Linux 4.6-rc2 Kernel Released

[JonathanM]

debianxfce Can you describe your performance issues with USB DACs? I'm curious.

Its not that hard for me to believe you when you say that most desktop motherboards have a decent enough DAC, but the one I have in my desktop isn't one of them.

[haagch]

Use gnome-search-tool

Oddly specific.

Anyway, I have 4 cpu cores in my laptop. Will the usb audio kernel code not run on a different core than the game if the game is using close to 100% of a core?

[Kano]

haagch

It is very unlikely that an external soundcard will slow down the system, just ignore him. He has problems with PA too, in his eyes 1 % CPU usage is too much. Of course there should not be any kernel bugs that prevent the use of your usb device.

[Kano]

Well as there is a bug report, it should get fixed, whats your problem? I know more ppl using usb audio without problems.

[JonathanM]

Also USB port is much slower than integrated audio that uses PCI.

My DAC uses USB 1.1, so the port wont be the limiting factor . I'm not going to look at kernel code right now, I was curious about your experience with it. CPU usage isn't the only factor when talking about performance, but here using USB audio raises the usage with 0.1-0.5%.

[haagch]

This haag made a bug report about usb-audio

I only made the radeon runpm bug report. The usb sound are the ones I found after looking for my problems in the bug tracker.

Of course there should not be any kernel bugs that prevent the use of your usb device.

Well, these *are* early rcs, so it's not that urgent. Still, it broke audio on Greg Kroah-Hartmann's machine - the most prominent linux maintainer, so it's weird that so broken code gets into mainline rcs...

It's also weird that we still don't have advanced static code analysis by default by now. For example there is a null checked access to quirks in the same function: if (quirk && quirk->ifnum
Seeing that reminded me of a blog post from PVS studio where they checked gtk3: http://www.viva64.com/en/b/0383/

In the function body, the programmer tests the 'completion' pointer for null and then uses it:
if (completion && completion->priv->completion_timeout) This check indicates the programmer's assumption that the pointer may be null. However, earlier in the code, this pointer was accessed without such check:

if (!completion->priv->popup_completion) return FALSE; If the pointer here appears to be null, we'll get undefined behavior. Either there is a missing check of the 'completion' pointer earlier in the code or the later check makes no sense and is not necessary.

If you have access to the AST, that is a really trivial and useful hint to give to the programmer. It's really weird that stuff like has still not arrived at mainstream...

[JonathanM]

That is away from your gaming. You do want to have all the cpu power for your game, dont you.

Even if it would cost a whole 1fps, I would still use it, but I think that my old nvidia GT 440 is the weakest link right now.

[darkbasic]

Download some kernel from kernel.org and extract. Use gnome-search-tool to find text USB_AUDIO from the linux-xxx folder. You will find several c files where is code for usb audio. That code takes many cpu cyckles. Also USB port is much slower than integrated audio that uses PCI.

LOL this is the most idiot thing I've ever heard. I did 8 channels recordings with USB 2.0 cards (M-Audio Fast Track Ultra) while doing headphones monitoring on the remaining 8 outputs, which means 16 channels simultaneously. Every single channel was mixed and filtered before being routed. I did even more complex setups with Firewire cards.

My motherboards since P35 chipset have had 90dB SNR. Your power amplifier is usually noisier, 85dB for example. Older brand was MSI and new ones Asus, like my alc887 example from the new Asus A88XM-E motherboard.

Judging audio quality by SNR alone is STUPID. Source: my Electronic Engineering's DSP background.

[darkbasic]

Oh dear, you use that thing in a desktop. I could have given you 5 soundblaster pci cards for free, if your motherboard audio sucks.

You have no knowledge of what you are talking about. Old soundblaster pci cards were such a shit that I don't want to touch them even with a pointed stick. I still clearly remember how BAD it sounded their in-hardware resampler. They were CHEAP for a reason.

[JonathanM]

Oh dear, you use that thing in a desktop. I could have given you 5 soundblaster pci cards for free, if your motherboard audio sucks.

Fear not, I mainly use it on my laptop.