Bug#354486: lastfm: Profile information

[Paul Telford]

tags 354486 +help -unreproducible
thanks


Here's some more info from another user that points to src_sinc as well:


--
I got 'qmake' functional with the g++ -g3--ggdb -p -pg flags,
callgrind points to high resource usage(~91% of total lastfm usage) in
function 'calc_output' (starting line# 420) in
'./lastfm-1.1.90/src/libsamplerate/src_sinc.c'

see the 'do-while' loops for the right an left channels, start @ line
#'s 435  454

(for both the right and left channels, 2x~22%)
icoeff = filter-coeffs [indx] + fraction * (filter-coeffs [indx
+ 1] - filter-coeffs [indx]) ;
/*and at 2x~9%*/

left += icoeff * filter-buffer [data_index + ch] ;

right += icoeff * filter-buffer [data_index + ch] ;

Any ideas? without 'callgrind' running 'top' reports lastfm  30% cpu
usage, lastfm has playback problems if I attempt to max out my cpu
with another process...

With lastfm running in the valgrind/callgrind virtual cpu callgrind
grabs 99.9% cpu time, playback is *gone* but terminal output, and
continued track submission suggests that the lastfm is functioning
normaly, albet slow...
The Data that led me to the above info can be found at
http://chaboo.casa-z.org/callgrind/callgrind.out.2203
--

Bug#354486: lastfm: Profile information

[Ian Wienand]

Package: lastfm
Version: 1.1.90-2
Followup-For: Bug #354486

Hi,

I'm seeing this very high CPU usage too.

I recompiled with gprof turned on, and I see that most of the time is
being spent in sinc_process() from libsndfile.

Flat profile:

Each sample counts as 0.01 seconds.
  %   cumulative   self  self total
 time   seconds   secondscalls  ms/call  ms/call  name
 93.06 22.1122.11  85825.7725.77  sinc_process
  3.32 22.90 0.7961776 0.01 0.01  synth_1to1

This appears to be being called from WaveThread::run()

[1] 99.90.28   23.46 WaveThread::run() [1]
   22.110.00 858/858 sinc_process [2]
0.001.231033/1033decodeMP3 [3]
0.070.00 858/858 src_short_to_float_array [11]
0.050.00 858/858 src_float_to_short_array [12]
0.000.00 858/858 src_process [26]
---
   22.110.00 858/858 WaveThread::run() [1]
[2] 93.1   22.110.00 858 sinc_process [2]
---

Now considering sinc_process has a comment

  Beware all ye who dare pass this point. There be dragons here.

I'm not inclined to dive in there :) I CC'd Erik incase he knows of
anything that might cause this.  I confirmed that the src_sinc.c file
is the same as the latest libsndfile download.  

Of course, it's probably something stupid that WaveThread::run() is
doing to invoke sinc_process.

The only thing I can think is that this is on my laptop, a Pentium M.
I wonder if the other reporters are running similar hardware?

Thanks,

-i

-- System Information:
Debian Release: testing/unstable
  APT prefers unstable
  APT policy: (500, 'unstable'), (1, 'experimental')
Architecture: i386 (i686)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.16
Locale: LANG=en_AU.UTF-8, LC_CTYPE=en_AU.UTF-8 (charmap=UTF-8)

Versions of packages lastfm depends on:
ii  libasound21.0.10-2   ALSA library
ii  libc6 2.3.6-4GNU C Library: Shared libraries an
ii  libgcc1   1:4.0.3-1  GCC support library
ii  libqt4-core   4.1.1-1+b1 Qt 4 core non-GUI functionality ru
ii  libqt4-gui4.1.1-1+b1 Qt 4 core GUI functionality runtim
ii  libstdc++64.0.3-1The GNU Standard C++ Library v3
ii  libx11-6  6.9.0.dfsg.1-5 X Window System protocol client li
ii  xlibs 6.9.0.dfsg.1-5 X Window System client libraries m

Versions of packages lastfm recommends:
ii  epiphany-browser [www 1.8.5-2Intuitive GNOME web browser
ii  firefox [www-browser] 1.5.dfsg+1.5.0.1-4 lightweight web browser based on M
ii  lynx [www-browser]2.8.5-2sarge1  Text-mode WWW Browser
ii  mozilla-browser [www- 2:1.7.12-1.1   The Mozilla Internet application s
ii  opera-static [www-bro 9.0-20060206.1 The Opera Web Browser
ii  w3m [www-browser] 0.5.1-4WWW browsable pager with excellent

-- no debconf information


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Bug#354486: lastfm: Profile information

[Paul Telford]

On 3/31/06, Ian Wienand [EMAIL PROTECTED] wrote:

 I recompiled with gprof turned on, and I see that most of the time is
 being spent in sinc_process() from libsndfile.

I have to admit that most of this is over my head.  I hope that Erik
or someone else can provide some hints, because I have no clue how to
proceed.  Thanks for the valuable debugging information, however.


 The only thing I can think is that this is on my laptop, a Pentium M.
 I wonder if the other reporters are running similar hardware?

At least one other person has said they see this on a 1.7ghzP4 and
2.2ghzAthXp which seems to rule out the Pentium-M specifically.



 Paul.