CVI 8.5 crashes on call to nianlys.dll

[blakney]

Hello All,

I see some quriky behavior in CVI 8.5 (that may well be introduced in 8.5 but hard to tell). At least one computer is crashing on a call to nianlys.dll. The Windows error reporting suggest the following.

Appname: ftmsdata.exe Appver 3.54.27 Modname: nianlys.dll Modver : 8.5.0.279 offset: 0003e70a

This may well be related to attempting  a complex FFT in NI analysis package.

One other thing to note is that Task Mananger notes a huge jump in memeory consumption at the time of the crash. (roughly a 1 GB jump in RAM use). The affected computer is a laptop with 2GB RAM and the ram never shows to get totally tapped - at least not as reported by Task Mananger.

The same software runs fine on multiple other computers with a total memory use of less than 200 MB.

The same laptop ran very similar versions of the software created under CVI 8.1.1 - but not identical.

Thanks in advance for your time.

[blakney]

I just reread my earlier post and realized I missed some info I meant to add originally.

The afflicted laptop is running WinXP 2005 Tablet edition. Is there any reason to suspect the OS in this situation.

The complex FFT would have been a 4 Megasample FFT as double.

Greg

[DRock]

Hey Greg,

You indicate that other PCs run the identical program (using 8.5) with no issue at all - what O/S are they running?  Are they also running Tablet 2005?  Are there any other significant differences between the configurations of the machines?

Did this program ever run correctly on the troublesome machine, in 8.5?  I know you said it ran a similar version under 8.1.1, but am I reading correctly that the software has been troublesome the entire time you've had CVI 8.5?

[blakney]

The software seems to run fine on several of lab computers but most of these are near identical Dells Optiplex 755s and at least one XPS M1730 😉 cofigured with WinXP Pro ( most likely 2002 version ) - will check tomorrow. The affected computer is a grad student's personal computer - HP laptop don't recall model. As for the software under 8.1.1, the software had undergone minor revisions at the same time as the CVI update. I may try to build load 8.1.1 on a free computer, see if the newer code will build and then see if the problem resolves itself. Give me a day or two and I will report back.

I was mainly wondering if there was some known issue with the Analysis package v8.5 under XP tablet. I guess there is enough weirdness with that configuration a match might be a longshot.

Greg

[blakney]

New information. Not specific to CVI 8.5

The version of analysis package does not matter (at least for 8.0.1 to current) but the size of the array sent to ReFFT function does. On this one laptop (Compaq TC4400 Intel processor T5600 2 GBRAM XP 2005 Tablet). A 8Mword array will crash and generate a Microsoft report for nianlys.dll different offsets for each version of the dll but I'll assume that is normal. The same data array when truncated once to 4MWord will run fine. Also same executable run on a couple of separate computers runs fine at the full 8 MWord.

Although we had already tried a new stick of RAM (in both slots and with only 1GB loaded), I am now back to thinking this might be a hardware issue. Any opposing arguments are welcome. If hardware, any suggestions on localizing the source (my guess is memory / mem controller (mobo) / processor)?

Any good suggestions for a freeware exerciser program that would best mimic the hardware a large FFTs?

I would like to figure out how the student could get this issue isolated to get a warranty repair, if needed.

Greg 

[DRock]

Hey Greg!

This one could be pretty tough to narrow down.  First and foremost, let me say that Tablet 2005 is not, in fact, a supported operating system for LabWindows/CVI.  We cannot predict the behavior of any of its features on this O/S (as we have not done thorough testing), and this could always be the very heart of the issue.

That being said, we'd obviously like to help you as best we can.

My first suggestion is for us to try to determine if this is really something that is being handled different between CVI 8.1.1 and CVI 8.5.  You already indicate that a slightly different revision of the software seems to operate without problems in CVI 8.1.1 on a different PC - can we try 8.5 on that machine?  Can we try 8.1.1 on the student's laptop?

One thing we could do is to "exchange" versions of the analysis DLL you're using between the different PCs.  The nianlys.dll can be found located in the <National Instruments>\Shared\Analysis directory.  What happens when you copy the CVI 8.1.1 version of this DLL to the student's laptop and (temporarily) replace the DLL found on the student's machine?  By "temporarily" I mean, ensure that you move the student's 8.5 version of the DLL to a different directory / make a copy in order to back it up.  Does this affect the behavior of the software?  What about a transfer of the 8.5 DLL to the 8.1.1 machines?

What information do you have about your data set?  Do other functions in the analysis library provoke the same crashing behavior with the same data set (even if using the functions with that data set don't inherently "make sense")?