mpiexec patch for InfiniPath MPI

[Christian Bell]

On Sat, 15 Mar 2008, Pete Wyckoff wrote:

> In psm.c:
> 
> +    u64ptr = (uint64_t *) (data + 3 * sizeof (uint32_t));
> +    epids[rank] = ntohu64 (*u64ptr);
> 
> This u64ptr will be unaligned on 64-bit architectures.  While you
> can get away with it on opteron, it will cause complaints on ia64
> and errors on mips.  Suggest you do memcpy into a u64 then ntohu64
> on it.
> 
> Same deal here.  This is at offset 20 in a malloced block, so also
> unaligned.  I'm not particularly worked up about it, but you may
> run into trouble eventually.
> 
> +    einfo_epids = (uint64_t *) (einfo_hdr + 4);
> 

I spent more time generating test cases with torque testing rather
than reviewing the code as closely as you have.  I find them all
annoying enough to correct them.

> As far as release, I'd like to hold off until we get this mvapich
> situation figured out.  Hopefully not too long.  Then there will be
> two bright features to talk about.  If you're desperate, I will
> silently bump to 0.84 and you can ship that, but I'd rather not have
> two different 0.84s out there, so let me know.

I have a new patch I could send you but I'd like to test it first
(even though the changes look trivial).  That may be later this week,
but don't hold off if 0.84 is ready and you want to push it out.
Also, I don't think releasing 0.84 is a priority -- I think we can
now safely quote 0.84 in our docs.  Our software release requires a
lot of coordination and various non-engineering tasks before the bits
hit the web so I'd say that unless you can't get 0.84 out in the next
month, wait for whatever you need from mvapich.

thanks,

    . . christian

-- 
christian.bell at qlogic.com
(QLogic Host Solutions Group, formerly Pathscale)