jogamp › jogl

High-resolution timer

Classic

List

Threaded

7 messages Options

foxua

High-resolution timer

Is there platforn-independent high-resolution Timer in JOGL (like the Timer class in LWJGL)?

gouessej

Re: High-resolution timer

Administrator

Hi!

foxua wrote

Is there platforn-independent high-resolution Timer in JOGL (like the Timer class in LWJGL)?

As far as I know, there is no plan for such a feature in JOGL but feel free to contribute.

If you really need an high-resolution timer, maybe look at the one used in Java3D (com.sun.j3d.utils.timer.J3DTimer) or in java.util.concurrent.locks.LockSupport (parkNanos uses a native high-resolution timer under the hood). Is System.nanoTime() accurate enough for your needs?

Julien Gouesse | Personal blog | Website

foxua

Re: High-resolution timer

Thanks for the reply
AFAIK nanotime() implementation has a bug - on some multicore systems it behaves incorrectly.
I'll try to use nanotime() - maybe Oracle fixed it - I don't know.

gouessej

Re: High-resolution timer

Administrator

foxua wrote

Thanks for the reply
AFAIK nanotime() implementation has a bug - on some multicore systems it behaves incorrectly.
I'll try to use nanotime() - maybe Oracle fixed it - I don't know.

I know this bug, I use a small workaround in TUER to detect when the time "seems" to come back :s It concerns only AMD microprocessors as far as I know.

Julien Gouesse | Personal blog | Website

Sven Gothel

Re: High-resolution timer

Administrator

On Tuesday, November 16, 2010 12:36:49 gouessej [via jogamp] wrote:

>
> foxua wrote:
> >
> > Thanks for the reply
> > AFAIK nanotime() implementation has a bug - on some multicore systems it
> > behaves incorrectly.
> > I'll try to use nanotime() - maybe Oracle fixed it - I don't know.
> >
> I know this bug, I use a small workaround in TUER to detect when the time
> "seems" to come back :s It concerns only AMD microprocessors as far as I
> know.

We may add something in gluegen Platform class:
long getNanoTime(); // @return >=0 supported, <0 unsupported

just tell me where to fetch your implementation .. thank you.
Other ideas welcome, but I won't do it myself now, no time.

~Sven

gouessej

Re: High-resolution timer

Administrator

Sven Gothel wrote

On Tuesday, November 16, 2010 12:36:49 gouessej [via jogamp] wrote:
>
> foxua wrote:
> >
> > Thanks for the reply
> > AFAIK nanotime() implementation has a bug - on some multicore systems it
> > behaves incorrectly.
> > I'll try to use nanotime() - maybe Oracle fixed it - I don't know.
> >
> I know this bug, I use a small workaround in TUER to detect when the time
> "seems" to come back :s It concerns only AMD microprocessors as far as I
> know.

We may add something in gluegen Platform class:
long getNanoTime(); // @return >=0 supported, <0 unsupported

just tell me where to fetch your implementation .. thank you.
Other ideas welcome, but I won't do it myself now, no time.

The native high-resolution timer used in LockSupport is not concerned by this bug. I have looked at my source code, I don't find the hack I used. Actually, I called very often System.nanoTime() and when 2 successive calls gave absurd result, I used the Olympic average to compute the elapsed time.

Julien Gouesse | Personal blog | Website

Michael Bien

Re: High-resolution timer

On 11/16/2010 04:38 PM, gouessej [via jogamp] wrote:

Sven Gothel wrote:

On Tuesday, November 16, 2010 12:36:49 gouessej [via jogamp] wrote:
>
> foxua wrote:
> >
> > Thanks for the reply
> > AFAIK nanotime() implementation has a bug - on some multicore systems it
> > behaves incorrectly.
> > I'll try to use nanotime() - maybe Oracle fixed it - I don't know.
> >
> I know this bug, I use a small workaround in TUER to detect when the time
> "seems" to come back :s It concerns only AMD microprocessors as far as I
> know.

We may add something in gluegen Platform class:
long getNanoTime(); // @return >=0 supported, <0 unsupported

just tell me where to fetch your implementation .. thank you.
Other ideas welcome, but I won't do it myself now, no time.

The native high-resolution timer used in LockSupport is not concerned by this bug. I have looked at my source code, I don't find the hack I used. Actually, I called very often System.nanoTime() and when 2 successive calls gave absurd result, I used the Olympic average to compute the elapsed time.
http://tuer.sourceforge.net
http://gouessej.wordpress.com

View message @ http://jogamp.762907.n3.nabble.com/High-resolution-timer-tp1901680p1911462.html
To start a new topic under jogamp, email [hidden email]
To unsubscribe from jogamp, click here.

btw AMD provides an "optimizer" to workaround the multicore timing issues:
http://support.amd.com/us/Pages/dynamicDetails.aspx?ListID=c5cd2c08-1432-4756-aafa-4d9dc646342f&ItemID=153

i am not aware of any timing issues regarding System.nanotime() on other multicore CPUs.

regards,
michael

-- 
http://michael-bien.com/