[pulseaudio-discuss] whistle on Ubuntu 10.04

Alexander E. Patrakov patrakov at gmail.com
Fri Jul 9 09:25:41 PDT 2010


07.07.2010 00:00, Daniel Chen wrote:
> 2010/7/6<rulet1 at meta.ua>:
>>   Anybode knows how to solve this bug?
>> https://bugs.launchpad.net/ubuntu/+source/pulseaudio/+bug/475997?comments=all
>
> Just a note that I asked about this symptom some months ago in the IRC
> channel, and there was no conclusive or unified method of resolving it
> (though that should not be an indication of whether efforts should be
> expended to do so).

After some private e-mails with the original reporter, I gathered the 
following additional information. Does it match any known recently 
resolved SIMD bugs? If so, please tell the git commit ID that should be 
backported.

1) The bug can be reproduced by the original reporter if he modifies the 
LFE volume without changing volumes of other channels.

2) The bug manifests itself as PulseAudio setting every fourth sample in 
a channel to something very low, but not zero (found by setting a tee 
PCM as an output device for PulseAudio). This distortion applies only to 
odd-numbered channels.

# printing the headerless file as six columns of signed shorts
# so that each channel goes to its own column
od -s -An -w12 sound.raw

   -3340   -230  -3340  -2140  -2740  -2740
   -3281  -2148  -3281   -231  -2714  -2714
   -3248  -2127  -3248  -2127  -2688   -289
   -3301  -2125  -3301  -2125  -2713  -2713
   -3362   -229  -3362  -2131  -2746  -2746
   -3329  -2141  -3329   -231  -2735  -2735
   -3281  -2122  -3281  -2122  -2702   -291
   -3257  -2061  -3257  -2061  -2659  -2659
   -3200   -222  -3200  -2061  -2630  -2630
   -3085  -2063  -3085   -222  -2574  -2574
   -2999  -1986  -2999  -1986  -2492   -268
   -3049  -1959  -3049  -1959  -2504  -2504
   -3078   -213  -3078  -1978  -2528  -2528
   -3008  -1966  -3008   -212  -2487  -2487

I.e., every eighth value in the interleaved stream is corrupted.

3) PULSE_NO_SIMD=1 helps, as does setting the default sample format to 
something else than s16le.

4) The original reporter uses a 32-bit kernel and has these flags in 
/proc/cpuinfo:

fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx lm constant_tsc
pebs bts pni dtes64 monitor ds_cpl cid cx16 xtpr lahf_lm

-- 
Alexander E. Patrakov



More information about the pulseaudio-discuss mailing list