Re: [Gnash-dev] youtube profiling (-O2 and inlining)

[strk]

On Fri, May 04, 2007 at 05:45:13PM +0200, Udo Giacomozzi wrote:
> Hello strk,
> 
> Friday, May 4, 2007, 3:45:59 PM, you wrote:
> >>From 20.5 seconds to 8 seconds spent in movie advancement !!
> s> A far more acceptable performance now, really don't think this is
> s> an issue anymore. 
> 
> Well, I still don't like the high CPU usage with a 30x30 pixel player
> window (on a Pentium M 1.5 GHz!).

Not related to AGG anymore, it seems.

Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total
 time   seconds   seconds    calls  us/call  us/call  name
  9.72      0.14     0.14                             
FLVParser::nextAudioFrame()
  8.33      0.26     0.12                             
curl_adapter::CurlStreamFile::fill_cache(long)
  7.64      0.37     0.11                             char* 
std::string::_S_construct<__gnu_cxx::__normal_iterator<char*, std::string> 
>(__gnu_cxx::__normal_iterator<char*, std::string>, 
__gnu_cxx::__normal_iterator<char*, std::string>, std::allocator<char> const&, 
std::forward_iterator_tag)
  7.29      0.48     0.11  9225389     0.01     0.01  
gnash::NetStream::setStatus(gnash::NetStream::StatusCode)
  6.94      0.58     0.10                             
FLVParser::nextMediaFrame()
  6.25      0.67     0.09                             
FLVParser::parseNextFrame()
  6.25      0.76     0.09                             
gnash::NetStreamGst::audio_callback_handoff(_GstElement*, _GstBuffer*, 
_GstPad*, void*)
  5.56      0.84     0.08                             LoadThread::read(void*, 
unsigned int)
  4.86      0.91     0.07                             
gnash::NetStreamGst::video_callback_handoff(_GstElement*, _GstBuffer*, 
_GstPad*, void*)

About setStatus (lots of calls!) ... I keep getting status notifications even 
after the movie ended... 

In any case, loader and decoder seem to be the bottleneck now (-s 0.3)
Extreme locking seems a good candidate to look at.

--strk;