[linux-audio-user] Re: Update on savihost + Vsti's under wine
markknecht at gmail.com
Tue Nov 8 21:03:11 EST 2005
On 11/8/05, Nigel Henry <cave.dnb at tiscali.fr> wrote:
> Hi Mark. Right. Here goes. This is on FC2, not Gentoo. I'll have a go on
> Gentoo when I get the time. Wine is version 0.9-1fc2winehq, savihost is
> version 1.23. I borrowed mfc42.dll from my XP install and put it
> in .wine/drive_c/windows/system32. Savihost.exe I put in the program files,
> and also put Lin-Plugs FreeAlpha synth's .dll in there. Renamed
> savihosts .exe to FreeAlpha, and it worked first go. Now FreeAlpha synth has
> a separate file for it's patches. it will only use the default patch that is
> on the synth But it works. I've been trying a lot of freebie synths, and
> notice that they use a lot of cpu. Ksim is showing, when idle, between 70 &
> 80% cpu useage, and yet the synth itself only shows less than 5% cpu useage.
> VirSyn's MicroTERA. a free synth uses 100% cpu. Not enough to freeze the
> mouse, and even though I did get some sounds out of it, it would not load the
> sound bank. Native Instruments Reactor Session One 4.0, another freebie fully
> operational synth. When installed, this has a .exe and a .dll. Running
> the .exe brings up the GUI, but the only menu item that works is the one for
> sample rate, so cannot connect usb midi keyboard or audio output. The .dll
> through savihost works fine, but again at idle, is showing 50% cpu on Ksim,
> and the synth is showing about 18.8%. The complete patch bank is available on
> this synth. Single notes play ok, but push the cpu on Ksim to about 95%.
> Chords push the cpu useage to 100%+ and become quite broken up. All in all,
> I'm very happy with the way this is going. Hardly anything worked on
> wine-20041019-1fc2winehq, but 0.9-1fc2winehq has certainly made more things
> work. One observation. I tried crystal.dll on numerous occasions, and it
> would not open. At that time only mfc42.dll was in the system32 directory.
> After installing Reactor Session One synth, and perhaps something else,
> more .dll's turned up in system32. I thought I'd try crystal synth again, and
> this time it worked. The extra .dll's are listed below.
> Here are a few of the freebie synths I tried, and which worked ok with
> Triangle I & Triangle II from http://www.rgcaudio.com
> Interruptor Dub Siren from http://www.interruptor.ch
> JXPlugins JX220 from http://www.jxplugins.co.uk
> LFObia Mocheese from http://www.electronicscene.com/elektracidz
> LRsynths Trionimox from http://www.alphastrike.de
> Psychic Modulation Pandemonium from http://www.psychicmodulation.com
> Tobybear Screamer from http://www.tobybear.de
> LinPlug Free Alpha from http://www.linplug.com
> I'd be really interested in knowing why these synths use so much cpu when just
> lying idle. Nigel.
> Machine: Aiii-friend 1.3Ghz, 1GB Ram, Audigy2 soundblaster, Built-in
> Cyberblade i1 graphics.
Hey - you've been busy!
I had a couple of minutes so I fired up Savihost using Triangle 1
and Crystal to see if I could duplicate your results. For those synths
I see (using top)
Triangle 1 - 6% usr, 6%sys
Crystal - 20% usr, 80% sys
This is on an AMD64 3000+ machine.
Since Savihost is the main app in both cases (just renamed to the
dll name) then the difference has to be something in the way the dll
is written. (duh...) How we find that I'm not so sure. What I need is
an app that shows CPU usage by thread or process I suppose.
I'll ask about that on the Wine lists and see what I find out.
Assuming that we do find out something then possibly one day the
Wine developers (or kernel developers) can do something about it.
Note that in both cases above I ended up using the OSS seting in
winecfg. The Jack interface doesn't work at all and the Alsa interface
seemed to have an awful lot of noise. The kernel is 2.6.14-rt6, Ingo's
latest and the machine is Gentoo based.
I have the Alpha demo downloaded but I haven't installed it yet.
I think we need to work on a way to report this info, and maybe a
place to store it online. I can see a Wiki where we all can enter
numbers as being of value, assuming it starys clean.
More information about the Linux-audio-user