printf("%.2f\n" , 90.625)

[Dario (drinking coffee in the office…)]

The following program

int main(){printf(" %.2f\n", 90.625);}

what it must print out?

I try it on several compiler/operating system
and I obtaing different results.

On some compilers it writes 90.62,
but with other compilers it writes 90.63.

- Dario

[Lew Pitcher]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dario (drinking co
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFAqPpIagV FX4UWr64RApR5AJ 9tUJX+9j50vtrBa Ut7zROkwKbkdACe L+/m
DF4Mo3cFDbqmVAw I4ekW7R8=
=5AwR
-----END PGP SIGNATURE-----

[Lew Pitcher]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Third time's a charm??

Dario wrote:

The following program

int main(){printf(" %.2f\n", 90.625);}

what it must print out?

See the C faq (http:/www.faqs.org/) question 14.1

- --
Lew Pitcher
IT Consultant, Enterprise Application Architecture,
Enterprise Technology Solutions, TD Bank Financial Group

(Opinions expressed are my own, not my employers')
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFAqP72agV FX4UWr64RAvXKAK Cl8sn9Tpr1kM3D8 fsH2zE0gZb56ACd HepF
PE66iwK0pwleh9C kMZBtaoE=
=jVzY
-----END PGP SIGNATURE-----

[Lew Pitcher]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

OK, that was wierd. Thunderbird definitely messed that one up.

I'll try again...

Dario (drinking co
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFAqP6iagV FX4UWr64RAollAK CugSLdkGM+0jzi5 fOVUy+ihKV3lQCg iKMH
WYbGHjocD3vDI9E u1XiLI6w=
=0uhC
-----END PGP SIGNATURE-----

[CBFalconer]

Lew Pitcher wrote:

Dario wrote:

The following program

int main(){printf(" %.2f\n", 90.625);}

what it must print out?

See the C faq (http:/www.faqs.org/) question 14.1

The binary value is: 1011010.101 with . representing the binary
point. The fractional portion represents 5/8. The whole thing
only needs 10 significant bits. So the OPs question is about how
his printf interpreter does its precision rounding, which is a
matter for the particular implementation. This is almost
certainly outside the FP implementation.

I don't know just where this dog buried its bone.

--
Chuck F (cb********@yah oo.com) (cb********@wor ldnet.att.net)
Available for consulting/temporary embedded and systems.
<http://cbfalconer.home .att.net> USE worldnet address!

[Dario (drinking coffee in the office…)]

Lew Pitcher wrote:

Dario wrote:

The following program

int main(){printf(" %.2f\n", 90.625);}

what it must print out?

See the C faq (http:/www.faqs.org/) question 14.1

Question 14.1 is not related to my question.
The number 90.625 is exactly represented with a double value.

My question was:
Why printf("%.2f\n" , 90.625) prints 90.62 or 90.63
depending on compiler ?

Using GCC it writes 90.62.
Using Microsoft compiler it writes 90.63.

What is the correct result?

- Dario

[CBFalconer]

"Dario (drinking co?ee in the o?ce…)" wrote:

.... snip ...
My question was:
Why printf("%.2f\n" , 90.625) prints 90.62 or 90.63
depending on compiler ?

Using GCC it writes 90.62.
Using Microsoft compiler it writes 90.63.

What is the correct result?

I already answered this. However, my guess is that Microsoft
lazily does rounding up, while what ever library you are using for
gcc rounds to even.

--
Chuck F (cb********@yah oo.com) (cb********@wor ldnet.att.net)
Available for consulting/temporary embedded and systems.
<http://cbfalconer.home .att.net> USE worldnet address!

[Dan Pop]

In <L9************ ********@news20 .bellglobal.com > Lew Pitcher <Le*********@td .com> writes:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Could you, please, drop this junk from your posts. It's annoying...
Dario wrote:

The following program

int main(){printf(" %.2f\n", 90.625);}

what it must print out?

Let's see what the standard says:

The value is rounded to the appropriate number of digits.

But it doesn't provide any rounding algorithm. So, you can realistically
expect both 90.62 and 90.63, depending on the implementor's preference.

OTOH, calling printf with no prototype declaration in scope is undefined
behaviour in C89, so the program in question could print *anything*.
See the C faq (http:/www.faqs.org/) question 14.1

Bogus advice: 0.625 = 1/2 + 1/8 and all "small" integers are exactly
represented in (binary) floating point.

Dan
--
Dan Pop
DESY Zeuthen, RZ group
Email: Da*****@ifh.de

[Dario (drinking coffee in the office…)]

Dan Pop wrote:

In <L9************ ********@news20 .bellglobal.com > Lew Pitcher <Le*********@td .com> writes:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Could you, please, drop this junk from your posts. It's annoying...

Seconded.

Dario wrote:

The following program

int main(){printf(" %.2f\n", 90.625);}

what it must print out?

Let's see what the standard says:

The value is rounded to the appropriate number of digits.

But it doesn't provide any rounding algorithm. So, you can realistically
expect both 90.62 and 90.63, depending on the implementor's preference.

Just to clarify:
I can expect both 90.56 and 90.57 even from the following program?

#include <stdio.h>
int main(int argc, char*argv[])
{
printf("%.2f\n" , 90.5625);
return 0;
}
OTOH, calling printf with no prototype declaration in scope is undefined
behaviour in C89, so the program in question could print *anything*.

Obviously!
In my original post I over-simplified the example
just to demonstrate my problem.

- Dario

[Dan Pop]

In <c8**********@b alena.cs.interb usiness.it> =?UTF-8?B?IkRhcmlvICh kcmlua2luZyBjb+ +sgGVlIGluIHRoZ SBv76yDY2XigKYp Ig==?= <da***@despamme d.com> writes:

Just to clarify:
I can expect both 90.56 and 90.57 even from the following program?

#include <stdio.h>
int main(int argc, char*argv[])
{
printf("%.2f\n" , 90.5625);
return 0;
}

I think so, but QoI (quality of implementation) considerations would
effectively rule out 90.57.

Dan
--
Dan Pop
DESY Zeuthen, RZ group
Email: Da*****@ifh.de