The DSPs, for example, can compute fixed-point math
at less power/MHz than GPUs using floating-point math.
The fixed-point math
requires programmers to pay significant attention to the number of coefficients used in each algorithm when multiplying and accumulating digital data to prevent distortion caused by register overflow and a decrease of the signal-to-noise ratio caused by truncation noise.
The change from implementing a control algorithm in floating-point to fixed-point math can take a great amount of time, and overflows and underflows inherent in fixed-point operations can wreak havoc with control loops.
A customer in Europe had a PID loop that we converted to fixed-point math in about 10 minutes.
Our Blocksets support high-precision fixed-point math
and a floating-point override simulation mode so developers don't get bogged down with quantization, underflow and overflow as they start to create algorithms rithms.
In some cases, even double-precision fixed-point math