%16 with double

can't i mod a double ?

Code:

---

#include<stdio.h>

int main()
{
        double a=100;
        printf("%d",a%16);
        return 0;
}

---

and i get:

Code:

---

root@cactus:~/PL# gcc -o mod mod.c
mod.c: In function ‘main’:
mod.c:6: error: invalid operands to binary %

---

Evidently not. modulus only makes sense for integer divisions. You could cast it to an int first.

Quote:

---

Originally Posted by spank

can't i mod a double ?

Code:

---

#include<stdio.h>

int main()
{
        double a=100;
        printf("%d",a%16);
        return 0;
}

---

and i get:

Code:

---

root@cactus:~/PL# gcc -o mod mod.c
mod.c: In function ‘main’:
mod.c:6: error: invalid operands to binary %

---

---

Isn't a double a floating point? I'm a C total n00b so I might be wrong but as far as I am aware...

1) When you assign a value to a float, you have to add a decimal point to the assigned value.
2) The signifier for a double is %lf. You have put in %d which I believe is for int
3) I think I read somewhere that double is not always available and it depends on the system you are using.

Could it be that one of these things is causing the problem? Maybe someone with more experience can say for sure. In my short time here, people have been incredibly helpful and knowledgable, I'm sure someone will be able to point out the problem.

Hope this helps,
TV

Code:

---

SYNOPSIS
      #include <math.h>

      double fmod(double x, double y);

      float fmodf(float x, float y);

      long double fmodl(long double x, long double y);

DESCRIPTION
      The  fmod()  function  computes the remainder of dividing x by y.  The return value is x - n * y,
      where n is the quotient of x / y, rounded towards zero to an integer.

---

Quote:

---

Originally Posted by tvsinesperanto

Isn't a double a floating point? I'm a C total n00b so I might be wrong but as far as I am aware...

1) When you assign a value to a float, you have to add a decimal point to the assigned value.

---

Not really, you need a decimal point for an expression like float f = 1/2.0; because otherwise 1/2 is integer and evaluates to 0 instead of the desired 0.5. Something like float f = 2; is fine. However, it would be good practice when initialising a floating point type to add the .0 for clarity.

Quote:

---

2) The signifier for a double is %lf. You have put in %d which I believe is for int

---

Actually with printf, doubles use %f. %lf is only valid for scanf. This is because floats and doubles both become doubles when passed to printf.

Quote:

---

3) I think I read somewhere that double is not always available and it depends on the system you are using.

---

Any conforming C compiler will have the double type.

Quote:

---

Originally Posted by cwr

Not really, you need a decimal point for an expression like float f = 1/2.0; because otherwise 1/2 is integer and evaluates to 0 instead of the desired 0.5. Something like float f = 2; is fine. However, it would be good practice when initialising a floating point type to add the .0 for clarity.

---

Ah, OK. My text book says that you "should always put in the decimal point". I took that to mean that it was required but I see now that it could be read as a suggestion. Thanks for putting me straight on that.

Quote:

---

Originally Posted by cwr

Actually with printf, doubles use %f. %lf is only valid for scanf. This is because floats and doubles both become doubles when passed to printf.

---

Ok, so does that mean that, when using a double in a printf() function, you must use %f and can't use %lf or that it makes no difference whether you use %lf or %f?

Referring, once again, to my text book, it seems to imply that you should use %lf. It actually says...

"These [referring to a table below that contains a list of signifiers and their corresponding types in which %lf is matched to double] are used in all *f() functions: printf(), scanf(), fprintf(), sscanf(), sprintf(), and so forth."*

* Pg 368 "C Programming" by Larry Ullman and Marc Liyanage

So, is the text incorrect then? I am starting to have concerns about my text book if this is the case.

Quote:

---

Originally Posted by cwr

Any conforming C compiler will have the double type.

---

This was actually my mistake this time, I simply misread the book. It actually says that a long double is not always available. Sorry about that.

Sorry to hijack the thread like this but I am concerned that my text book may be steering me wrong in places. Should I start a new thread perhaps?

Quote:

---

Ok, so does that mean that, when using a double in a printf() function, you must use %f and can't use %lf or that it makes no difference whether you use %lf or %f?

---

The former: you must use %f. %lf is invalid for printf(). (But you must use %lf for scanf(); %f in scanf() reads a float instead of a double.)

Your book is a bit misleading. Which one is it?
[edit]
Never mind, I see you posted its title.
[/edit]

Quote:

---

It actually says that a long double is not always available.

