LabVIEW versus LabWindows/CVI

[Max Weiss]

I have an order to program an application, which displays data (among
other things) in a multiplot chart with 150 Hz.
I have a Pentium 100 MHz and 48 MB memory. Is LabVIEW fast enough to
manage an application with 6 plots in on chart (scope chart) reliable on
WIN95. If not - does it make sense to use LabWindows/CVI instead of LV.
--
Ciao
Max


* Max Weiss/Kaiserstr. 188/76133 Karlsruhe/0721 2030339/Germany *
* max@mvmpc9.ciw.uni-karlsruhe.de/DB8MWE *

[Greg_McKaskle2]

> I have an order to program an application, which displays data (among
> other things) in a multiplot chart with 150 Hz.
> I have a Pentium 100 MHz and 48 MB memory. Is LabVIEW fast enough to
> manage an application with 6 plots in on chart (scope chart) reliable on
> WIN95. If not - does it make sense to use LabWindows/CVI instead of LV.

Updating a graph 150 times a second, when the screen only updates ~60
times per second won't help much, and takes lots of CPU horsepower. If
you update the chart with arrays of points, then you can throttle the
speed based on how often you want to draw and how much time it takes.
The speed is also very dependent on the size of the chart.

CVI is a great product, but its displays are very similar in capacity
and speed to what you wil
l find in LV. The deciding choice between it
and LV basically comes down to whether you want to write in C or diagrams.

Greg McKaskle

[PJ]

Hi,

Concerning the speed difference LV vs CVI:
In my opinion CVI goes much faster. For example: I did a program in LV using
the Levenberg Marquardt VI (GMath). I ran the program
and found that it took 13 minutes for the VI to deliver a proper fit.

Then I did a corresponding program in CVI, using the NonLinearFit function.
I ran exactly the same spectra with exactly the same
guess coefficients as in LV(and the same computer of course). It took 1
minute, 27 seconds. And that I found rather amazing, indeed.
I tested several other types of spectras and CVI always fits 10 times (at
least) faster.

Maybe in other cases the difference is not as striking as in this curve
fitting example, and it depends on how the algorithm is implemented, I know.
But until someone convinces me of the contrary, also in general I choose CVI
when I want speed. Out in the industry, I know that programmers often uses
LV to develop something quickly, and then they use C to speed it up.
I am open for any ideas, I'm just telling you what I've seen.
Note: No shadow over LV. LV will always have an important position.
Any comments of the mentioned curve fitting example would be interesting.
Would the LV program become faster if I use the
Application Build Toolkit to make a standalone exe.file?

/PJ



Greg McKaskle skrev i meddelandet <374CB693.DD515391@austin.rr.com>...
>> I have an order to program an application, which displays data (among
>> other things) in a multiplot chart with 150 Hz.
>> I have a Pentium 100 MHz and 48 MB memory. Is LabVIEW fast enough to
>> manage an application with 6 plots in on chart (scope chart) reliable on
>> WIN95. If not - does it make sense to use LabWindows/CVI instead of LV.
>
>Updating a graph 150 times a second, when the screen only updates ~60
>times per second won't help much, and takes lots of CPU horsepower. If
>you update the chart with arrays of points, then you can throttle the
>speed based on how often you want to draw and how much time it takes.
>The speed is also very dependent on the size of the chart.
>
>CVI is a great product, but its displays are very similar in capacity
>and speed to what you will find in LV. The deciding choice between it
>and LV basically comes down to whether you want to write in C or diagrams.
>
>Greg McKaskle

[Dr. Franz-Josef Ahlers]

Hi,

I think it's the parsed mathematics in GMath that makes the fitting in labview
so slow.
You should try the levenberg-Marquardt with the built LM-Fit (maybe it is only
available in the
advanced analysis library??).

Of course it is much more tedious to use the built in LM-Fit, since you
practically have to
write a new labview program for each type of function you want to fit

PJ schrieb:

> Hi,
>
> Concerning the speed difference LV vs CVI:
> In my opinion CVI goes much faster. For example: I did a program in LV using
> the Levenberg Marquardt VI (GMath). I ran the program
> and found that it took 13 minutes for the VI to deliver a proper fit.
>
> Then I did a corresponding program in CVI, using the NonLinearFit function.
> I ran exactly
the same spectra with exactly the same
> guess coefficients as in LV(and the same computer of course). It took 1
> minute, 27 seconds. And that I found rather amazing, indeed.
> I tested several other types of spectras and CVI always fits 10 times (at
> least) faster.
>
> Maybe in other cases the difference is not as striking as in this curve
> fitting example, and it depends on how the algorithm is implemented, I know.
> But until someone convinces me of the contrary, also in general I choose CVI
> when I want speed. Out in the industry, I know that programmers often uses
> LV to develop something quickly, and then they use C to speed it up.
> I am open for any ideas, I'm just telling you what I've seen.
> Note: No shadow over LV. LV will always have an important position.
> Any comments of the mentioned curve fitting example would be interesting.
> Would the LV program become faster if I use the
> Application Build Toolkit to make a standalone exe.file?
>
> /PJ
>

--
Dr. Franz Jose
f Ahlers
Physikalisch-Technische Bundesanstalt
Bundesallee 100
D-38116 Braunschweig
tel +49 531 592 2410
fax +49 531 592 2405
email franz.ahlers@ptb.de

[Per Johansson]

The standard Lev-Mar.VI faster? Interesting, but of no interest since I have to change the model
function very often.
The 13 minutes example was very "easy" fit. (three clean gaussian peaks).
I some cases (with a 350 MHz PII), I have been waiting one whole hour for the fit, and that's not
acceptable at all. Without CVI in this case I wouldn't be able to do my job.

/P

Dr. Franz-Josef Ahlers wrote:

> Hi,
>
> I think it's the parsed mathematics in GMath that makes the fitting in labview
> so slow.
> You should try the levenberg-Marquardt with the built LM-Fit (maybe it is only
> available in the
> advanced analysis library??).
>
> Of course it is much more tedious to use the built in LM-Fit, since you
> practically have to
> write a new labview program for each type of function you want to fit
>
> PJ schrieb:
>
> > Hi,
> >
> > Concerning the speed difference LV vs CVI:
> > In my opinion CVI goes much faster. For example: I did a program in LV using
> > the Levenberg Marquardt VI (GMath). I ran the program
> > and found that it took 13 minutes for the VI to deliver a proper fit.
> >
> > Then I did a corresponding program in CVI, using the NonLinearFit function.
> > I ran exactly the same spectra with exactly the same
> > guess coefficients as in LV(and the same computer of course). It took 1
> > minute, 27 seconds. And that I found rather amazing, indeed.
> > I tested several other types of spectras and CVI always fits 10 times (at
> > least) faster.
> >
> > Maybe in other cases the difference is not as striking as in this curve
> > fitting example, and it depends on how the algorithm is implemented, I know.
> > But until someone convinces me of the contrary, also in general I choose CVI
> > when I want speed. Out in the industry, I know that programmers often uses
> > LV to develop something quickly, and then they use C to speed it up.
> > I am open for any ideas, I'm just telling you what I've seen.
> > Note: No shadow over LV. LV will always have an important position.
> > Any comments of the mentioned curve fitting example would be interesting.
> > Would the LV program become faster if I use the
> > Application Build Toolkit to make a standalone exe.file?
> >
> > /PJ
> >
>
> --
> Dr. Franz Josef Ahlers
> Physikalisch-Technische Bundesanstalt
> Bundesallee 100
> D-38116 Braunschweig
> tel +49 531 592 2410
> fax +49 531 592 2405
> email franz.ahlers@ptb.de

[Paul Rowntree]

Hello All;

The need to recreate a new LV program for each fitting model IS tedious ... If
anyone is interested, I have adaptedthe stock version to use a loadable VI
reference, and am creating a growing list of model vi's; as long as each model has
a common design for the connector panel, it can load up any model on-the-fly. My
program uses a very general connector that can be used for any mathematical
function, as well as allowing selected parameters to be fixed, calculates
parameter uncertainties, etc. It seems to work well ...

Best wishes

Dr. Franz-Josef Ahlers wrote: Of course it is much more tedious to use the built
in LM-Fit, since you

> practically have to
> write a new labview program for each type of function you want to fit
>
>
PJ schrieb:
>
> >

[sahilga]

hey guys,

LabWindows/CVI is a C language compiler that is made by the same people who are making LabView. ... You can control the same hardware as LabView, 100% compatible. There is cases where an algorithm is more expressive in C language than LabView.

thank me later..

[rolfk]

Resurrecting a almost exactly 20 year old thread? Way to go mate!

[RavensFan]

Well, he just joined the forums today.  So maybe he is starting at the beginning and is going to read and comment on all nearly 1.1 million messages in the LabVIEW forum until he catches up! 

[altenbach]

---

@sahilga wrote:

LabWindows/CVI is a C language compiler that is made by the same people who are making LabView. ... You can control the same hardware as LabView, 100% compatible. There is cases where an algorithm is more expressive in C language than LabView.

---

All three sentences are taken verbatim from an old article found on the internet. This forum typically favors original content and if outside sources are quoted, should include proper links and references.

---

@sahilga wrote:

thank me later..

---

Sorry, we cannot do that, because you did not compose any of it.

Typically, the quality of information is suspect unless it contains the term "LabVIEW" in the correct lettercase. Something to keep in mind! 😄