How do you round floats in C, using VisualDSP++?

Question

Most C compilers offer some variation of the Standard C Library. The VisualDSP++ compiler is no different. However, I've recently encountered a scenario where I need to take a float value, round it to the nearest whole number and then convert it to an int.

This is rather easy using most variations of C, but after attempting to use roundf, I've found that math.h does not include the roundf function.

How can I correctly round my float variables to the nearest int value with VisualDSP++?

For those who are unaware of VisualDSP, it is an embedded systems development IDE an compiler. — RLH, Jun 19 '13 at 18:34
Would the strategy in [this answer](http://stackoverflow.com/a/497079/7116) work for you? — user7116, Jun 19 '13 at 18:37

score 4 · Accepted Answer · 2013-08-04T20:23:56.783

4

I am not familiar with VisualDSP, but you can round a float to an integer like so:

float number = 3.14;
int rounded = (int)(number + (number > 0 ? 1 : -1) * 0.5);

So, you could write your own function for this:

int roundFloatToInt(float number)
{
    return (int)(number + (number > 0 ? 1 : -1) * 0.5);
}

The above outputs the following for the given input:

roundFloatToInt(0.10) --> 0

roundFloatToInt(3.14) --> 3

roundFloatToInt(3.54) --> 4

It appears to work in all the cases I tested, including negatives.

edited Aug 04 '13 at 20:23

answered Jun 19 '13 at 18:37

1

I wouldn't name it `round`, since that's the name of a standard function in C99. Re-using the name could hurt the portability of the code. – Keith Thompson Jun 19 '13 at 18:44
And added support for negative numbers. – Jun 19 '13 at 18:56

Keith Thompson · Answer 2 · 2013-06-19T18:54:55.863

1

The *f() functions, including roundf, were added in, as I recall, the 1999 version of the ISO C standard, along with the roundd function that operates on long doubles.

It's ~~likely~~ possible that your <math.h> declares round, which operates on double rather than float. (C90 actually didn't have any of round, roundf, or roundl, but your implementation might provide round as an extension.) (EDIT: No, it doesn't.)

Untested code follows:

float x = 123.456;
int n = round(x);

round takes a double argument; the float argument passed to it will be implicitly converted. The double result is then implicitly converted to int.

You could make all the conversions explicit by adding casts:

int n = (int)round((double)x);

but that's just clutter.

EDIT: C89/C90 did define the floor and ceil functions. If round isn't available, it shouldn't be difficult to define it in terms of floor. The details depend on exactly what kind of rounding you want to do.

edited Jun 19 '13 at 18:54

answered Jun 19 '13 at 18:38

Keith Thompson

254,901
44
429
631

Sorry, I should have explained my thoroughness a bit better. I did a search throughout the entire VisualDSP++ application directory, which includes the include files and all internal C files. "round" doesn't exist as a function name or as part of one within the entire compilers internal library. – RLH Jun 19 '13 at 18:47
Regarding your EDIT post. There is a chance that this is a C89/90 compiler. I inherited this project and since my primary background is Windows development, I'm use to the "conveniences" of more modern compilers that are kept up-to-date. Visual C++ 6.0 is probably more current that VisualDSP++. – RLH Jun 19 '13 at 18:49

How do you round floats in C, using VisualDSP++?

2 Answers2