Open MPI logo

Hardware Locality Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Hardware Locality Development mailing list

Subject: Re: [hwloc-devel] release status
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2009-10-02 18:00:53


On Oct 2, 2009, at 5:45 PM, Fawzi Mohamed wrote:

> about the problems with linking, passing only the handle I can
> recompile the .so, but I cannot initialize constants, at least on my
> system.
> So exporting handles works, it is having addressing directly of
> external variables that have that type that is problematic, because
> then the address is a constant, but the variable is checked by the
> liker against its size. I have seen the trick with constant size
> struct that you use in openmpi.
>

Yes, I'm guessing/assuming all my fears are moot.

> Adding functions that alloc & dealloc does not mean that the struct
> has to be purely mallocated
>

mallocated -- great word. :-)

> , I think it might be worthwhile to keep
> something size+pointer so that if the size is small say less than
> size_t the cpuset is stored where otherwise there is the pointer... or
> something like that.
>
> Indeed I would keep a minimal struct...
>

Especially with a large number of OS processor IDs, won't the size of
the array dwarf that of the struct? I think we're quibbling over just
a few bytes here in an area where performance and space really aren't
all that important...

-- 
Jeff Squyres
jsquyres_at_[hidden]