On 8/26/06, Janos Vegh wrote:

I just used the default settings, no extra switches. BTW, you can edit and run the test program, maybe you can conclude something interesting through varying the optimization. Anyhow, the difference between simple and [] access is too big for a simple optimization, I think so.

main has to return int; Your code isn't legal C++. Also, it seems like you end up using memory in the vector after only reserving it, not resizing it. Your empty cycle time makes it obvious you're not even doing the most rudimentary of optimizations. Here's what I get in g++ (g++ (GCC) 4.1.1 (Gentoo 4.1.1)) after making a few changes while using -O3 -DNDEBUG as arguments to the compiler (though -DNDEBUG didn't make any signifigant difference): Empty cycle real 0m0.003s user 0m0.000s sys 0m0.000s STL=1, SIMPLE=1, size=1000 real 0m1.357s user 0m1.168s sys 0m0.008s STL=1, size=1000 real 0m1.337s user 0m1.172s sys 0m0.016s STL=1, ITERATOR=1, size=1000 real 0m1.315s user 0m1.172s sys 0m0.012s STL=1, AT=1, size=1000 real 0m1.357s user 0m1.176s sys 0m0.004s I really don't see an issue here. ~ Scott McMurray Code used: #define STL 1 #define SIMPLE 0 #define ITERATOR 0 #define AT 1 #define EMPTY 0 #if STL #include <vector> #endif using namespace std; #include <iostream> #define ARR_SIZE 1000 #define CYCLES 1000000 int main (int argc, char** argv) { long int i,j,*P; #if EMPTY cout << "Empty cycle"; #else #if STL #if ITERATOR || AT std::vector<long int>::iterator pos; #endif std::vector<long int> arr(ARR_SIZE); cout << "STL=1"; #if SIMPLE cout << ", SIMPLE=1"; #endif #if ITERATOR cout << ", ITERATOR=1"; #endif #if AT cout << ", AT=1"; #endif cout << ", size=" << arr.size(); #else long int *arr; arr = new long int[ARR_SIZE]; cout << "STL=0"; #endif #endif cout << endl; #if SIMPLE && STL P = &arr[0]; #endif for(j=0;j

Use g++ -O3 (not -o3) for optimization. You might try posting this to comp.lang.c++ since this has nothing to do with boost. Chris On 8/26/06, Janos Vegh wrote:

On 8/26/06, me22 wrote:

...

main has to return int; Your code isn't legal C++. Also, it seems like you end up using memory in the vector after only reserving it, not resizing it.

Yes, a forgotten int, my mistake. After reserving the memory, I guess I can write the memory locations. I think neither of the above affects the subject.

...

Your empty cycle time makes it obvious you're not even doing the most rudimentary of optimizations.

My goal was to measure the time of the empty processing, in order to estimate the time of the net calculation, rather than to find out that for no operation I need no time.

...

Here's what I get in g++ (g++ (GCC) 4.1.1 (Gentoo 4.1.1)) after making a few changes while using -O3 -DNDEBUG as arguments to the compiler (though -DNDEBUG didn't make any signifigant difference):

Empty cycle real 0m0.003s user 0m0.000s sys 0m0.000s

STL=1, SIMPLE=1, size=1000 real 0m1.357s user 0m1.168s sys 0m0.008s

STL=1, size=1000 real 0m1.337s user 0m1.172s sys 0m0.016s

STL=1, ITERATOR=1, size=1000 real 0m1.315s user 0m1.172s sys 0m0.012s

STL=1, AT=1, size=1000 real 0m1.357s user 0m1.176s sys 0m0.004s

I really don't see an issue here.

I neither, except that it looks like that in the case of AT checking the lower and upper limits of the index needs no extra time(??).

Otherwise, on my system $ g++ -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=1, ITERATOR=0, AT=0, size=1000 real 0m4.775s user 0m4.476s sys 0m0.080s

$ g++ -o3 -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=1, ITERATOR=0, AT=0, size=1000 real 0m4.749s user 0m4.496s sys 0m0.060s

and

$ g++ -o3 - oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=0, size=1000 real 0m33.413s user 0m32.076s sys 0m0.100s

$ g++ -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=0, size=1000 real 0m34.554s user 0m33.668s sys 0m0.060s

and

$ g++ -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=1, size=1000 real 1m26.753s user 1m24.301s sys 0m0.080s

$ g++ -o3 -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=1, size=1000 real 1m29.311s user 1m27.035s sys 0m0.050s

After that you were so kind to run my test on your Linux, maybe I need to ask MingW people about this problem.

Anyhow, thanks for replying and sorry for bothering the list with my problem

Janos

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

This is most likely that you are not getting your operator[] inlined. The function call overhead for such a small operation is going to be very large. Optimizations are your friends. On 8/26/06, Janos Vegh wrote:

On 8/26/06, me22 wrote:

...

main has to return int; Your code isn't legal C++. Also, it seems like you end up using memory in the vector after only reserving it, not resizing it.

Yes, a forgotten int, my mistake. After reserving the memory, I guess I can write the memory locations. I think neither of the above affects the subject.

...

Your empty cycle time makes it obvious you're not even doing the most rudimentary of optimizations.

My goal was to measure the time of the empty processing, in order to estimate the time of the net calculation, rather than to find out that for no operation I need no time.

...

Here's what I get in g++ (g++ (GCC) 4.1.1 (Gentoo 4.1.1)) after making a few changes while using -O3 -DNDEBUG as arguments to the compiler (though -DNDEBUG didn't make any signifigant difference):

Empty cycle real 0m0.003s user 0m0.000s sys 0m0.000s

STL=1, SIMPLE=1, size=1000 real 0m1.357s user 0m1.168s sys 0m0.008s

STL=1, size=1000 real 0m1.337s user 0m1.172s sys 0m0.016s

STL=1, ITERATOR=1, size=1000 real 0m1.315s user 0m1.172s sys 0m0.012s

STL=1, AT=1, size=1000 real 0m1.357s user 0m1.176s sys 0m0.004s

I really don't see an issue here.

I neither, except that it looks like that in the case of AT checking the lower and upper limits of the index needs no extra time(??).

Otherwise, on my system $ g++ -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=1, ITERATOR=0, AT=0, size=1000 real 0m4.775s user 0m4.476s sys 0m0.080s

$ g++ -o3 -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=1, ITERATOR=0, AT=0, size=1000 real 0m4.749s user 0m4.496s sys 0m0.060s

and

$ g++ -o3 - oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=0, size=1000 real 0m33.413s user 0m32.076s sys 0m0.100s

$ g++ -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=0, size=1000 real 0m34.554s user 0m33.668s sys 0m0.060s

and

$ g++ -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=1, size=1000 real 1m26.753s user 1m24.301s sys 0m0.080s

$ g++ -o3 -oeffic2.exe effic2.cpp $ time ./effic2.exe STL=1, SIMPLE=0, ITERATOR=0, AT=1, size=1000 real 1m29.311s user 1m27.035s sys 0m0.050s

After that you were so kind to run my test on your Linux, maybe I need to ask MingW people about this problem.

Anyhow, thanks for replying and sorry for bothering the list with my problem

Janos

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users