Re: Alternative Matlab/Octave compiler

[Juan Pablo Carbajal]

On Fri, Aug 21, 2015 at 12:09 AM, Shmuel Safonov <address@hidden> wrote:
> Hi,
>
> I've read the old threads
>
> http://octave.1599824.n4.nabble.com/octave-to-independent-C-code-td1630298.html
> <http://octave.1599824.n4.nabble.com/octave-to-independent-C-code-td1630298.html>
> and
> http://octave.1599824.n4.nabble.com/curious-about-octave-compiler-td1635344.html#none
>  where a question was posted about possibility of alternative m-files
> compilation and with reference to
> Stratego Octave Compiler but this product eventually seems to disappear. I
> worked by myself on such an alternative compiler with restricted syntax and
> accepted some results:
> - Matlab syntax subset is supported
> - Scalar and cell arrays are supported
> - Control constructions (if/while/for/switch) are supported
> - 2D and partially 3D matrixes are supported
> - try/catch are supported at SEH-supported platform (Windows with MSVC
> C-compiler)
> What is not supported is:
> - Arrays of structures
> - Full support of >2D matrixes. Actually the support is absent only at the
> level of real-time support library
> - Classes
> - "end"-index in arrays in deep indexed constructions (like A(end).bb).
> - GUI is not supported by I added debug output to plots (at PC).
>
>  I also created  an debugger application for simple viewing of the contents
> of variables(matrixes etc.) during EXE-file debugging with some existing
> debugger (e.g. MSVC) using the process memory viewing.
> The compiler package contains the code generator and the support library.
> The support library is very compact, much less than MATLAB MCR e.g. C-code
> extension may be added to the application easily. I tried the compilation of
> a large project (> 500 files) whose time-critical part was  written in C-MEX
> code; the performance is approximately as in MATLAB. I also succeeded to
> compile this project as a library for Android and added to it Java GUI
> provided by this OS.
>
>
> *Yet I am not sure if the issue of the alternative compilation is still
> relevant?*   Thus at this stage I only published the M-to-NASM/EXE generator
> but I am ready to get available M-to-C code generator as well if e.g. it is
> important to somebody personally. My binaries, an example and documentation
> are  here <http://csafonov.blogspot.co.il/p/tmc.html>  .
>
>
>
>
>
> --
> View this message in context: 
> http://octave.1599824.n4.nabble.com/Alternative-Matlab-Octave-compiler-tp4672195.html
> Sent from the Octave - General mailing list archive at Nabble.com.
>
> _______________________________________________
> Help-octave mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/help-octave

I haven't tried your software, but assuming it is libre software, it
would be nice to provide it as a package. Clearly some functions with
long iterations (like ode45 and some in the optim package) could
benefit from being able to compile the user-provided functions that
need to be evaluated on each iteration.

Do you think that would be a possiblity?