Why does compiling over 100,000 lines of std::vector::push_back take a long time?

Question

I\'m compiling a C++ library which defines a single function that randomly samples from a set of data points. The data points are stored in a std::vector. There a

Answer 1

This question was completely answered by the great answer from osgx.

Maybe one additional aspect: push_back() vs initialization list

When running the above test with 100000 push_backs, I get the following result with a gcc 4.4.6 on a Debian 6.0.6 system:

$ time g++ -std=c++0x -ftime-report ./pb100k.cc 

Execution times (seconds)
 garbage collection    :   0.55 ( 1%) usr   0.00 ( 0%) sys   0.55 ( 1%) wall       0 kB ( 0%) ggc
 ...
 reload                :  33.95 (58%) usr   0.13 ( 6%) sys  34.14 (56%) wall   65723 kB ( 9%) ggc
 thread pro- & epilogue:   0.66 ( 1%) usr   0.00 ( 0%) sys   0.66 ( 1%) wall      84 kB ( 0%) ggc
 final                 :   1.82 ( 3%) usr   0.01 ( 0%) sys   1.81 ( 3%) wall      21 kB ( 0%) ggc
 TOTAL                 :  58.65             2.13            60.92             737584 kB

real    1m2.804s
user    1m0.348s
sys     0m2.328s

When using an initialization list, it is much faster:

$ cat pbi100k.cc
#include 
using namespace std;

int main()
{
   vector d {
   0.190987822870774,
   /* 100000 lines with doubles generated with:
          perl -e 'print(rand(10),",\n") for 1..100000'
   */
   7.45608614801021};

  return d.size();
}

$ time g++ -std=c++0x -ftime-report ./pbi100k.cc 

Execution times (seconds)
 callgraph construction:   0.02 ( 2%) usr   0.00 ( 0%) sys   0.02 ( 1%) wall      25 kB ( 0%) ggc
 preprocessing         :   0.72 (59%) usr   0.06 (25%) sys   0.80 (54%) wall    8004 kB (12%) ggc
 parser                :   0.24 (20%) usr   0.12 (50%) sys   0.36 (24%) wall   43185 kB (65%) ggc
 name lookup           :   0.01 ( 1%) usr   0.05 (21%) sys   0.03 ( 2%) wall    1447 kB ( 2%) ggc
 tree gimplify         :   0.01 ( 1%) usr   0.00 ( 0%) sys   0.02 ( 1%) wall     277 kB ( 0%) ggc
 tree find ref. vars   :   0.01 ( 1%) usr   0.00 ( 0%) sys   0.01 ( 1%) wall      15 kB ( 0%) ggc
 varconst              :   0.19 (15%) usr   0.01 ( 4%) sys   0.20 (14%) wall   11288 kB (17%) ggc
 integrated RA         :   0.02 ( 2%) usr   0.00 ( 0%) sys   0.02 ( 1%) wall      74 kB ( 0%) ggc
 reload                :   0.01 ( 1%) usr   0.00 ( 0%) sys   0.01 ( 1%) wall      61 kB ( 0%) ggc
 TOTAL                 :   1.23             0.24             1.48              66378 kB

real    0m1.701s
user    0m1.416s
sys     0m0.276s

This is about 30+ times faster!