on my side, I always make one file library, all objects are included the same binary, thus I link only once to FFTW (or something else) and then the size it not so big when linking dynamically its always a mess with rpath at runtime for a crossplatform project. moreover you can have a better control of the version and the feature of the library you use.

cheers a

-- do it yourself http://antoine.villeret.free.fr

2017-07-11 0:11 GMT+02:00 William Brent william.brent@gmail.com:

Yeah, that was one of my dilemmas...static or shared. The individual binaries were bigger when I linked statically, so I opted for shared on Mac to make the whole package smaller. On Windows I could only figure out how to set it up as shared. But maybe it's easier to have large binaries than it is to require that people install FFTW...

On Jul 10, 2017 5:26 PM, "Antoine Villeret" antoine.villeret@gmail.com wrote:

...

Thanks for that work !

Concerning the FFTW dependency, I made a restoration tool (by porting a LV2 plugin) that uses FFTW library and the easiest way to deploy on OSX was to link statically to FFTW. I have to rebuild FFTW from scratch because there is no more static version of that library in brew, but that was quite easy so far.

Cheers

Antoine

-- do it yourself http://antoine.villeret.free.fr

2017-07-10 22:45 GMT+02:00 William Brent william.brent@gmail.com:

...

Hi all,

I've finally gotten around to making some improvements and updates to the timbreID library, so I'd be grateful for any feedback and bug reports at this point. You can currently get source code and Linux/Mac/Windows binaries via deken. Below is a short list of the main additions. One major point is that I decided to use FFTW this time around. I've managed to get that working fine on Linux/Mac/Windows, but it would be great to get any advice on how to minimize the trouble that that dependency brings up. Makefile edits and suggestions are also very welcome, especially aspects that involve linking to FFTW on these 3 different platforms.

UPDATES: Bark-based versions of all spectral features (barkSpecCentroid~, barkSpecSpread~, etc.) A cepstrum-based pitch tracker (cepstrumPitch~) An attack time analysis object (attackTime~) Spectral slope analysis objects (specSlope~, barkSpecSlope~) A waveform slope analysis object (waveSlope~) A DCT object (dct~) Various simple time-domain objects (peakSample~, minSample~, maxSample~, minSampleDelta~, maxSampleDelta~) Various conversion objects (bin2freq, bark2freq, etc.) Additional [tabletool] methods (clip, round, ceiling, floor, maximum magnitude, find zero crossings, mtof, ftom, dbtorms, rmstodb, bin2freq, freq2bin, bark2freq, freq2bark, auto-fit boundaries) Various improvements to the database/classification object [timbreID]

There is also an updated examples package, which mainly addresses the change in functionality of [timbreID]'s fourth outlet, but has significant improvements for the concatenative and timbre space examples. It also includes a new example directory for audio segmentation. You can get that at http://williambrent.conflations.com/pages/research.html#timbreID. I've also set up a GitHub repo at wbrent/timbreID, which has everything (source, binaries, example patches).

Thanks and feel free to write me on or off list. William

-- William Brent www.williambrent.com

“Great minds flock together” Conflations: conversational idiom for the 21st century

www.conflations.com

---

Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/li stinfo/pd-list

Yes, one library file is sounding more and more like the right way to go. I've never done that, so if you can offer any simple examples to look over, please do. But apart from individual binaries vs one big binary for the whole library - I'm not even sure how to link statically on Windows with the FFTW .dll. I did succeed in producing a .a file using MinGW/dlltool ( http://www.mingw.org/wiki/createimportlibraries), but couldn't get it to work in the end. If anyone who has done static linking of FFTW 3.3.5 on Windows is wiling to hear my gripes, please let me know. I can post more detailed info on the errors I got.

On Tue, Jul 11, 2017 at 11:43 AM, Antoine Villeret < antoine.villeret@gmail.com> wrote:

on my side, I always make one file library, all objects are included the same binary, thus I link only once to FFTW (or something else) and then the size it not so big when linking dynamically its always a mess with rpath at runtime for a crossplatform project. moreover you can have a better control of the version and the feature of the library you use.

cheers a

-- do it yourself http://antoine.villeret.free.fr

2017-07-11 0:11 GMT+02:00 William Brent william.brent@gmail.com:

...

Yeah, that was one of my dilemmas...static or shared. The individual binaries were bigger when I linked statically, so I opted for shared on Mac to make the whole package smaller. On Windows I could only figure out how to set it up as shared. But maybe it's easier to have large binaries than it is to require that people install FFTW...

On Jul 10, 2017 5:26 PM, "Antoine Villeret" antoine.villeret@gmail.com wrote:

...

Thanks for that work !

Concerning the FFTW dependency, I made a restoration tool (by porting a LV2 plugin) that uses FFTW library and the easiest way to deploy on OSX was to link statically to FFTW. I have to rebuild FFTW from scratch because there is no more static version of that library in brew, but that was quite easy so far.

Cheers

Antoine

-- do it yourself http://antoine.villeret.free.fr

2017-07-10 22:45 GMT+02:00 William Brent william.brent@gmail.com:

...

Hi all,

I've finally gotten around to making some improvements and updates to the timbreID library, so I'd be grateful for any feedback and bug reports at this point. You can currently get source code and Linux/Mac/Windows binaries via deken. Below is a short list of the main additions. One major point is that I decided to use FFTW this time around. I've managed to get that working fine on Linux/Mac/Windows, but it would be great to get any advice on how to minimize the trouble that that dependency brings up. Makefile edits and suggestions are also very welcome, especially aspects that involve linking to FFTW on these 3 different platforms.

UPDATES: Bark-based versions of all spectral features (barkSpecCentroid~, barkSpecSpread~, etc.) A cepstrum-based pitch tracker (cepstrumPitch~) An attack time analysis object (attackTime~) Spectral slope analysis objects (specSlope~, barkSpecSlope~) A waveform slope analysis object (waveSlope~) A DCT object (dct~) Various simple time-domain objects (peakSample~, minSample~, maxSample~, minSampleDelta~, maxSampleDelta~) Various conversion objects (bin2freq, bark2freq, etc.) Additional [tabletool] methods (clip, round, ceiling, floor, maximum magnitude, find zero crossings, mtof, ftom, dbtorms, rmstodb, bin2freq, freq2bin, bark2freq, freq2bark, auto-fit boundaries) Various improvements to the database/classification object [timbreID]

There is also an updated examples package, which mainly addresses the change in functionality of [timbreID]'s fourth outlet, but has significant improvements for the concatenative and timbre space examples. It also includes a new example directory for audio segmentation. You can get that at http://williambrent.conflations.com/pages/research.html#timbreID. I've also set up a GitHub repo at wbrent/timbreID, which has everything (source, binaries, example patches).

Thanks and feel free to write me on or off list. William

-- William Brent www.williambrent.com

“Great minds flock together” Conflations: conversational idiom for the 21st century

www.conflations.com

---

Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/li stinfo/pd-list

Thanks Antoine - got it working as a single library file now. That plus static linking to FFTW makes it pretty compact and easy to install.

On Tue, Jul 11, 2017 at 4:12 PM, Antoine Villeret < antoine.villeret@gmail.com> wrote:

There are lots of library that uses one big file for all objects : Gem, zexy, Cream library (from Cicm) amongs others. The key is to call each object's setup method in the library setup method.

Cheers

antoine

-- do it yourself http://antoine.villeret.free.fr

2017-07-11 20:43 GMT+02:00 William Brent william.brent@gmail.com:

...

Yes, one library file is sounding more and more like the right way to go. I've never done that, so if you can offer any simple examples to look over, please do. But apart from individual binaries vs one big binary for the whole library - I'm not even sure how to link statically on Windows with the FFTW .dll. I did succeed in producing a .a file using MinGW/dlltool ( http://www.mingw.org/wiki/createimportlibraries), but couldn't get it to work in the end. If anyone who has done static linking of FFTW 3.3.5 on Windows is wiling to hear my gripes, please let me know. I can post more detailed info on the errors I got.

On Tue, Jul 11, 2017 at 11:43 AM, Antoine Villeret < antoine.villeret@gmail.com> wrote:

...

on my side, I always make one file library, all objects are included the same binary, thus I link only once to FFTW (or something else) and then the size it not so big when linking dynamically its always a mess with rpath at runtime for a crossplatform project. moreover you can have a better control of the version and the feature of the library you use.

cheers a

-- do it yourself http://antoine.villeret.free.fr

2017-07-11 0:11 GMT+02:00 William Brent william.brent@gmail.com:

...

Yeah, that was one of my dilemmas...static or shared. The individual binaries were bigger when I linked statically, so I opted for shared on Mac to make the whole package smaller. On Windows I could only figure out how to set it up as shared. But maybe it's easier to have large binaries than it is to require that people install FFTW...

On Jul 10, 2017 5:26 PM, "Antoine Villeret" antoine.villeret@gmail.com wrote:

...

Thanks for that work !

Concerning the FFTW dependency, I made a restoration tool (by porting a LV2 plugin) that uses FFTW library and the easiest way to deploy on OSX was to link statically to FFTW. I have to rebuild FFTW from scratch because there is no more static version of that library in brew, but that was quite easy so far.

Cheers

Antoine

-- do it yourself http://antoine.villeret.free.fr

2017-07-10 22:45 GMT+02:00 William Brent william.brent@gmail.com:

...

Hi all,

I've finally gotten around to making some improvements and updates to the timbreID library, so I'd be grateful for any feedback and bug reports at this point. You can currently get source code and Linux/Mac/Windows binaries via deken. Below is a short list of the main additions. One major point is that I decided to use FFTW this time around. I've managed to get that working fine on Linux/Mac/Windows, but it would be great to get any advice on how to minimize the trouble that that dependency brings up. Makefile edits and suggestions are also very welcome, especially aspects that involve linking to FFTW on these 3 different platforms.

UPDATES: Bark-based versions of all spectral features (barkSpecCentroid~, barkSpecSpread~, etc.) A cepstrum-based pitch tracker (cepstrumPitch~) An attack time analysis object (attackTime~) Spectral slope analysis objects (specSlope~, barkSpecSlope~) A waveform slope analysis object (waveSlope~) A DCT object (dct~) Various simple time-domain objects (peakSample~, minSample~, maxSample~, minSampleDelta~, maxSampleDelta~) Various conversion objects (bin2freq, bark2freq, etc.) Additional [tabletool] methods (clip, round, ceiling, floor, maximum magnitude, find zero crossings, mtof, ftom, dbtorms, rmstodb, bin2freq, freq2bin, bark2freq, freq2bark, auto-fit boundaries) Various improvements to the database/classification object [timbreID]

There is also an updated examples package, which mainly addresses the change in functionality of [timbreID]'s fourth outlet, but has significant improvements for the concatenative and timbre space examples. It also includes a new example directory for audio segmentation. You can get that at http://williambrent.conflations.com/pages/research.html#timbreID. I've also set up a GitHub repo at wbrent/timbreID, which has everything (source, binaries, example patches).

Thanks and feel free to write me on or off list. William

-- William Brent www.williambrent.com

“Great minds flock together” Conflations: conversational idiom for the 21st century

www.conflations.com

---

Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/li stinfo/pd-list

-- William Brent www.williambrent.com

“Great minds flock together” Conflations: conversational idiom for the 21st century

www.conflations.com