CINT can generate optimized byte code for very simple loops not involving
calls to procedures or I/O.
In your example below, the first procedure is optimized.
The second cannot be optimized.
Masa will very likely provide more info.
Rene Brun
Pasha Murat wrote:
> Hello,
>
> in the example below I'm interactively running 2 test routines which I
> consider to be exactly the same from the point of view of execution:
> the only difference between them consists in that one line which
> *should never be executed* is uncommented in the second routine.
>
> I'd also expect the execution times to be exactly the same.
> However ROOT TStopwatch timer reports *very* strange results:
> the reported times differ by more than a factor of 100...
>
> One could either suspect TStopwatch timer in reporting wrong time in one
> of 2 cases or CINT in doing much more than it has been asked to in case
> when `cout ...' is uncommented.
>
> I should also mention that replacing `cout' in test2 with `printf' makes
> both times *exactly* the same in agreement with the expectations...
>
> Could anybody comment on it?
> Thanks, Pasha
>
> P.S. For the information to be complete <iostream.h> is included in logon macro.
> --------------------------------------------------------------------------------
> void test1 () {
> TStopwatch t;
> t.Start();
> int k;
> for (int i=0; i<100000; i++) {
> k = i;
> if (k == -1) {
> // cout << " oops! " << endl;
> }
> }
> t.Stop();
> printf("RealTime=%f seconds, CpuTime=%f seconds\n",t.RealTime(),t.CpuTime());
> }
>
> void test2 () {
> TStopwatch t;
> t.Start();
> int k;
> for (int i=0; i<100000; i++) {
> k = i;
> if (k == -1) {
> cout << " oops! " << endl;
> }
> }
> t.Stop();
> printf("RealTime=%f seconds, CpuTime=%f seconds\n",t.RealTime(),t.CpuTime());
> }
> --------------------------------------------------------------------------------
> root [1] test1()
> RealTime=0.230000 seconds, CpuTime=0.240000 seconds
> root [2] test2()
> RealTime=28.810000 seconds, CpuTime=27.770000 seconds
This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 00:34:29 MET