Does EVE has denormal problems? (problem solved)

Official support for: nusofting.net
RELATED
PRODUCTS

Post

[DELETED]

Post

[DELETED]

Post

ttoz wrote:ok, so i have come with a negative attitude before, when i had problems of course. but that didn't stop me from buying, every single vsti you have EVER made, and ALL you reaktor synths (when they were first released), and ALL your sync modular synths, and ALL your infinity patches, and even presets at dashsounds before it closed. since I have spent close to $1000 USD with dash, look at THAT, before you think I am trying to flame you :wink:
Yes, but haven't I been nice to you all the time, and helped you and even changed specs of products mostly just for you? :wink:

Anyway, I'm looking this thing over and over, to see if there could be something more than what we though. Don't worry, we will deal this over.

Are we OK again? 8)

Best Regards, WilliamK

Post

ttoz wrote:i hit play, the 5 clips played, cpu was at about 12%. then, when the clips reached their end, and the cursor went into silence, cpu momentarly jumped to 40%, then to nothing. so what it means is, that the effect is compounded by the amount of instances you use
This indeed shows a denormal problem. ;) Now I think you got it right, I mean, you got where it happens. Still, the code allready checks for denormal, strange. I did the same test here, it should show 12% usage, drop to 8% usage and to 1%. How fast it goes to 40% and to 0%? Maybe the denormal check I do is not lower enought...

I will need your help in order to find if it is indeed a denormal problem or something else. Is that ok with you?

If so, turn off all Effects on all EVE instances, does the 40% spike still happens? (to turn off the effects, just turn all Echo and Modulator knobs to zero.)

Wk

Post

[DELETED]

Post

[DELETED]

Post

ttoz wrote:you have been nice to me MOST of the time :lol:
"I'm only human..." :lol: At least I try to be, and I keep in trying. And I see that you also try to.
ttoz wrote:dude, i have stuck up for you many times in the other threads..as far as "are we ok", when were we NOT okay? we're not kids, this is just talk, not stuff we hold a grudge for for the rest of our lives :wink: btw, did you read the above test with multiple instances?
Hey, I'm a big kid, don't call me not a kid!! :lol: Ok, just to break the ice...

Read my post above. I think you did find a problem now. ;-)

Here is late now, 1:00 am, that's why I'm a bit grungy. But is just that I'm honest all the time, if I'm upset or happy. :P I may have to work on this in the morning, but don't worry, we will solve this thing up. It just occoured to me. LinPlug also had the same problem, on the end, it was due to some new P4 processors that has even worse low-float-point math processing. So it needs special denormal code. I have to check that out and see.

Best Regards, WilliamK

Post

ttoz wrote:...so it doesn't mater if you've pressed on note or 10..does that make sense?
Yes, actually it does. It could be the effects module. The way we check for IDLE on each instance. :o Is good you did that test, so is easier for me to track things out. But I still need you to do the test above, can you do that quickly?

Wk

Post

Ttoz, here is the list of quick tests I want you to do, so I can find the problem here. ;)

1) turn off DEGRADE on the TapeDelay (Echo) Effect. (set DEGRADE to ZERO) does that stop the 40% spike?

2) turn off echo send on all layers and instances. does that stop the 40% spike?

That's all.

Wk

Post

WilliamK wrote:Ttoz, here is the list of quick tests I want you to do, so I can find the problem here. ;)

1) turn off DEGRADE on the TapeDelay (Echo) Effect. (set DEGRADE to ZERO) does that stop the 40% spike?

2) turn off echo send on all layers and instances. does that stop the 40% spike?

That's all.

Wk
William, it is the delay module that's doing it. Spends way too much time processing content below the noise floor. I see it struggling on -70 to -90 and below material. If the delay moduel is off, it works fine, no CPU spikes at the end.

Devon

Post

[DELETED]

Post

ttoz wrote:OK! i just sepnt the last hour on it touching all various knobs and I can indeed confirm 100% it's the echo:) i also managed to produce it a couple of times with the "chorus" effect in the modulator. the other mod fx seemed ok.

by the way, turning degrade to zero does not solve it, you have to turn all the echo sends off..

i say this is good news! because we now know for sure what the problem is! see, told you it was a denormal :wink:
Got to admit, it's the smallest denormal issue I've seen at least. :) I'm used to 40%, or 60% or just TOTAL cpu spike. :)

Devon

Post

[DELETED]

Post

Hy guys, I will take a closer look into the Echo code, to see what could be producing lower float-point values (called denormal). Just keep in mind one thing Ttoz, this is not exactly our fault, but Intel fault for doing a processor that can't handle small numbers. :lol: But we know, that we need to be carefull with all denormal isues.

Anyway, I will check the code this morning and release a beta version so you guys can test it out.

Best Regards, WilliamK

Post

WilliamK wrote: Anyway, I will check the code this morning and release a beta version so you guys can test it out.

Best Regards, WilliamK
And the sustain pedal fix too? :)

Devon

Post Reply

Return to “NUSofting”