DonationCoder.com Software > Drag&Drop Robot

Major CPU flaw?

(1/1)

Rasczak:
G'day.

First. I've gotta say: This program has great potential! A really convenient program for turning my server into a batch-robot, without the hassle of making batch-files.


The problem:
I've got dual-cores on all my machines.
When I start DD-robot, it uses one whole core! Even when doing multithreaded operations (using both cores for batch-work) - DD-robot is competing with the other programs for cycles. This slows down operations tremendously.

Solution:
By holding the mouse-button down inside - or otherwise drag/move - the window of DD-robot, the percent of CPU-cycles used instantly drops to zero.
What causes this? Seems like there's a major flaw within the app framework.


Here's two images showing the action.
The first shows DDrobot in action without any involvement from my side.
In the second image, I'm dragging the window.

Major CPU flaw?Major CPU flaw?

Major CPU flaw?Major CPU flaw?

mouser:
there are too possibilities here.. either i have forgotten something and am using a huge amount of cpu monitoring the launched application.. or more likely, what's happening is that D+D Robot is being "blamed" for the cpu use of the program you are launching.

In other words, i suspect that cpu usage you are seeing is really the cpu use of the launched program -- it's just being assigned to D+D Robot because of the way D+D launches the program.

Rasczak:
Initially, I also thought that D+D was only displaying CPU usage, because of the processes being run under it.
However, the difference between running commands through D+D and directly through CMD is noticably

I did a test with PAR2 just recently. By running the command thrugh D+D, it took 48,8 sec to get from 0,1 to 0,5 percent. The same process took 36,2 sec by running the command directly.
D+D fluctuates between 0 and 50% cpu usage (50% is one core). Actually, it was at 0% CPU usage about 60% of the time. The difference is also noticeably, in the sense that the percentage of work done, increases more rapidly when run directly.
Also, if the process were 'hiding' under D+D, it makes no sense that moving/clicking the D+D window would instantly drop CPU-usage to zero.

Something is definitely wrong.

f0dder:
Mouser, what kind of "monitoring" of the launched app are you doing?

If it's simply to wait for termination, I hope you're doing WaitForSingleObject and not polling GetExitCodeProcess... :)

mouser:
Ok so it does sound like something is going wrong with D+D robot.  It's been a while since i was in the source code -- i'll take a look and see what i see and report back.

Navigation

[0] Message Index