Hi all,

I have a query regarding vector weighted sum.equation to represent this is
Z[i] = W * X[i] + y[i]. what is the MAX or Min value of W = weighting factor ???
If I use  multiply and accumulate instruction then what is the default mode selected and
if the inputs are fractional what should be my Qformat?? I have to maintain same Qformat for
all??

Thanks in advance

Regards
Ravi chandran

On Fri, 4 May 2007, ravichandran wrote:

> Hi all,
>
> I have a query regarding vector weighted sum.equation to represent this is
> Z[i] = W * X[i] + y[i]. what is the MAX or Min value of W = weighting factor ???
> If I use  multiply and accumulate instruction then what is the default mode selected and
> if the inputs are fractional what should be my Qformat?? I have to maintain same Qformat
for all??

Howdy Ravi,

Depends on the processor.  I assume you are using a blackfin with integer
mac, but if you have sharc or tigersharc then it could be floating point
too.

You can scale between every multiply if needed, so you don't have to
maintain the same Q format.  That really depends on the full algorithm and
how many steps you'll have and the input data.  If possible, simulate the
problem on a PC and look at intermediate values to see where problems
happen, then build in scale factors to reduce those problems.

A min of zero will eliminate that term, a max of "infinity" (i.e. largest
possible number you can represent) wipes you out.  Anything in between is
possible, and probably more useful!

Patience, persistence, truth,
Dr. mike