You might want to look at ticket #1791 while you are doing this - Brad
added some valuable data earlier today.
On Feb 12, 2009, at 12:13 PM, Eugene Loh wrote:
> Jeff Squyres wrote:
>> This should probably include the disclaimer that we talked about
>> this extensively yesterday at the sm/btl meeting after the MPI
>> Forum meeting in San Jose.
>> Eugene has been working on new sm stuff (as he has posted), and we
>> had a very productive meeting yesterday discussing what he has
>> done and how to get it all into the trunk. Have a look at his
>> slides at the bottom of this wiki page:
>> Eugene will be replacing the current sm btl with a new one
> Let's downgrade that statement. I intend to make a series of
> putbacks to the current sm BTL to improve latency. I'm not so much
> "replacing" the current BTL as doing some optimization work.
>> that uses a single queue and various other improvements. Hence,
>> the current ompi_*fifo.h files will no longer be necessary.
>> Additionally, we resolved the "abstraction break" issues that were
>> discussed here on the list a little while ago -- we think all the
>> improvements can be done in the current architecture without
>> disrupting the abstraction barriers. This RFC is one step in the
>> process; eliminate some no- longer-necessary kruft.
>> On Feb 12, 2009, at 8:53 AM, Eugene Loh wrote:
>>> RFC: Eliminate ompi/class/ompi_[circular_buffer_]fifo.h
>>> WHAT: Eliminate those two include files.
>>> WHY: These include files are only used by the sm BTL. They
>>> are not generally usable. Further, the sm BTL will soon no
>>> longer use them. The current FIFOs support only a single sender
>>> each and we want multiple senders to be able to share a common
>>> FIFO. The current FIFOs also can grow without bound, which is
>>> complicated support that we don't think we need.
>>> WHERE: Two include files: ompi/class/ompi_[circular_buffer_]fifo.h
>>> WHEN: "Immediately". Intended for 1.3.x.
>>> TIMEOUT: February 20, 2009.
> devel mailing list