Re: binary notebook
- To: mathgroup at smc.vnet.net
- Subject: [mg119096] Re: binary notebook
- From: Oliver Ruebenkoenig <ruebenko at wolfram.com>
- Date: Sat, 21 May 2011 06:51:10 -0400 (EDT)
On Fri, 20 May 2011, ibmichuco wrote: > Hi all, > > Would Wolfram consider an option to save notebook as a binary/ executable > file? The file would still require the presence of a copy of a > compatible version > of Mathematica via, say, a library link. I can see that this feature > would have > three benefits: > > 1) The "nbx" file would be much faster, possibly much closer to that > of a > compiled fortran/c version, at least for similar numerical tasks. This > would > make Mathematica an excellent tool both in development and in > production. You could generate a *.m file and store everything you need in that. DumpSave it to create the a *.mx file and run that as a MathematicaScript. > > 2) People are more willing to distribute propriety codes if they can > maintain > some kind of control over them. > > 3) You can send people several modules bundled in to a single file, > and > hopefully, due to the library links proposal, the file can still be > reasonably > small in size. > > Recently I have had to convert some of my codes (pure number > crunching) > that I had developed in Mathematica to f90 just to make some > calculation > practical and thought that this additional feature would have saved > me > considerable time. Why do you think that running a binary version of a notebook would improve performance for number crunching? Perhaps the Mathematica code could have been improved? Did you use packed arrays and check for PackedArrayQ, the compiler is quite useful for these kind of things. Hth, Oliver > > Regards, > > Michuco > >