forums.silverfrost.com Forum Index forums.silverfrost.com
Welcome to the Silverfrost forums
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Math Errors?

 
Post new topic   Reply to topic    forums.silverfrost.com Forum Index -> Support
View previous topic :: View next topic  
Author Message
jetcheve



Joined: 23 Aug 2006
Posts: 23
PostPosted: Wed Sep 06, 2006 9:47 am    Post subject: Math Errors? Reply with quote
I have a piece of software originally written as a standalone program that I have turned into a dll using FTN95. It now recieves variables from a namelist file created by visual basic, and runs some computations. The standalone program is fine, it's been working for years. The variables pass in correctly, I've tested that. However, when the computations start, I'm getting ridiculous math errors in equations that I never changed, ones that I literally copied and pasted, save for some syntax changes such as line continuations. Take this block of code for example:

Code:
Do 90, INC = 1, INC1
RAD = DEG / 57.296

MAINVALS(INDEX,1) = DEG
PINVALS(INDEX,1) = DEG
XHVALS(INDEX,1) = DEG
SHOEVALS(INDEX,1) = DEG

SN = R * SIN(RAD)
CO = SQRT((L ** 2 - (R * SIN(RAD)) ** 2))
Call SUB530(ANG, SN, CO)
FEE = ANG
FT1 = NRL1(INC) * SIN(RAD) * (1 + COS(RAD) / ((L / R) ** 2 - SIN(RAD) ** 2) ** .5)
FT2 = NRL2(INC) * SIN(RAD) * (1 + COS(RAD) / ((L / R) ** 2 - SIN(RAD) ** 2) ** .5)
FR1 = NRL1(INC) * (COS(RAD) - (SIN(RAD)) ** 2 / ((L / R) ** 2 - SIN(RAD) ** 2) ** .5) - 0.0000284 * R * N ** 2 * WRT
FR2 = NRL2(INC) * (COS(RAD) - (SIN(RAD)) ** 2 / ((L / R) ** 2 - SIN(RAD) ** 2) ** .5) - 0.0000284 * R * N ** 2 * WRT
FRC1 = NRL1(INC) * (COS(RAD) - SIN(RAD) ** 2 / ((L / R) ** 2 - SIN(RAD) ** 2) ** .5) - 0.0000284 * R * N * N * (WRT - CP)
FRC2 = NRL2(INC) * (COS(RAD) - SIN(RAD) ** 2 / ((L / R) ** 2 - SIN(RAD) ** 2) ** .5) - 0.0000284 * R * N * N * (WRT - CP)
RA = (90.0 - FEE - DEG) / 57.296
FC1Y = -FRC1 * COS(RA) + FT1 * SIN(RA)
FC1X = -FRC1 * SIN(RA) - FT1 * COS(RA)
FC1 = SQRT((FC1X ** 2 + FC1Y ** 2))
FC2Y = -FRC2 * COS(RA) + FT2 * SIN(RA)
FC2X = -FRC2 * SIN(RA) - FT2 * COS(RA)
FC2 = SQRT((FC2X ** 2 + FC2Y ** 2))
SN = FC1Y
CO = FC1X



Starting with RAD = 0 or DEG = 90, obviously sin and cosine should be 0 and 1 respectively. Thus, as you'd expect, FT1 and FT2 are also 0. Now comes the first error... When FR1 is calculated using NRL1(1) = 15467 and rad still 0, you should get -15470.3 according to my hand calculations, maybe I made a mistake maybe I didn't, there are still others, but either way, the computer returns -18404.2. However, the most interesting one I have found is at the point where it calculate FC1Y or FC1X etc... Upon performing this calculation the first time, you have RA = 90 - FEE - DEG ( take my word that FEE and DEG are both, and correctly so, 90) and thus equal to 0. thus RA is PI/2 and cos is 0 and sin is 1. Therefore, since FT1 is 0 from earlier, and cos is 0, FC1Y should be -FRC1*0 + 0 * 1 which is 0. But to my surprise, I've discovered that 0+0 is now .107879. Now, if this were a smaller error, I could attribute it to rounding error, truncation, etc... But it's a fairly significant difference, and it's by far not the only occurance of it. What
Back to top
View user's profile Send private message
jetcheve



Joined: 23 Aug 2006
Posts: 23
PostPosted: Wed Sep 06, 2006 10:00 am    Post subject: Math Errors? Reply with quote
do you think is going on?
Back to top
View user's profile Send private message
brucebowler
Guest
PostPosted: Wed Sep 06, 2006 2:14 pm    Post subject: Math Errors? Reply with quote
A couple questions...

What do the declarations of everything look like? Do you have an IMPLICIT NONE in there to make sure you've type'd all variables?

Using either the debugger or "diagnostic writes", where does the calculation start differing from what you calculate by hand?

Why do you use **0.5 rather than sqrt()? They may not do *exactly* the same thing (Paul can probably answer that better than I can)

Back to top
JohnHorspool



Joined: 26 Sep 2005
Posts: 270
Location: Gloucestershire UK
PostPosted: Thu Sep 07, 2006 1:24 am    Post subject: Math Errors? Reply with quote
This may be trivial, but like Bruce's comment on **0.5 I always thought that it was better programming practice to also avoid using **2 as well
_________________
John Horspool
Roshaz Software Ltd.
Gloucestershire
Back to top
View user's profile Send private message Visit poster's website
PaulLaidler
Site Admin


Joined: 21 Feb 2005
Posts: 7927
Location: Salford, UK
PostPosted: Thu Sep 07, 2006 2:38 am    Post subject: Math Errors? Reply with quote
John

Have you tested the code using /CHECKMATE?

Is it possible that line continuations have gone wrong?
For example, with fixed format Fortran it is possible to have a valid equation but part of it is actually commented out.

If you can post a very short working example of the problem then fault should be easy to find.
Back to top
View user's profile Send private message AIM Address
Display posts from previous:   
Post new topic   Reply to topic    forums.silverfrost.com Forum Index -> Support All times are GMT + 1 Hour
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group