Re: [Discuss-gnuradio] Re: [Commit-gnuradio] r10410 -gnuradio/trunk/gnuradio-core/src/lib/missing

[Don Ward]

Michael Dickens wrote:


> On Feb 10, 2009, at 11:17 AM, Don Ward wrote:
> > As far as I can tell, MinGW has no standard aligned-memory
> > allocation functions. Can we use malloc16.c (in gnuradio-core/.../
> > general) or something like gcell/lib/runtime/gc_aligned_alloc.cc
> > instead?
>
> malloc16.c requires the use of both the specific allocating
> (malloc16Align () and calloc16Align ()) and freeing (free16Align ())
> routines to do what it does, but it would work (even though it might
> waste a bit of memory when allocating, in order to guarantee both
> alignment as well as a means for storing the actual allocated pointer
> value) ... This could be an option for certain issues.
>
> gc_aligned_alloc () uses posix_memalign (), so that's not an option.
>
> A better solution makes use of some built-in memory-alloc-with-
> alignment function and uses the standard "free ()" to dealloc.  That's
> what the posix_memalign () function is meant to, and currently does,
> provide.

Agreed, but as far as I can tell, Windows does not have an aligned memory 
allocation function that works with free() so this solution is not 
available.

> The best way to do this would be to add on to the posix_memalign.cc
> file for the specific OS and how it does allocating.  Use the
> AC_CHECK_HEADER macro to check for the correct headers, then
> AC_CHECK_FUNCS to make sure the functions exist and are callable.
> Then #ifdef around the appropriate functions after the end of
> HAVE_VALLOC in posix_memalign.cc for how this OS does it.

How about a test to automatically disable components that require 
posix_memalign during configuration?

Thanks,

-- Don W.