Has anyone else seen issue with trying to allocate more than 1Gig of memory for a buffer?

[Ben]

HI Unclebump,

THe write to file approach is the correct answer if that was an option.

This project requires NO data written to disk until everything els is done.

Second to acquiring the data, that is the most rigid requirement of the application.

Ben

[unclebump]

Talk about tying your hands and feet together. Sounds like you need more toes to count with. Could you use DSC memory tags and Citadel to hold some of the data in process?? Maybe this link will help.

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/memory/base/4gt_ram_tuning.asp

[Raistlin]

I would also like to note it is NOT a Windows limitation and has nothing to do with the limitations of 32 bit. With a C DLL we can adress up to 2 GB on the same 2 or 4 GB RAM machine (higher gives problems due to 32 bit) where LabVIEW stops at 1 GB or less. So changing boot.ini flags and similar things will not change a thing.

[Ben]

"Talk about tying your hands and feet together."

I believe DSC would freak over my 1K update rate. I also have to timestamp based on a IRIG-B feed and an internal simulation time.

I will chase down you link.

Thank you!

Ben

[Ben]

Hi Raistlin,

Ouch!

Well I never did promise myself I would never C again. It was a dream though.

Ben

[unclebump]

I meant to take your whole data array, flatten it to a binary string. Then write it to a single memory string tag.

[Ben]

Hi Unclebump,

Now that is an interesting thought.

Let me chase down that option.

Ben

Message Edited by Ben on 05-12-2005 09:53 AM

[Ben]

Hi Unnclebump,

I checked with support under service req #673401.

LV-DSC will still be limited to the same memory space that I have in LV.

Good thought!

It is one of those "out of the box" solutions that I am looking for.

Ben

[unclebump]

Sounds like you will need to have two or more distributed computers(PXI boxes?) to acquire and buffer the data. Somehow synchronized between them. What sort of hardware are you using to grab 448 channels? Can you get analog cards(or external hardware) that can buffer the data inside them?? Then Labview could write the file as it reads from the external data sources.

Here is what I was looking at. I'm not trying to steer you away from NI hardware.

http://www.daytronic.com/products/10K/10K488.htm
http://www.daytronic.com/products/8K/8kspec.htm#model

[Ben]

Hi uncle!

In the thread I mentioned in my reply to myself, I list some of the details.

I am actually using a SCRAMNet interface to monitor the interaction of about 10 PC's that use SCRAMNet to create a distributed real-time environment.

SCRAMNet is distributed shared memory that is inked via fiber. 400ns after one node updates a memory address, my machine gets the update and fires an interupt.

I have someone counting them now but it looks like between 2000 and 4000 channels.

NOT ALL will be sampled at 1KHz. The DAQ stuff will. For DAQ I am using an PXI chassis that I talk to via Ethernet.

Time syncronization will be via SCRAMNet or IRIG-B in later phases.

Ben