Kcachegrind/callgrind is inaccurate for dispatcher

I have a model code on which kcachegrind/callgrind reports strange results. It is kind of dispatcher function. The dispatcher is called from 4 places; each call says, which actual do_J function to run (so the first2 will call only do_1 and do_2 and so on)

Source (this is a model of actual code)

#define N 1000000

int a[N];
int do_1(int *a) { int i; for(i=0;i<N/4;i++) a[i]+=1; }
int do_2(int *a) { int i; for(i=0;i<N/2;i++) a[i]+=2; }
int do_3(int *a) { int i; for(i=0;i<N*3/4;i++) a[i]+=3; }
int do_4(int *a) { int i; for(i=0;i<N;i++) a[i]+=4; }

int dispatcher(int *a, int j) {
    if(j==1) do_1(a);
    else if(j==2) do_2(a);
    else if(j==3) do_3(a);
    else do_4(a);
}

int first2(int *a) { dispatcher(a,1); dispatcher(a,2); }
int last2(int *a) { dispatcher(a,4); dispatcher(a,3); }
int inner2(int *a) { dispatcher(a,2); dispatcher(a,3); }
int outer2(int *a) { dispatcher(a,1); dispatcher(a,4); }

int main(){
    first2(a);
    last2(a);
    inner2(a);
    outer2(a);
}

Compiled with gcc -O0; Callgrinded with valgrind --tool=callgrind; kcachegrinded with kcachegrind and qcachegrind-0.7.

Here is a full callgraph of the application. All paths to do_J go through dispatcher and this is good (the do_1 is just hided as too fast, but it is here really, just left to do_2)

Lets focus on do_1 and check, who called it (this picture is incorrect):

And this is very strange, I think, only first2 and outer2 called do_1 but not all.

Is it a limitation of callgrind/kcachegrind? How can I get accurate callgraph with weights (proportional to running time of every function, with and without its childs)?

Yes, this is limitation of callgrind format. It doesn't store full trace; it only stores parent-child calls information.

There is a google-perftools project with pprof/libprofiler.so CPU profiler, http://google-perftools.googlecode.com/svn/trunk/doc/cpuprofile.html . libprofiler.so can get profile with calltraces and it will store every trace event with full backtrace. pprof is converter of libprofile's output to graphic formats or to callgrind format. In full view the result will be the same as in kcachegrind; but if you will focus on some function, e.g. do_1 using pprof's option focus; it will show accurate calltree when focused on function.