Comments on: Negative MPI tags: Just. Say. No. http://blogs.cisco.com/performance/negative-mpi-tags-just-say-no/ Wed, 22 May 2013 00:09:23 +0000 hourly 1 By: Jeff Squyres http://blogs.cisco.com/performance/negative-mpi-tags-just-say-no/#comment-679292 Jeff Squyres Fri, 12 Oct 2012 19:31:52 +0000 http://blogs.cisco.com/?p=86826#comment-679292 I was not around when MPI-1 was designed, but I *suspect* that at least one of the reasons was to allow MPI implementations to send in-band control messages.

   0 likes

]]> By: Denis http://blogs.cisco.com/performance/negative-mpi-tags-just-say-no/#comment-679290 Denis Fri, 12 Oct 2012 19:27:57 +0000 http://blogs.cisco.com/?p=86826#comment-679290 Is there a reason behind using signed type for tags?

   0 likes

]]> By: Jeff Squyres http://blogs.cisco.com/performance/negative-mpi-tags-just-say-no/#comment-679219 Jeff Squyres Fri, 12 Oct 2012 14:34:50 +0000 http://blogs.cisco.com/?p=86826#comment-679219 Absolutely! Heck, I can’t even cook a reasonable dinner for my family without at least 2^17 or 2^18 tags.

:-)

   0 likes

]]> By: Geoffrey Irving http://blogs.cisco.com/performance/negative-mpi-tags-just-say-no/#comment-679214 Geoffrey Irving Fri, 12 Oct 2012 14:28:14 +0000 http://blogs.cisco.com/?p=86826#comment-679214 Indeed, to be completely portable you have to be ready to use no more than 2^15 tags, which is far, far too few for many useful protocols.

   0 likes

]]>