Software operation is slowing down after large number of loops

[ianhf]

Hi

Hopefully somebody can help me with this.

I have used LabWindowsCVI to generate test code for a semiconductor product.

I have used LabWindows Test Executive as the sequencer to control the test and test operations. i have nested the test sequence

operation into three levels, first level  batch information  and wafer operation, second level individual selection of die on wafer and third

level actually tests the individual die.

However, when i have run a large number of tests (> 8000), the test executive software slows to a halt.

When i have looked at the code the software is slowing down / stopping when the tests leaves the third sequence and returns to the second

sequence. All sequences pre-loads the tests.

I have checked memory leakage and there seems to be a small amount of leakage but not enough to stop the program running.

I have checked my code but can't see an obvious cause of memory leakage.

Has anybody seen this problem before using the test executive sequence or have any suggestions what could be causing it.

Thanks

Ian

[JamesC]

Hi Ian,

Please can you tell me what version of CVI and Test Executive you are using. Also what version of Windows are you running (Inc. Service packs)?

Also what is the memory usage when you start your application, and what is the memory usage after 8000 loops?

What do each of your sequences do?
How many times does the 3rd sequence run for the 2nd Sequence? (I.e 2nd Seq runs once, 3rd Seq runs once. Or 2nd Seq runs once 3rd Seq runs 5 times).

Regards
JamesC
NIUK and Ireland

[ianhf]

Hi James

I am running the following software

LabWindowsCVi 8.0.0 (273)

Test Executive 2.0.2

Microsoft Windows XP version 2002 Service Pack 2

I haven't actually run 8000 loops yet but the memory increase is 15M for 400 runs  with a start memory of 22,520K.

I hadn't realised this was increasing by 40K per run and wil try and find out why this is increasing so much.

The first sequence controls and selects a batch of wafers by calling sequence 2 for each individual wafer in the batch. At the moment we are using 12 wafers per batch

The second sequence contains 11 tests that loops back and calls sequence 3 to test each individual die on the wafer There is typically 810 die on each wafer

The third sequence contains 105 tests for testing the individual die.

ie first sequence run once

   second sequence run 12 times

   third sequence run 810

reagrds

Ian

[cdk52]

Hi.

Not sure whether it is related, but I have observed this for some time:

While my CVI program is running, its memory use (according to Windows Task Manager) steadily increases over a period of hours, and plateaus at some level which seems to depend on how many analog channels are acquired, and the type of processing being done. On some systems, I have seen memory use peak at over 100MB. However, when I minimise the app, its memory use immediately falls to below 10MB (suggesting that there is no memory leak), then begins to grow again.

The app uses a number of threads, with most communication between them involving the allocation and freeing of memory.

I use CVI v7.1.1, and the target machines run Windows NT4 SP6, or 2000 SP4.

Does anyone have a clue about this?

Cheers,
Colin.

[JamesC]

Hi Colin,

I would not think your issue is related to the one Ian is seeing, firstly because you are using different versions of teststand, and secondly because I think Ian's issue is due to Test Executive and reporting options.

I would suggest you start a new forum thread in the LabWindows/CVI board, so we can keep issues "on-topic"

Regards
JamesC
NIUK and Ireland

[JamesC]

Ian,

As I emailed you, I think your issue is caused by the hardware you are using.
256MB is the minimum memory requirement for Windows XP SP2, and as you only have 128MB memory in your controller, I think alot of needed data (test modules/windows components) are being moved to the page file and slowing down your system, when in the next processor time slice Test Executive need the paged data.

Regards
JamesC
NIUK and Ireland

[ianhf]

Hi James

I have looked at the virtual memory and the debug / cvi processes are swopping out approx 70M and 60M to virtual memory.

This will obviously slow the whole process down and i am looking to increase the memory in the machine.

However, do you think this will give the memory increase / commit charge constatnt climb as i am not confident in this fixing all my probelms.

Thanks

Ian

[JamesC]

Hi Ian,

In an earlier post you said your memory usage increases by 15MB for 400 runs, can you clarify if this is 400 runs of Sequence 3, so you are in your first run of Sequence 1. Or is this 400 runs for Sequence 1.

If the 400 runs is of sequence 3 then the memory increase should help, if a reference is not being closed until sequence 2 completes.

If however the 400 runs is of sequence 1 then a memory upgrade would only be a tempory fix, as there must be a larger memory leak that will need to be found.

Regards
JamesC
NIUK and Ireland

[ianhf]

Hi James

Sequence 1 was only run once, sequence 2 ran five times and sequence 3 ran 400 times.

I have the reporting turned off by using menu features in the test executive GUI. when i check the code the report file returns before writing data to file.

thanks

Ian