CurveFit/ODR=3 with weight crashes Igor - CPU dependency?

Share icon

Sun, 01/22/2012 - 12:19 pm


I've been fitting 2D data to implicit function.
CurveFit/ODR=3 keeps crashing Igor program when /XW flag is added.
But the worse problem is that the same code runs with no problem in other PC.
I tried three PCs - all on Windows 7 Enterprise for 64bit SP1 & Igor 6.2.2.2.
One laptop has AMD Turion 64 X2, two desktops have Intel core2 6700 / Intel core i5.
The last out of the three is the only which runs the code normally, and the other two gives problem.

And I suspect that the laptop has changed from normal to problematic during the last few days.
Because it fails to run the same code it run a few days ago.

The ODR package has CPU dependency? or is there anything I can check first in this case?
This drives me nuts..
I am not aware of a CPU dependency. It is more likely some sort of memory corruption or uninitialized variable bug. Such bugs can depend on the detailed layout of data and code in memory, and on what you did before you ran the code with problems.

If you can provide a simple example of the problem, please send an example with data and fit function to support@wavemetrics.com.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

January 23, 2012 at 10:46 am - Permalink

Was there a solution to this? I am currently getting some very frustrating crashes with CurveFit/ODR=2 ... it's not entirely reproducible and sometimes results in crashes when trying to save, or close, the experiment (hence the frustrating part).

January 30, 2015 at 04:52 am - Permalink

Was there a solution to this? I am currently getting some very frustrating crashes with CurveFit/ODR=2 ... it's not entirely reproducible and sometimes results in crashes when trying to save, or close, the experiment (hence the frustrating part).

January 30, 2015 at 04:52 am - Permalink

@jcor Maybe you can write a loop which involves calling SaveExperiment to reproduce it?
Can you post the experiment?

January 30, 2015 at 09:56 am - Permalink

In the end, I emailed WaveMetrics and it turned out there was a small bug when using X residual calculation ( /XR flag). The new build should have fixed it, though I haven't tested it yet. For now, I just do /XR manually.

February 2, 2015 at 10:07 am - Permalink