Problem in time calculation.

[Bennet Xavier]

Hi All,

I have a method in which i have to calcute the time take for a function to execute.
here is the way i am trying to get the time.

public void Sample()
{

long starttime = System.currentTimeMillis();
. 
.
Function Call()
.
.
System.out.println(rmsOperationId+" -Time taken to retreive the configured plugin-type is : "
    + (System.currentTimeMillis() - starttime)+" MS ");

}

-1 is getting displayed as the result. in which scenario this can happen ? as System.currentTimeMillis should be always greter than starttime .

2010-06-03 21:06:41,129 this is the logged time.

Thanks in advance.
Bennet.

[Martin Vanyavchich]

Weird. Does it happend like this every time you run it?

[Bennet Xavier]

Hi ,

Not every time... once in a while this problem occurs.

I have tested for 5 times...2 times i recevied the -1.

I am unable to reproduce the same.

Thanks,
Bennet

[Campbell Ritchie]

There may inaccuracies and imprecisions in the milliseconds method. You would do better to find the other methods of the System class which may give you a more precise time.

Some computer clocks only calculate their time to the nearest 0.1s or 0.01s, so that introduces more inaccuracy into the millisecond method. Please confirm that you are applying all your calls to the same machine.

[Rob Spoor]

Campbell Ritchie wrote:There may inaccuracies and imprecisions in the milliseconds method. You would do better to find the other methods of the System class which may give you a more precise time.

Some computer clocks only calculate their time to the nearest 0.1s or 0.01s, so that introduces more inaccuracy into the millisecond method. Please confirm that you are applying all your calls to the same machine.

Although System.currentTimeMillis() is inaccurate and imprecise, this is the first time I've ever heard of the return value decreasing between subsequent calls. Returning the same value sure, but a lower one? Even if it rounds (up or down) from 0.01s, it should never get a lower value unless the system clock is decreased.

[Paul Clapham]

Rob Prime wrote:Even if it rounds (up or down) from 0.01s, it should never get a lower value unless the system clock is decreased.

It's possible the system clock could be decreased. Many computers these days have a process which checks the computer time against an outside server periodically, adjusting it as required. This adjustment could certainly change the system clock backwards.

[Edwin Dalorzo]

This is rather awkward. I think I would like to see the whole code. You just wrote an excerpt from it. The problem could be in the code were are not seeing. Could you post an executable example of your code?

[Henry Wong]

Paul Clapham wrote:
It's possible the system clock could be decreased. Many computers these days have a process which checks the computer time against an outside server periodically, adjusting it as required. This adjustment could certainly change the system clock backwards.

While I agree that this can happen, it needs to happen between the two calls to fetch the time. And I am very skeptical that it happens often enough for this to happen 2 times out 5 tries. Also, for this to happen, the difference of time must be small -- since you won't see negative numbers if the total is not near zero. Add to that that the drift needs to be greater than the time readings itself... ie. too many things to conclude that there is something else going on.

Henry

[Campbell Ritchie]

No longer a "beginning" question. Moving thread.