Re: [Audacity-devel] [audacity] r11215 committed - Added an sse enabled mixed-radix-fft implementation and a menu/view/{n...

[Steve t. F. <ste...@gm...>]

Having tested the new time-stretch (on Ubuntu 10.10),
after applying the #include <stdlib.h> fix,
I've not found any problems with it and I can confirm that the quality
of time-stretch is improved (it is slow but I'm impressed by the very
good quality). Hopefully it will not be too long before this
enhancement can be reintroduced.

Steve

On Thu, Jul 14, 2011 at 6:30 PM, Michael Chinen <mc...@gm...> wrote:
> On Thu, Jul 7, 2011 at 10:09 AM, Clayton Otey <cla...@gm...> wrote:
>> Here is a summary of my changes from r11214 & r11215 which I reverted in
>> r11217
>> To preface:
>> I would like r11214, the TimeScale/sbsms changes to be considered for
>> acceptance now, despite the feature freeze.  Please comment.
>> r11215, the mixed radix fft / spectrogram window adjustment shortcut can
>> wait.
>
> Hi Clayton,
>
> Since this is a week old and no one has replied, I will give my thoughts on it.
>
> I think we can't take either patch right now because with the feature
> additions and functionality changes it adds there is no clear reason
> why it should go in despite the feature freeze.
>
> Right now we are sacrificing new features and better functionality for
> stability and bug fixes.  We would need more reason than just that you
> want the code to go in because it is improved.
>
> On the other hand there is only one bug blocking the stable release,
> so it might be not so long before a candidate and then actual release
> happens, at which point we should re-evaluate your patches.
>
> Even after the freeze is called off please mention on devel the
> changes before committing them, especially if you make further
> changes.
>
> Best,
>
> Michael
>
>>
>> * r11214: TimeScale improvement & gui update
>> - I synced lib/src/sbsms to libsbsms-2.0.0.  This version contains
>> significant quality improvements to time-stretching including multi-window,
>> multi-pass sinusoid tracking & synthesis, stereo phase matching & much fine
>> tuning.  I think it sounds pretty darn good, even with transient low
>> frequency material & dense polyphony.  The stereo matching is far from
>> perfect, but better than before.  If interested, I have a multithreaded
>> implementation which is currently disabled by default.
>> - The 'Sliding Time Scale / Pitch Shift" gui now has more control over the
>> rate/pitch slide
>> There are 3 "types": "linear rate", "linear stretch", "geometric"
>> There are 2 "references": "input", "output"
>> e.g. tempo slide type="linear rate", reference="input" means the rate of the
>> input will be interpolated linearly between the initial & final values to
>> produce the output
>> e.g. pitch slide type="geometric", reference="output" means the pitch of the
>> output audio will interpolate exponentially (like you'd expect in
>> portamento) between the initial & final values
>> slide type="linear stretch" means the stretch factor (1/rate) is linearly
>> interpolated, as opposed to the rate being linearly interpolated as in
>> "linear rate".
>> Link Initial/Final just forces the initial & final values to be the same,
>> for convenience.
>> Link Rate/Pitch forces the rate & pitch values to be the same, so that
>> resampling can efficiently & cleanly implement the effect.
>> e.g. A record slowing down can be simulated with 'link rate/pitch'=on &
>> slide type="linear rate" & reference="output"
>> - I've tested it pretty thoroughly on mac & win.
>> - We have (I think) fixed the linux build (just add some #include
>> <stdlib.h>).
>> - I have fixed the xcode build on my mac.
>> * r11215 - mixed radix fft & spectrogram window adjustment
>> inside #ifdef EXPERIMENTAL_MIXEDRADIX_FFT
>> Replaces the spectrum calculation with a faster sse enabled mixed radix fft
>> This allows window sizes that aren't powers of 2 to be efficiently
>> calculated
>> Added menu items:
>> Narrow Spectrum ctrl-[
>> Broaden Spectrum ctrl-]
>> which globally adjust the spectrogram window size between 8 & 32768 with 25
>> roughly equally spaced window sizes
>> Since r11215 I have almost finished implementing a per-track slider bar
>> underneath the pan/gain sliders, which shows up only in the spectrogram view
>> & allows the user to choose the spectrogram window size.  It is labeled with
>> "T" on the left (high time resolution/short window) & "F" on the right (high
>> frequency resolution/long window).  The window size in the preferences is
>> used only as an initial value & any changes made with the slider or the
>> shortcut are not written to preferences.
>> ------------------------------------------------------------------------------
>> All of the data generated in your IT infrastructure is seriously valuable.
>> Why? It contains a definitive record of application performance, security
>> threats, fraudulent activity, and more. Splunk takes this data and makes
>> sense of it. IT sense. And common sense.
>> http://p.sf.net/sfu/splunk-d2d-c2
>> _______________________________________________
>> audacity-devel mailing list
>> aud...@li...
>> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>>
>>
>
> ------------------------------------------------------------------------------
> AppSumo Presents a FREE Video for the SourceForge Community by Eric
> Ries, the creator of the Lean Startup Methodology on "Lean Startup
> Secrets Revealed." This video shows you how to validate your ideas,
> optimize your ideas and identify your business strategy.
> http://p.sf.net/sfu/appsumosfdev2dev
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>