

These commands can appear in different HALFILES. The HALFILES are processed in the order of their appearance in the ini file.

The TWOPASS option can be specified with options to add output for debugging (verbose) and to prevent deletion of temporary files (nodelete). The options are separated with commas.

Example:

With TWOPASS processing, all [HAL]HALFILES are first read and multiple appearances of loadrt directives for each module are accumulated. No hal commands are executed in this initial pass.

After the initial pass, the modules are loaded automatically with a number equal to the total number when using the count= option or with all of the individual names specified when using the names= option.

A second pass is then made to execute all of the other hal instructions specified in the HALFILES. The addf commands that associate a component’s functions with thread execution are executed in the order of appearance with other commands during this second pass.

While you can use either the count= or names= options, they are mutually exclusive — only one type can be specified for a given module.

TWOPASS processing is most effective when using the names= option. This option allows you to provide unique names that are mnemonic or otherwise relevant to the configuration. For example, if you use a derivative component to estimate the velocities and accelerations on each (x,y,z) coordinate, using the count= method will give arcane component names like ddt.0, ddt.1, ddt.2, etc.

Alternatively, using the names= option like:

results in components sensibly named xvel,yvel,zvel, xacel,yacel,zacel.

Many comps supplied with the distribution are created with the comp utility and support the names= option. These include the common logic components that are the glue of many hal configurations.

User-created comps that use the comp utility automatically support the names= option as well. In addition to comps generated with the comp utility, numerous other comps support the names=option. Comps that support names= option include: at_pid, encoder, encoder_ratio, pid, siggen, and sim_encoder.

Twopass processing occurs before the loading of a gui. When using a [HAL]POSTGUI_HALFILE, it is convenient to place all the loadrt statements for components needed in a halfile that is loaded earlier.

Example of a HAL section when using a POSTGUI_HALFILE :

Some GUIs support halfiles that are processed after the GUI is started in order to connect hal pins that are created by the GUI. When using a postgui halfile with TWOPASS processing, include all loadrt items for components added by postgui halfiles in a separate halfile that is processed before the GUI. The addf commands can also be included in the file. Example:

Examples of TWOPASS usage for a simulator are included in the directories: