Problem with C# Testing tool

[James dean]

I am using devpartner to try to locate the slow areas in my application.
I am using the free community edition version. Devpartner slows down the
application alot when i run it with my application. I want to know is
the information gathered by devpartner actually correct. What i mean is
if i run my application without devpartner it is obviously much slower.
Does this have any affect on the actual results that i get. The way i am
using the tool at the moment is when i want to measure for exaample how
long it takes to draw a Line i take the following steps:

I use the "Clear all(F8)" button to clear the data before i start
drawing my line.

Then i draw the line.

Then i go back and take a snapshot of the data(F7).

The idea that in between clearing the data and taking a snapshot of the
data all the functions used will be checked and the slowest ones will be
located. At the moment i look in the "% with children" column to check
for bottlenecks in the method.

What i want to know is.......Is the above method correct. If it is not
could somebody please inform me the correct way to use Devpartner.
Thanks in advance

 

[Frank Schmitt]

I am using devpartner to try to locate the slow areas in my application.
I am using the free community edition version. Devpartner slows down the
application alot when i run it with my application. I want to know is
the information gathered by devpartner actually correct.

I don't know about DevPartner, but every Profiler I've used until now
slows down the application it profiles - sometimes by an order of magnitude.

What i mean is
if i run my application without devpartner it is obviously much slower.
Does this have any affect on the actual results that i get.

It shouldn't - otherwise profiling would be pretty useless, wouldn't it?

The way i am
using the tool at the moment is when i want to measure for exaample how
long it takes to draw a Line i take the following steps:

Huh? You want to measure the time it takes to draw *one single line*?

Either, your application is slow - then profile a *complete run* of your
application, using it exactly how you'd use it without the profiler; or,
your application isn't slow - then why bother profiling at all?

The idea that in between clearing the data and taking a snapshot of the
data all the functions used will be checked and the slowest ones will be
located. At the moment i look in the "% with children" column to check
for bottlenecks in the method.

% with children is good, but look at % without children, also - might be
some low-level gets called a lot from different points in the application.

HTH & kind regards
frank

 

[James dean]

Frank,
Thanks for the advice. The "Drawing one line" test was only an example
just to explain how i would test a function. I actually discovered the
problem. I did not have the "Exclude system images" checkbox checked in
the "Exclusions" settings. Its now working much faster. Problem before
was it was going into Gdi+ methods etc... also as well as the
application methods.....as u could imagine this caused it to be much
slower.
Thanks,
James