---

long double should be around on newer compilers, like Dev-C++. It's part of the C99 standard.

Quote:

---

Originally Posted by dwks

The former: you must use %f. %lf is invalid for printf(). (But you must use %lf for scanf(); %f in scanf() reads a float instead of a double.)

Your book is a bit misleading. Which one is it?
[edit]
Never mind, I see you posted its title.
[/edit]

---

Yeah, I am really impressed with the layout and the way it explains things but I'm now quite concerned that it may contain erorrs.

Oh well, I guess that is what forums are for...

Thanks for the help.

Cheers,
TV

Quote:

---

This is because floats and doubles both become doubles when passed to printf.

---

That actually applies to any function whever passing a float by value. The compiler expands a float to double, push the double on the stack, then in the receiving function that declares it as a float in the parameter list, the compiler has to shrink it back to a float. Programs that may be speed critical should just use doubles everywhere to avoid all those conversions.

Quote:

---

Originally Posted by Ancient Dragon

That actually applies to any function whever passing a float by value. The compiler expands a float to double, push the double on the stack, then in the receiving function that declares it as a float in the parameter list, the compiler has to shrink it back to a float. Programs that may be speed critical should just use doubles everywhere to avoid all those conversions.

---

I'd be interested in more information on this. Some of my googling has led me to things such as "default argument promotions" and "when a function is called with no prototype in scope" and stuff about K&R1 vs C89. I'll continue digging, but if you have quickie links I would be grateful.

Quote:

---

Originally Posted by Ancient Dragon

That actually applies to any function whever passing a float by value. The compiler expands a float to double, push the double on the stack, then in the receiving function that declares it as a float in the parameter list, the compiler has to shrink it back to a float. Programs that may be speed critical should just use doubles everywhere to avoid all those conversions.

---

Are you talking about the specific behaviour of one architecture, or are you saying this happens generally?

Edit: Followup:

I made this simple program below that calls a function taking a float many many times...

Code:

---

#include <stdio.h>
float foo(float f)
{
    return f+1.0;
}
int main(void)
{
    int i;
    float f = 0;
    for (i = 0; i < 100000000; i++)
        f = foo(f);
    printf("%f", foo);
    return 0;
}

---

I then timed it:

Code:

---

$ gcc -o float float.c
$ for ((i=0; i<5; i++)); do time ./float; done 2>&1 | grep real
real    0m1.110s
real    0m1.098s
real    0m1.097s
real    0m1.108s
real    0m1.097s

---

I then replaced the above code to use doubles everywhere instead of float:

Code:

---

$ cat float.c | sed 's/float/double/g' > double.c
$ gcc -o double double.c
$ for ((i=0; i<5; i++)); do time ./double; done 2>&1 | grep real
real    0m4.235s
real    0m4.325s
real    0m4.390s
real    0m4.301s
real    0m4.270s

---

Here we see double is much slower. Let's try with -O2:

Code:

---

$ gcc -O2 -o float float.c
$ for ((i=0; i<5; i++)); do time ./float; done 2>&1 | grep real
real    0m0.931s
real    0m0.923s
real    0m0.933s
real    0m0.923s
real    0m0.933s
$ gcc -O2 -o double double.c
$ for ((i=0; i<5; i++)); do time ./double; done 2>&1 | grep real
real    0m0.934s
real    0m0.922s
real    0m0.935s
real    0m0.922s
real    0m0.933s

---

The timing is now basically identical. I'm not sure there is any significant overhead for function calls with double versus float?

Oh, the above timings are on a Pentium 4 3.0Ghz using gcc 3.3 under linux 2.6. I did have a look at the assembly output, but my x86 asm is a little rusty, so I couldn't really comment constructively.

Update:

Here are the same timings with -O2 using gcc 4.0 and a PowerPC (ppc) G4 1.0GHz under MacOS X 10.4:

Code:

---

$ for ((i=0; i<5; i++)); do time ./float ; done 2>&1 | grep real
real    0m4.170s
real    0m4.212s
real    0m4.071s
real    0m4.187s
real    0m4.096s
$ for ((i=0; i<5; i++)); do time ./double ; done 2>&1 | grep real
real    0m2.754s
real    0m3.216s
real    0m3.023s
real    0m2.797s
real    0m2.762s

---

Here we see double is indeed faster, but this may also be to do with the add operation, not just the function call overhead.