Hey,

PETSc addresses the problem of dealing with many command line flags 
through a (hierarchy of) configuration file(s) (see here: 
http://www.mcs.anl.gov/petsc/petsc-current/docs/manual.pdf, Section 14.2 
"Runtime Options"). In such a configuration file the same flags are 
listed one after another as you would pass through the command line 
directly. This way one can use the same code for parsing and quickly add 
additional flags as needed.

As for a curses GUI: I wouldn't rely on this without a pure 
terminal-driven approach. For example, on Windows machines you won't 
have a curses GUI available, but you always have the command line. Thus, 
I'd always want to have configuration files, particularly as I (as a 
user) can generate them through scripts, which I can't through a GUI.

Best regards,
Karli



On 07/13/2014 05:47 AM, Philippe Tillet wrote:
> Another option I forgot to mention is having a configuration file.
>
> python autotune.py autotune.conf
>
> Where autotune.conf would contain things such as:
>
> output-path my/output/path
>
> module vector-axpy
> local-size 2 4 8 16 32 64 128 256
> num-groups 2 4 8 16 32 64 128 256
> simd-width 1 2 4
>
> module matrix-matrix-product
> ms ..
> ks ...
> ns ...
> etc ..
>
> It seems like a reasonable (and portable) solution too.
> If you were to use to auto-tuner, which option would you prefer?
>
> Philippe
>
>
> 2014-07-13 5:35 GMT+02:00 Philippe Tillet <phil.til...@gmail.com
> <mailto:phil.til...@gmail.com>>:
>
>     Hey,
>
>     So, basically, the auto-tuner is starting to have a lot of
>     parameters, and using command line options becomes really tedious.
>     Since the auto-tuner is also highly expected to run on servers, I've
>     thought about making a curses GUI to it. Now, there are two options,
>     if I want to do this:
>     - Using curses, provided in the default libraries on unix, but not
>     on windows.
>     - Using Unicurses, a portable port :
>     https://pypi.python.org/pypi/UniCurses
>     I like option (2) better. Since it already requires pyviennacl
>     anyway, it seems reasonable to add one more dependency. Any thoughts?
>
>     Philippe
>
>
>
>
> ------------------------------------------------------------------------------
>
>
>
> _______________________________________________
> ViennaCL-devel mailing list
> ViennaCL-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/viennacl-devel
>


------------------------------------------------------------------------------
_______________________________________________
ViennaCL-devel mailing list
ViennaCL-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/viennacl-devel

Reply via email to