[pulseaudio-discuss] Merging soxr

[Andrey Semashev]

[quoted text parts reordered]

On Wed, Jan 7, 2015 at 8:48 PM, Alexander E. Patrakov
<patrakov at gmail.com> wrote:
> 07.01.2015 15:14, Alexander E. Patrakov wrote:
>>
>> I will retest with qemu (via its alsa backend) later today.
>
> Done, both for "pa" and "alsa" backends. I must say that soxr-based
> resamplers break a configuration that worked fine (except the initial period
> of several seconds) with speex-float-5. Namely, after some seconds of
> initial crackling (which is a KVM limitation), speex-float-5 produces clear
> sound for at least 5 minutes. soxr-based resamplers produce a constant
> stream of underruns.

[snip]

> export PULSE_LATENCY_MSEC=8  # gets clamped to 2 ms for unknown reason
> export QEMU_AUDIO_TIMER_PERIOD=1000
>
> nohup \
> flock -n hda.dsk \
>         qemu-system-x86_64 -enable-kvm -m 2048 -cpu core2duo -smp 2 \
>         -machine q35 \
>         -vga std -sdl \
>         -soundhw hda \
>         -usb -device usb-kbd -device usb-mouse \
>         -device isa-applesmc,osk="...(c)AppleComputerInc" \
>         -smbios type=2 \
>         -device ide-drive,bus=ide.2,drive=MacHDD \
>         -drive id=MacHDD,if=none,file=hda.dsk \
>         -netdev user,id=hub0port0 \
>         -device virtio-net,netdev=hub0port0,id=mac_vnet0 \
>         -monitor stdio \
> </dev/null >/dev/null 2>&1 &
>
> (search Google for OSK0 and OSK1, the command line was mostly copy-pasted
> from http://blog.ostanin.org/2014/02/11/playing-with-mac-os-x-on-kvm/ )
>
> If you don't have Mac OS X, you can also reproduce the first bug (which,
> BTW, also applies to -mq and -hq) with paplay:
>
> paplay --latency-msec=8 any-44100hz-stereo-file.flac
>
> Note again: there are constant underruns in this case in the pulseaudio log,
> which are not present with speex-based resamplers.
>
> Or you can just dismiss the issue, because the testcases just request
> latency that cannot be satisfied with soxr.

Do I understand correctly that you tried to enforce 8 ms (or 2?)
latency in PA? If so then I would rather dismiss it since, as we know,
soxr latency is around 20 ms, sometimes more, and you can't achieve 8
ms with it.

> Also, with PulseAudio forced to 44.1 kHz, FooBar2000 v1.2 (which uses
> DirectSound and thus, by default, resamples everything to 48 kHz) just plays
> silence (with a neverending stream of underruns in pulseaudio log) over
> soxr-vhq and works fine over speex-float-5. soxr-hq and soxr-mq also work
> fine with wine.

I didn't quite understand this test, could you elaborate? Are you
running Windows in a VM here? Which one?