LabVIEW 2009 vs LV8.6.1 Queue performance

[Mellroth]

I'm just starting to play with LV2009, and tested the DVR against a Single Element Queue (SEQ).

Compared to a SEQ in LabVIEW 8.6.1 the DVR runs about 2 times faster, but if I compare the SEQ speed in LV 2009 against SEQ speed in LV 8.6.1 the speed has decreased. The access time in LV 2009 goes up ~30%.

After seeing this I also measured the time it takes to add 100 000 elements (I32 values) to a queue and then remove them. In LV8.6.1 this takes ~80ms, in LV2009 it takes ~105ms, which is pretty close to 30% increase in access time.

Have anyone else seen this, or any other decrease in performance?

/J

[Yair]

LV 2009 has had some significant changes made under the hood. I don't know the specific details, but my understanding was that it could cause decreased performance in some cases and that NI is expecting this to improve in future versions.

[Mellroth]

LV 2009 has had some significant changes made under the hood. I don't know the specific details, but my understanding was that it could cause decreased performance in some cases and that NI is expecting this to improve in future versions.

It was my understanding that the changes made to the compiler shouldn't have made things worse, but allow for more future optimizations.

Obviously I was wrong

The good part, however, is that the DVR outperforms the SEQ. It is therefore possible to tweak all instances where SEQs are used.

All Producer-Consumer patterns, using queues, will still be 30% slower.

Do you know of any other slowdown in LabVIEW 2009 to be aware of?

/J

[Yair]

I know of no specific slowdowns.

Also, I'm guessing some of the performance issues will be mitigated in future versions. Your understanding wasn't wrong. Just consider it a bug.