The example above depicts a navigation process running on one type of processor communicating its results to a variety of other types of processors via a variety of data buses. The nav process produces a solution containing floating point values (e.g., longitude, direction cosines, velocity components, etc.). On its originating processor, these values exist in the native format of the underlying hardware. If transmitted intact, they would be gibberish on a different processor.
GHOST provides utilities to encode floating point values into a Global Format which can be transmitted to any other processor, where it can be converted back into a floating point value in the format of its underlying hardware so it can be used by applications running there.
These conversion utilities perform some simple algebra to break the original value into integers which maintain the original precision of the value. For example, ENCODE_RADIANS splits a floating point value with 6 digits of accuracy (a common floating point precision) into integer minutes and integer hundredths of seconds. 0.000001 radians (i.e., 5.72958e-5 degrees, or 0 degrees,, 0 minutes, 0.21 seconds) would therefore be passed as the pair 0 and 21 (0 minutes, 21 hundredths of seconds). When recombined, the original value arrives intact.
The overhead associated with such encoding and decoding is minimal at worst. Numerous attempts to quantify the overhead incurred during a normal messaging cycle have resulted in no measurable time loss due to this extra processing. In comparision to the transport delays and bus arbitrations, the overhead you pay for this is infintesimal.
Therefore, using GHOST, you get true heterogeneous processing with no measurable degradation in run-time performace.