[Clam-devel] Re: asking for advice about a callgrind session (sms harmonizer)

[Xavier Amatriain]

Hernan, could you elaborate a bit more on what you are testing and what 
kind of clicks you are getting?
(maybe post a sound example and a dump of the Analysis and Synthesis 
configurations).

Hernán Ordiales wrote:
> On 8/2/07, Hernán Ordiales <h at ordia.com.ar> wrote:
>   
>> I'm sending you a callgrind session I've made from sms harmonizer
>> because and I'm still getting clicks and I couldn't find nothing worth
>> in the analysis with kcachegrind (i'm not much experienced with this
>> either)
>>
>> i've done it in this way:
>>
>> valgrind --tool=callgrind ./OfflinePlayer
>> example-data/test/harmonizer-without-visualization.clamnetwork
>> Elvis44k.wav elvis-output-test.wav
>>     
> [snip]
>
> One comment more about this, in fact now seems to work better with mic
> voice as input and with the elvis file the artifact is heard always in
> the same part of the file... maybe that could be a clue, don't? I'm
> started to think that isn't a problem about performance, my processors
> are always below 50%...
>
> I'll also try greg's patch too... how do you think is the best way to
> compare (real-time audio) performances? (with callgrind NE gets
> zombified)
>
>