Re: [LAU] 64-bit improves audio performance? - Linux-audio-user

18 Dec 2009

Are you sure it's 64 bit that is actually making an improvement. 
If the distros are the same then this maybe a fair comparison, however if they
are different distros being compared I doubt that the performance difference is a
result of 64 bit versus 32 bit. 

I have run both and not been able to notice any substantial difference with
almost identical Gentoo installs. Even on a system with 4GB of RAM using highmem,
there is no perceivable difference.

My installs are tuned for audio work, I haven't had issues with latency for a
number of years now..  

On Fri Dec 18 15:50 , "Jonathan E. Brickman"  sent:

...
 Well.  No change in Jack behavior with LXDE, no change
from total 
removal of Pulse.  Now I really wonder.  64-bit has definitely upped my 
GUI speed, often tripled my WWW speed -- but it has either not touched 
audio (synth and Jack) performance and latency, or has hurt it 
slightly!  Anyone seen the same thing?

J.E.B.

    Should CONFIG_HPET_EMULATE_RTC be set to n on a
realtime kernel?  Is
 there something else which might invalidate CONFIG_HPET=y ?

          the RTC device is unrelated to HPET. your kernel can have HPET support
 but if your h/w doesn't, you don't get /dev/hpet. My motherboard, for
 example, does not have an HPET device.

        That is interesting.  I am interested principally because I'm seeing
 xruns and am hunting for causes, and that stood out.

 Checked the BIOS; HPET is there, already turned on.  Running the
 vanilla-install Debian Testing (AMD64) kernel, package
 "linux-image-2.6.30-2-amd64" version 2.6.30-8, I do have a /dev/hpet.
 Running any of six or seven slightly different but very clean rtlinux
 builds (vanilla kernel source of 2.6.31.6, plus rtlinux
 patch-2.6.31.6-rt19.bz2), .config options verified and reverified very
 carefully, I don't have a /dev/hpet.  Anything I should check?  Do you
 think I should get on a kernel dev list?

 But I understand now that hpet may have little or nothing to do with the
 xrun problem.  At least part of the symptomatology, is that Pulse
 talking to Jack on 64-bit Debian Testing / Gnome with GUI sound events
 off, seems to eat a whole lot more of Jack's DSP capacity than the same
 combination on 32-bit / LXDE.  On 32-bit, Pulse at idle ate zero CPU;
 now on 64-bit, Pulse at idle is eating about 2%.  I'm wondering right
 this minute if Gnome keeps its default sound open, delivering full-bore
 (albeit silent) audio even when it's told not to do so.

 I suppose I'll try LXDE.  But any suggestions will be very much appreciated.

 J.E.B.

 _______________________________________________
 Linux-audio-user mailing list
 Linux-audio-user(a)lists.linuxaudio.org
 http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user

_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
) 