Experimental Physics and
Industrial Control System

Benjamin Franksen <[email protected]> · Tue, 21 Oct 1997 09:45:07 +0200

Hi again,

I fear the aoRecord is 'infected' by the same kind of disease, as one
can see by looking into the init_record() and convert() functions. In
the first case the solution is the same as in the aiRecord, replace 

	val = pai->rval + pai->roff;

in init_record() by

	val = (double)pai->rval + (double)pai->roff;

In the second case (convert()) it goes the other way around: First val
is streched according to eslo as in

	value = (value - pao->egul) / pao->eslo;

Afterwards roff is subtracted:

	pao->rval -= pao->roff;

As in the aiRecord, if egul and eguf are assuming 32 bit raw values,
this can lead to an overflow. Proposed solution: replace the last 4
lines of convert():

	if (value >= 0.0) pao->rval = value + 0.5;
	else	pao->rval = value - 0.5;
	pao->rval -= pao->roff;
	return;

by these three:

	if (value >= 0.0) pao->rval = value - pao->roff + 0.5;
	else	pao->rval = value - pao->roff- 0.5;
	return;

As in the aiRecord, this imposes a slight loss of efficiency (in case of
less than 32 bit raw values) because of the additional conversion of
pao->roff from long to double. Don't know how many nanoseconds (or, more
interesting, how many processor cycles) that takes. I assume one can
afford it in most cases.

Caution: I didn't test the code. Don't trust me blindly.

	Ben
-- 
The Notorious Neb Nesknarf
// snail: BESSY II, Rudower Chaussee 5, D-12489 Berlin, Germany
// email: [email protected]
// phone/fax: +49(30)6392-4865 / 6392-4859

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System