Impulse convolver/guitarix

[funkmuscle]

hey brummer, I heard you spoke to my friend Ken and he gave you some goodies...
I sent you a message in you inbox...

[brummer]

Yea, funkmuscle, it looks like a start of a partnership, . . . .
Hopefully we could see some of the acmebargig heads as linux native plugs in guitarix soon.


well done funkmuscle

[Pablo]

by the way, I have check in SVN a new MultiClient structure (guitarix_amp -> guitarix_fx) witch allow to include other jack app's (for example Rakarrack or jack-rack) in the effect chain between amp and convolver. That's only the first introduction and may be change next time.

[funkmuscle]

guitarix_amp -> guitarix_fx

and the acmebargig products are awesome... Ken has put a lot into it... He always wanted to work with linux but he said he had nothing but attitude amongst the devs... Not at him but at each other so he pulled out.

I don't even think he wanted it native vst, I think he wanted it pure linux as slv2 and ladspa.. Sorry but native vst plugins needing the SDK and a host ain't too native.

[brummer]

I don't even think he wanted it native vst, I think he wanted it pure linux as slv2 and ladspa.. Sorry but native vst plugins needing the SDK and a host ain't too native.

No, don't get me wrong, I mean a plugin amp for guitarix, the next stage is a amp chooser, were you can choose the amp to use, we have the old default amp, developed by me , additional we have now the new amp model witch is in the experimental widget developed by adegert, next step is to include them both to the main widget and let the user choose witch one he wone use.

We wouldn't port just the heads from windows VST to linux VST, that, would change nothing in the sense of free software, the VST headers aren't free and we wouldn't port them to LV2.
We would like to add some of the acmebargig heads and makes guitarix a multi amp.
That's what I talk about with Ken, ad he like this Idea to.
But note, such things needs often more time then one could imagine.

[funkmuscle]

yeah, that's what I mean.. Ken wants to do what you're doing that's why I introduced you two.
he wants to help improve guitarix because he likes what you've done and he wants to some day have his vsts be LADSPA or LV2, not native vsts.

[funkmuscle]

revision 718...WOW... I'm using only the Experimental Preamp and selected amp 1 through an IR file of a Deizel cab...amazing....

[GarryO]

Hermann, sorry for the very late reply. My CPU is AMD sempron 2800+
With the --optimization switch all the latest svn versions now compile ok,
but _no_ improvement in CPU usage at all measured with top and also qps
I tried several different settings: tubes, effects etc.
As the default compile is -o3 anyway I would expect any improvement to be very small.
memory use is +2 megs with --optimization switch.

best wishes, G.

[brummer]

Hi Garry

but _no_ improvement in CPU usage at all

mmm, so what is the output from configure when you run the --optimization flag ?
Witch flags are shown in the line "CPU version" and "C++ flags " ?

Code: Select all

CPU version                              : Intel(R) Pentium(R) 4 CPU 2.60GHz
C++ flags                                : -Wall -std=c++0x -O3 -DNDEBUG -march=i686 -mmmx -msse2 -mfpmath=sse -fomit-frame-pointer -ftree-loop-linear -ffinite-math-only -fno-math-errno -fno-signed-zeros -fstrength-reduce -pipe

like that ?

that's the configure line I use :

Code: Select all

./waf configure --prefix=/usr --includeresampler --includeconvolver --experimental --optimization

[funkmuscle]

revision 719 a is not resampling the IR files anymore. It just loads them as is and I need them at 48000. Did I mess up or miss something?

[brummer]

revision 719 resample the files on the fly, have a look at the log box on bottom of guitarix.

The opposite is the case, all the time before, the IR-files wasn't resampled, my fool, here is a Extract of the latest release notes.

________________________________________________________________________

NOTE:// Andreas have found and solved a bug in the IR-file resampler,
that means that all the files you have resampled with guitarix, are
"NOT RESAMPLED", only the header was change.
To get rid of this, please remove the resampled files and rename the
files with the extension .orig (remove .orig)
That sucks, sorry for that.
hermann
_________________________________________________________________________

[GarryO]

Hermann, my config line was

Code: Select all

 ./waf configure  --optimization  --experimental 

and the resulting C++ flags:

Code: Select all

 -Wall -std=c++0x -O3 -DNDEBUG -march=i686 -mmmx -m3dnow -msse -mfpmath=sse -fomit-frame-pointer -ftree-loop-linear -ffinite-math-only -fno-math-errno -fno-signed-zeros -fstrength-reduce -pipe

this was svn718. I monitored CPU use with top, with various combinations of FX etc enabled, and got the same figures for the optimized version as the unoptimized.

~G.

EDIT: sorry, it was not svn718, but an slightly earlier version.

[funkmuscle]

revision 719 resample the files on the fly, have a look at the log box on bottom of guitarix.

The opposite is the case, all the time before, the IR-files wasn't resampled, my fool, here is a Extract of the latest release notes.

________________________________________________________________________

NOTE:// Andreas have found and solved a bug in the IR-file resampler,
that means that all the files you have resampled with guitarix, are
"NOT RESAMPLED", only the header was change.
To get rid of this, please remove the resampled files and rename the
files with the extension .orig (remove .orig)
That sucks, sorry for that.
hermann
_________________________________________________________________________

that is right... it is doing that and what a difference it makes... it sounds way better... dude, by any chance could you just do that for JcGui??? I know you're done with it but that's a major plus if you can do that..

[brummer]

this was svn718. I monitored CPU use with top, with various combinations of FX etc enabled, and got the same figures for the optimized version as the unoptimized.

mm, may I need to rework the check for a Semperon, normal it will be sinec rev@708 -march=native, like it is unoptimized, with the additional falgs for optimized. I have change the check now, try if you get now the -march=native flag instead -march=i686

dude, by any chance could you just do that for JcGui???

No, you know, it's all free source, . . . . ., anyone could do that.
If, I would do it, then not before middle of next year.

[GarryO]

mmm, still got

Code: Select all

 -Wall -std=c++0x -O3 -DNDEBUG -march=i686 -mmmx -m3dnow -msse -mfpmath=sse -fomit-frame-pointer -ftree-loop-linear -ffinite-math-only -fno-math-errno -fno-signed-zeros -fstrength-reduce -pipe


even though

Code: Select all

 cat /proc/cpuinfo
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 6
model           : 10
model name      : AMD Sempron(tm) 2800+
stepping        : 0


and your wscript:

Code: Select all

  elif "AMD" in model and "Sempron" in model:
        cxxflags.append ("-march=native")