So it looks like the two overflow modes (OVM=1 forcing saturation on each add/subtract, OVM=0 uses an overflow counter on add/subtract, with explicit SAT ACC to saturate) only pay attention to add/subtract operations. Is there any way to get them to work with shifts or multiplication?
At one point in some assembly-language computation, I need to apply a constant gain factor to a 32-bit number in the accumulator, before extracting a 16-bit portion to an output variable, and need it not to overflow. The only way I can seem to apply any multiplicative factor is to do:
SAT ACC
ADDL ACC, @ACC
SAT ACC
ADDL ACC, @ACC
SAT ACC
or
SAT ACC
SETC OVM
ADDL ACC, @ACC
ADDL ACC, @ACC
CLRC OVM
(e.g. add accumulator to itself multiple times, and I have to saturate first before the first one of these so that the accumulator is not overflowed, e.g. 0x7fffffff + 2 needs to be interpreted as a positive number but it shows up in the accumulator prior to saturation as a negative number 0x80000001)
It would have been nice to have an ASL (arithmetic shift left) operation in addition to LSL, such that for ASL the overflow/saturation logic were obeyed.
