[Dev-C++] C versus C++

[Ioannis V. <no...@ya...>]

It is great you are interested for the reality of the things. :) At
first your code had some errors (for 100% ansi standard) and omitions
(you forgot #include <iostream> for example) which i fixed. I did not
dare to see your code and think how exactly your algorithm work but the
results showed the C++ versions to be slower than C's.

I attach you your code fixed as old.cpp.

I wrote my code using the standard library sorting function thus
applying the same algorithm to both stuff. The code is attached as=20
temp.cpp.

The results are:

C:\c>temp
begin: 2002 cend: 24885 cppend: 47558

char * sorting speed: 22883 std::string sorting speed: 22673

As you see the string speed here is smaller than char *. :)

In less elements the opposite happened. In both situations the
difference is small. In worse case scenario C++ code must function at
the same speed as C's code. And this because C++ was designed for 0
space and time overhead.

Also as i said you before when you do not use string and use char
pointers you end up using more code than those provided by built in
member functions(=3Dspace overhead). Also you end up with time overhead!
And this because function calls are more expensive than member function
calls, and especially in a good compiler where template member functions
are inlined. So in practice using C++ mechanisms is far better than
using C's because of the less head aches, clearer code, more bug free
code (you can end up copying things after the end of a char array for
example, but not in a string), less code, far easier maintainability and
probably better speed!

As i told before take also a look at:=20

 http://www.research.att.com/~bs/esc99.html


The reason that your code produces big differences is obviously bugs.
The algorithms aren't applied fairly the same at both worlds, e.t.c..

Also i am glad to discuss with you these matters. I am waiting fro your
reply on the issue.

I am sending my reply in the list because there are others who will
benefit from this discussion. :)


Ioannis

* Ioannis Vranos
* Programming pages: http://www.noicys.f2s.com
* Alternative URL: http://run.to/noicys




> -----Original Message-----
> From: Jason Hardman [mailto:jas...@bt...]
> Sent: Monday, October 08, 2001 4:44 AM
> To: Ioannis Vranos
> Subject: Re: [Dev-C++] Help with functions
>=20
>=20
> Hi again,
>=20
> I know this is an old issue but I wanted to demonstrate a
> point. I understand the points you were making about function=20
> calls versus method calls and how the varying overheads are=20
> minimal. However I wanted to show you that C string versus=20
> stl::string does provide a lot of difference.
>=20
> If you take a look at the attached code you will find 4
> different versions of the sort algorithm functions I had=20
> posted earlier. The two original ones using C style strings,=20
> slightly redesiged to not modify the input string, and then=20
> two new version of the same algorithms, using the stl::string=20
> class. If you run the program it will report how long it took=20
> a certain number of iterations of each function. The results=20
> have been quite interesting on my computer, and I assume they=20
> will be similar on yours. I recomend that you run it with and=20
> without compiler optimisations switched on. The compiler I=20
> use is MinGW, so I don't know how other compilers will=20
> perform, but I don't think the results will change a great=20
> deal in others. Be prepared to wait a bit, my computer is a=20
> PIII 750 and the whole program when fully optimised takes=20
> about 20 secs, unoptimised is much longer.
>=20
> The reason I have done this is out of an interest in such
> things, and also just sheer bloody mindedness. I haven't=20
> meant to do it to be argumentative, but meerly to illustrate=20
> an issue. Hope you don't mind? I have actualy enjoyed=20
> tinkering around with the program as I haven't had much=20
> opportunity in the past to analyse speeds in programs. There=20
> are probably a few more optimisations I could introduce,=20
> especially in both V1 functions, mainly by using pointers and=20
> iterators respectivly.
>=20
> This doesn't mean that I recomend C strings over stl::string
> objects, but it does demonstrate that a person should choose=20
> wisely depending upon the task they will be performing. I=20
> welcome any thoughts you have.
>=20
> I will note that the case I have demonstrated isn't of any
> import unless working with a very large set of data, when=20
> execution time is critical. Also the string I have supplied=20
> as a test case is one that tries to provide a representative=20
> case for normal use of the algorithms. It is interesting (to
> me) how using different input strings changes the results,=20
> mainly when comparing the two algorithms. Try strings of=20
> differing lengths, and vary how unsorted the strings are,=20
> i.e. give it a string that doesn't need any sorting.
>=20
> Thanx for putting up with me, Jason.
>=20
> ----- Original Message -----
> From: "Ioannis Vranos" <no...@ya...>
> To: "Dev-C++" <dev...@li...>
> Sent: Friday, October 05, 2001 6:45 AM
> Subject: RE: [Dev-C++] Help with functions
>=20
> The notion that the C subset is fastest than C++ is a myth.
> See http://www.research.att.com/~bs/esc99.html and=20
> http://www.research.att.com/~bs/crc.pdf
>=20
> C++ string is a
> typedef of basic_string<char> type that is:
>=20
> typedef basic_string<char> string;
>=20
> basic_string is a template class and template classes tend to
> expanded inline when compiled. Also a good compiler creates=20
> instances only of the member functions of a template class=20
> which are used and ignores the rest. Thus since the code is=20
> expanded inline in the application you have 0 runtime overhead.
>=20
>=20
>=20
>=20