Re: LEG: Optimisation and porting - assembly
* Andreas Tille <tille@debian.org> [2013-04-04 08:44]:
Thanks for your analysis. I wonder whether it might be feasible for
Praat upstream to integrate the adaptations to the according upstream
libraries. If I where the Praat author I would really try to avoid
maintaining separate code for common libraries. So perhaps the code
analysis for assembly code might have some positive side effekt to step
by step getting rid of these patches and stick to the original upstream
libraries.
That seems difficult to happen. See the files:
external/espeak/READ_ME.TXT
external/glpk/READ_ME.TXT
external/flac/READ_ME.TXT
external/portaudio/READ_ME.TXT
external/mp3/mp3.h
My understanding of the development of Praat is that the authors tend to
be quite idiosyncratic, and that for two reasons: first, Praat is a tool
written by linguists and used mainly by linguists. Second, the majority
of Praat users seem to run it on MacOS and Windows. That said, my
feeling is that any Unix-related improvements (like porting to autotools
or CMake) will be hard to get integrated upstream.
I admit I do not really rise my hand to volunteer writing a portable
build system for praat but from a very quick look onto the code it does
not look that hard to port it to autotools or cmake. If upstream might
consider a patch for the build system this could even reduce the
maintenance work inside Debian in the long term.
I wrote a minimal patch for compiling Praat against libgsl0-dev that
seems to work. However, porting the building system to use the native
glpk, flac, mad, portaudio and espeak libraries seems complicated to me.
I will contact the upstream authors about this and will commit my changes
as soon as I can.
Rafael
Reply to: