Slide 17 of 67
Notes:
Having two accumulators is a benefit in many instances.
This enables a non-pipelined multiply and add. The MAC unit is not just a MAC unit. It's a multiplier and an adder. We can have two operands go into the multiplier and the output go to the accumulator, we can have the output of the multiplier go to the adder, or we can have operands go into the adder without going through the multiplier. So it's a non-pipeline multiply and add, which provides significant flexibility in how you use the device.
The C54x supports both signed and unsigned operations. We can have fractional and integer modes. It features 2’s complement rounding, and overflow detection. In addition, the C54x accumulators are 40 bits, which means that if overflow is detected on 32 bit results, we can correct for saturation.
Another important feature is that the A accumulator can be an input to the multiplier. These two accumulators are considered to be identical accumulators, with one difference. The output of the A accumulator is allowed to be an input to the multiplier, but an output of the B accumulator may not be an input to the multiplier.
