Stereo Tool

Oops! I think I forgot to upload them. They are there now.

It is not broken , it`s there .
And this time the redesign is OK . it is like before. yes it have almost no load - a kind of 1% of total load . But multicore option still causes load jumps . if only one core is used, the load is about 3-4 % less. I do`t know if this is the right behavour.
I will restart the machine to see how it will be and will tell you if there is some difference.
P.S: After restart i got one time a little CPu load , one time big load . No difference and looks like the less CPU load is fact now. I can see it.

cpu power is 1% down, sometimes... it pumps from 23 to 27 28 back to 24 and so on..

_________________
Sexy Girls in your city

Chris ,
did you try to turn off the milticore checkbox and then again to tell us what you see ?

I think the "new" design of the clipper is sounding better in the highs , and the problem is solved . So now i can say : Well done !

I see no big DIFF Doggy just +1%, cause i have affinity to 2 cores from 4 i have.

_________________
Sexy Girls in your city

So the same thing like me .
sometimes go +1% on one core.
Hans said he will look into this multi-thread code soon.

About non-constant CPU usage.

Task Manager takes 'snapshots', it doesn't continuously measure what each process and each core is doing. In the past in Windows XP, I have seen occasions where changing a single line of code could cause the measured CPU load to change from 0 to 30% - and 0% (with a complete preset) is very unlikely. I'm not sure how 'perfect' the measurement is in Windows 7.

If you run everything in a single thread, then the CPU will just be busy for a while - that period is nearly constant, so TaskManager *should* show a roughly constant CPU load.

If you run things in 2 threads, then both threads will have to wait for the other one very often. And if Task Manager still takes snapshots, then even if the CPU load is constant, the measured values can vary wildly.

Having said this, the current code can be improved as well. At this moment, each filter that supports multi-threading creates a separate thread for each channel of audio. (Composite clipper is an exception, it creates 2 threads). If Windows organizes things properly, then the main thread is executed on one core, and (assuming 2 channels) the other thread - for each filter - is ran on another core.

The improvement that I want to make is to create all these threads once and use them for all the different filters. That should actually be pretty easy. That way, only 2 threads (assuming 2 channels) are running, which makes things easier.

I understand almost but the question here is why when enable the multicore the CPU Load goes UP. Is it the right behavour? And do the Multicore is really jumps like that or not ( just Task Manager is "lying" us )

WOW ! I just saw that FM processing is load about 15% . GUI up - with FM 28-30%, without FM 13-15%
I can`t remember how it was with older versions , meaning before 7 but may be it was not that much . What do you say for this one Hans?

It should not go up much (a bit from the synchronizing etc., and probably because accessing the data from a different CPU core means that it needs to move to another CPU cache).

Jumps could very well be Task Manager. Actually I never see such big jumps - but I've reduced Task Manager's update speed to the slowest setting.

Anyway, I'l try to move everything into a single 2nd thread instead of a whole bunch, we'll see if that helps.

Do you use onboard graphics? That can cause the jump maybe.