Experimental Physics and
Industrial Control System

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 <2016> 2017 2018 2019 2020 2021 2022 2023 2024	Index	1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 <2016> 2017 2018 2019 2020 2021 2022 2023 2024
<== Date ==>		<== Thread ==>

Subject:	RE: FLOOR function in EPICS CALC error
From:	<[email protected]>
To:	<[email protected]>
Cc:	[email protected]
Date:	Mon, 14 Nov 2016 10:45:49 +0000

Thank you Mark and Tim,

I understand what is happening now, I should have checked the precision set on PREC to see this correctly. With the correct PREC, I see the irrational part now.

Thank you for your time,

Rory

From: Engbretson, Mark S. [mailto:[email protected]]
Sent: 11 November 2016 20:24
To: Clarke, Rory (STFC,DL,TECH)
Subject: Re: FLOOR function in EPICS CALC error

I.e. 0.01 is a endlessly ongoing bit string in base 2. The number can not be represented without lose on a binary computer.

On Nov 11, 2016, at 14:19, Engbretson, Mark S. <[email protected]> wrote:

Wouldn't an easier test of FLOOR be just to use the floor statement with a constant? If it was really broken, order of parens should make little different. since it does, it seems as if the actual value of the calculation is suffering from a precision issue.

On Nov 11, 2016, at 14:08, <[email protected]> <[email protected]> wrote:

Hi,

I am seeing what I think may be a rounding error in the FLOOR function and I was wondering if others had seen this. I replaced the variables in my code with actual values to reproduce the problem. I am using EPICS R3.14.12.5 with Scientific Linux 7. I also tried it with the same version of EPICS running on Linux Mint

In Calc records, just using an empty Calc record and putting the values in by hand:

record(calc,"DLY:FSTEP")

{

field(CALC,"6.1-0.2/0.01")

}

6.1-0.2/0.01 gives 590 as it should

However:

FLOOR((6.1-0.2)/0.01)

Gives me 589

If I write it as follows:

FLOOR((6.1-0.2)*(1/0.01))

I again get 590 (as it should be)