5.9 Using Probability Variables

The ‘prob’ keyword in an assembler program defines a probability variable with a specific name. A jprob or case instruction can contain the name of a probability variable instead of a numeric probability value. There are three supported types of probability variables:

• Controlled variables. They are the means of modifying a probability profile loaded into a node. A single controlled variable can occur in multiple jprob and case instructions.
• Output variables. They are the means of fetching probabilities learned while executing a node. A single output variable occurs in a single jprob or case instruction.
• Auxiliary variables. They act as constants. The assembler merely replaces every occurrence of the name of an auxiliary variable with the value of this variable specified using the ‘prob’ keyword.

A controlled probability variable can simultaneously be an output probability variable.

An output probabilities array is the means of fetching probabilities learned for case instructions in a choice instruction block by their indices. A location label assigned to the choice instruction block is the name of this array.