best round function?

[sail3005]

right now, i am using this function to round double/float numbers:

int round(double num)
{
    return num + 0.5f;
} 

how could i get it so it would take another paramater, that would round it to a specific decimal place? and, is this a good way to accomplish this function?

[abdul]

use "floor" in math.h header file to round a double or float variable like:

cout<<floor(3.4); 

[sail3005]

i can't use math.h, or any other librarys.

[Vlatko]

i can't use math.h, or any other librarys.

Why Not?

[Wynd]

We're trying to make our own math library, and we don't want any dependency on the "official" one.

[sail3005]

yeah, so far it's going pretty well.

Anyone have any ideas on how to get a second paramater, that will allow it to roubd to a specific decimal place?

[kedaman]

Have a look @ http://www.vbforums.com/showthread.p...ighlight=radix

[sail3005]

thanks kedaman, thats just what i wanted.

but, i have a question...i put in this code:

	double z; 
	double x = 2.229;
	int r=10,d=2; 
	double y=(int)(x*(z=pow(r,d)))/z;
	cout<<y<<endl;

that code returns "2.22" but shouldn't it produce 2.23?

one more question...what exactly does the 'r' variable control. i know you said it was the radix, but what does it do?

[kedaman]

#1 it returns the floor (round down)
radix is the base of a numbering system, for decimal it is 10, for binary it is 2, etc...
to round up from ceil(radix/2) you add 1/2z in advance

[sail3005]

add (1/2)z to what?

[kedaman]

x

[sail3005]

sorry, but this just doesn't seem to be working...now it returns 0.02??

	double z; 
	double x = 2.229;
	int r=10,d=2; 
	double y=(int)(x+((1/2)*z)*(z=pow(r,d)))/z;
	cout<<y<<endl;

[sail3005]

i must not be adding it in the right place???

[sail3005]

*bump*

[kedaman]

    double z; 
    double x = 2.229;
    int r=10,d=2; 
    double y=(int)((z=pow(r,d))*(x+(1/(2*z))))/z;
    cout<<y<<endl; 

1/2z not z/2
btw you have to swap the factors since first term evaluates first, and z is in the second - will cause the the addition to be in vain.

[sail3005]

ahhh...i see. thanks a lot for all your help!

[jim mcnamara]

For what it's worth - pow() is in math.h

On PC's, before the math coprocessor (80286), you had to have
integer arithmetic libraries and floating point libraries. You still have to have them, but they are much easier to build.

What you guys are trying to implement requires two spearate libraries - one for integers, another for floating point.

Originally, in very early CPU's, what you had as native operations were integer add, subtract, plus: carrry, negative, and oveflow flags.

Now CPU's support divison & multiplication (FP and integer) as well as other primitve operations.

What I'm trying to say:
if you want to build math libraries you have to start with the four basic operations (+ - / *) and derive everything else - like with the sqrt() thing in another thread. You can't just use
pow() without writing a pow() function first.

[kedaman]

Here's something i wrote for integer powers:
http://www.vbforums.com/showthread.p...wer#post531771

[sail3005]

Originally posted by jim mcnamara
For what it's worth - pow() is in math.h

On PC's, before the math coprocessor (80286), you had to have
integer arithmetic libraries and floating point libraries. You still have to have them, but they are much easier to build.

What you guys are trying to implement requires two spearate libraries - one for integers, another for floating point.

Originally, in very early CPU's, what you had as native operations were integer add, subtract, plus: carrry, negative, and oveflow flags.

Now CPU's support divison & multiplication (FP and integer) as well as other primitve operations.

What I'm trying to say:
if you want to build math libraries you have to start with the four basic operations (+ - / *) and derive everything else - like with the sqrt() thing in another thread. You can't just use
pow() without writing a pow() function first.

Yeah, i know. we already have our own basic pow funtion developed, so we didn't need math.h.

"What you guys are trying to implement requires two spearate libraries - one for integers, another for floating point. "

I was gonna get around that by using templates. would that work?

[sail3005]

Kedaman, whats the difference between this:

double power(int num, int power)
{
double x = 1;
for (int i = 1; i <= power; i++)
x = x * num;
return x;
}

and your complicated code:

if (power){
mask=256;//or higher max power
for(;!(mask&power);mask>>1); // or something that finds the highest set bit
y=x;
mask>>1;
for (;mask; ){
if (mask&power)
y*=x*y;
else
y*=y;
mask>>1;
}
else
y=1;
}

does yours allow for x^(1/2), x^(1/3) etc.?

[kedaman]

That's what i'm trying to explain here:
http://www.vbforums.com/showthread.p...wer#post533384
My algoritm will run faster, typically logn multiplications while yours will run n multiplications.

[abdul]

I have seen that Kedaman posts pretty complicated code. Maybe, he find it more interesting to confuse people

[sail3005]

nah, i don't think he is trying to confuse people. He has been helping me a lot lately, and i actually find it fun trying to figure out what the heck he is talking about. I learn a lot more inadvertantly when trying to figure out the stuff. He really knows what he is talking about. thanks kedaman .

[jim mcnamara]

Abdul -

Kedaman is definitely not trying to mess people up. Just like you, he is writing in a language that isn't his native tongue.
That makes it harder all by itself. Plus, he tends to abstract algorithms very rapidly.

Sail -
yes, you can use templates to get around the float vs integer problem. The only downside is that some integer solutions are simpler algorithms than their float cousins, and some operations are not defined or don't work the same; roots & modulus arithmetic come to mind.

Just to mention it - the C99 standard defines complex arithmetic operations. For everything you can do to reals, you can now do to complex numbers: ccos, csin, carccos, cabs, etc.

VC++ implements the complex class instead.

[kedaman]

My intentions are the opposite Abdul, I try to make people understand things. I do tend to generalise and abstract things which often can be confusing, but once you understand the generalisation you will be a lot better off than with the details alone.
Everyone -
If something confuses you even a bit, then i'd be glad to know, since it will both help you and me

You can use template specifications to narrow the definitions for specific types, say you want to have modulus for integers as well as floating points:

template<class T>
T modulus(T x, T y){}; //default case

int modulus<int>(int x,int y){ //modulus for integers
    return x%y;
};

float modulus<float>(float x,float y){
    //do a floating point modulus
}; 

I don't happen to know how a floating point modulus is done though Just explaining how you can define different per type for a definition. If T is something else, it won't compile because the default specification template doesn't return anything. The compiler knows how to narrow; if you specify a template it will automatically search the most specified template which fits the expression.

[sail3005]

Cool, thanks. Man, thank god for templates, or i would have a LOT of extra code!!