Missing round_to() function in PPP_support.h

Question

I'm currently working through Bjarne Stroustrup's Programming: Principles and Practice Using C++ (3rd Edition).

The book mentions a function named round_to() is provided in PPP_support.h

When, on the other hand, we want rounding, we can use round_to() from PPP_support. For example:
void conv2(double y)
{
    int x = round_to<int>(y); // 4/5 rounding
}

The file provided on Stroustrup's website seems to be missing the definition for round_to().

Does anyone know the official definition for that?

round_to was never a standard function and it seems it's not in PPP_support.h (anymore at least). You could use std::round(y); from <cmath> instead. You may get a warning so you could do an explicit cast: int x = static_cast<int>(std::round(y));. You could also define your own round_to like template<class To, class From> To round_to(From y) { return static_cast<To>(std::round(y)); } — Ted Lyngmo
– Ted Lyngmo, Commented Oct 21 at 11:32

prapin · Accepted Answer · 2025-10-21 12:45:16Z

2

Function round_to is not standard and I don't know how it is actually implemented.

In our code base, we use floor_cast, ceil_cast and round_cast that probably are very similar to the mentioned round_to.

To convert floating point numbers to integers, we should use a template function like those. Compared to raw static_cast<int>(std::round(f)), round_cast<int>(f) has the following advantages:

Much faster, standard round is typically rather slow
No undefined behavior due to integer overflow
Cleaner
Easier to search in code base

// Adapted from: https://stackoverflow.com/a/30308919
#include <concepts>
template<std::integral I, std::floating_point F> [[nodiscard]] I floor_cast(F value)
{
    int64_t w = static_cast<int64_t>(value);
    return static_cast<I>(w - (value < w));
}
template<std::integral I, std::floating_point F> [[nodiscard]] I ceil_cast(F value)
{
    int64_t w = static_cast<int64_t>(value);
    return static_cast<I>(w + (value > w));
}
template<std::integral I, std::floating_point F> [[nodiscard]] I round_cast(F value)
{
    return floor_cast<I>(value + 0.5);
}

answered Oct 21 at 12:45

prapin

7,1395 gold badges32 silver badges59 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Ted Lyngmo Oct 22 at 18:32

Interesting that you get faster versions than std::floor, std::ceil and std::round. Do your versions also give the same result as those in the standard library (with a static_cast<I> added)?

prapin Oct 22 at 18:42

For floor and ceil, I am pretty sure the result will be identical for all numerical values inside [-2^54, 2^54]. For round, my version always rounds up for *.5, while most implementations of std::round will do round-to-even. I think this only matters in very rare cases, like precise statistics.

Ted Lyngmo Oct 22 at 21:10

Ok, but how did you come to the conclusion that your versions of floor and ceil are faster? I tried your floor_cast in quick-bench against the libstdc++ version and they seem to be equally fast. The standard version also seems "cleaner" i.m.o. If one uses libc++ instead, the standard version is actually quite a bit faster than your version: quick-bench.com/q/VwaTAKcHgIz_AeiFXVn0Z1Vd99M

Collectives™ on Stack Overflow

Missing round_to() function in PPP_support.h

1 Answer 1

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related