Hi Kustaa,
Thanks for answering ;-)
El lun, 06-04-2009 a las 09:05 +0300, Kustaa Nyholm escribió:
> Hi,
>
> I presume the floating point code has been tested and works,
> in my experience people don't release code that that has
> not been tested, but of course a bug occasionally creeps in...
I have been working with floating point and sdcc for a long time and
everything was ok. The problems started when I used negative floating
points numbers.
> Without running your code it would seem to me that
> the code and your hardware behaves correctly when it
> is showing all LEDs of ie the value is zero.
Yes, the hardware is ok. I have been using this hardware for many years
for conducting the experiment of my dissertation. It was about the
locomotion of modular robots.
The same result are obtained in simulation
> What you are attempting and assuming that -1 + 2 gives 1,
> but this presumes that -1 and 2 are presentable
> *exactly* in 'float' format, which is not the case, I think.
>
> So -1 + 2 is ever so slightly less than 1 which when
> truncated (cast to char) becomes 0 hence all the LEDs are off.
>
> You may want to study floating point math a bit more,
> it is full of pit falls like that, all too easy to
> step into one.
That was the first idea. Maybe there was a truncation error. I have made
some more test. This code works ok: (the complete code is attached)
volatile float test;
void main(void)
{
TRISB=0x00;
test=5.4;
PORTB=test;
while(1);
}
As expected, a number 5 is shown in the leds (also in the simulator)
This second code also works ok:
volatile float test;
void main(void)
{
TRISB=0x00;
test=5.6;
test= 5.6 + (-0.2);
PORTB=test;
while(1);
}
Again, the number 5 is shown in the leds.
But this code DOES not work:
volatile float test;
volatile float test2;
void main(void)
{
TRISB=0x00;
test=5.6;
test2=-0.2;
test= 5.6 + test2;
PORTB=test;
while(1);
}
The value 0x45 is shown in the leds. The only change was to substitute
the -0.2 constant by the test2 variable.
Maybe I am doing something wrong. I would like some guys in the list to
test this examples.
For compiling I am using:
sdcc -Wl-m -Wl-ainhx8m -mpic14 -p16f876a -o test-float3 test-float3.c
This is the sdcc version:
$ sdcc -v
SDCC : mcs51/gbz80/z80/avr/ds390/pic16/pic14/TININative/xa51/ds400/hc08
2.9.0 #5416 (Mar 22 2009) (UNIX)
Thank you very much
Best regards, Juan
>
> br Kusti
>
>
>
>
> > From: Juan Gonzalez Gomez <[email protected]>
> > Reply-To: <[email protected]>
> > Date: Mon, 6 Apr 2009 03:00:07 +0300
> > To: <[email protected]>
> > Conversation: [Sdcc-user] pic14: floating point error
> > Subject: [Sdcc-user] pic14: floating point error
> >
> > Hi,
> >
> > I am trying the sdcc 2.9.0 for the PIC16F876A device. I have found
> > that the floating point operations are not working well when negative
> > numbers are used.
> >
> > The following example works ok:
> >
> > #include <pic16f876a.h>
> > #include <float.h>
> >
> > volatile float test;
> >
> > void main(void)
> > {
> > TRISB=0x00;
> >
> > test=1;
> > PORTB=(unsigned char)(test+2);
> >
> > while(1);
> > }
> >
> > There is an 8-led board connected to the port B. When this program is
> > executed the number 3 (in binary) is shown.
> >
> > But if the test variable is assigned to the number -1, it does not work.
> > The following code does not work. The number 1 should be showed, but all
> > the leds are off
> >
> > #include <pic16f876a.h>
> > #include <float.h>
> >
> > volatile float test;
> >
> > void main(void)
> > {
> > TRISB=0x00;
> >
> > test=-1;
> > PORTB=(unsigned char)(test+2);
> >
> > while(1);
> > }
> >
> > If the test variable is declared as int, it works. So, it seems there
> > is a problems with negative floating points.
> >
> > The program is compiled like this:
> >
> > sdcc -Wl-m -Wl-ainhx8m -mpic14 -p16f876a -o float-error float-error.c
> >
> > Is this a known bug? Has anyone tested the floating point support for
> > the pic14?
> >
> > Thanks
> >
> > Best regards, Juan
> >
> >
> > --
> > Juan Gonzalez Gomez
> > Blog: http://www.iearobotics.com/blog/
> > Web : http://www.iearobotics.com/juan
> >
> >
> > ------------------------------------------------------------------------------
> > _______________________________________________
> > Sdcc-user mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/sdcc-user
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Sdcc-user mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/sdcc-user
--
Juan Gonzalez Gomez
Blog: http://www.iearobotics.com/blog/
Web : http://www.iearobotics.com/juan
#include <pic16f876a.h>
#include <float.h>
volatile float test;
void main(void)
{
TRISB=0x00;
test=5.4;
PORTB=test;
while(1);
}
#include <pic16f876a.h>
#include <float.h>
volatile float test;
void main(void)
{
TRISB=0x00;
test=5.6;
test= 5.6 + (-0.2);
PORTB=test;
while(1);
}
#include <pic16f876a.h>
#include <float.h>
volatile float test;
volatile float test2;
void main(void)
{
TRISB=0x00;
test=5.6;
test2=-0.2;
test= 5.6 + test2;
PORTB=test;
while(1);
}
------------------------------------------------------------------------------
This SF.net email is sponsored by:
High Quality Requirements in a Collaborative Environment.
Download a free trial of Rational Requirements Composer Now!
http://p.sf.net/sfu/www-ibm-com
_______________________________________________
Sdcc-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sdcc-user
