]> git.openfabrics.org - ~shefty/librdmacm.git/commit
projects / ~shefty / librdmacm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 81e15fa) | patch
rsocket: Reduce SQ from 2 SGE per WR to 1 SGE
authorSean Hefty <sean.hefty@intel.com>
Thu, 24 May 2012 21:36:41 +0000 (14:36 -0700)
committerSean Hefty <sean.hefty@intel.com>
Sun, 27 May 2012 23:55:13 +0000 (16:55 -0700)
commit87b32f6b164477e6d1917d1a4b7a52e7187f0ea9
treef72cf21aae6b04bdcb43ae41b0f2c25c6ac1d7dbtree | snapshot
parent81e15fa8ac976f85a9cc0842564ad0edd5744311commit | diff
rsocket: Reduce SQ from 2 SGE per WR to 1 SGE

We currently request 2 SGEs per WR when allocating a QP.  The
second SGE is only used when sending data at the end of
the circular send buffer and the start.  All other sends are
restricted to a single SGE.

Reduce the size of the SQ by only requesting 1 SGE per WR.  The
resulting performance is basically unaffected.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
src/rsocket.c diff | blob | history
RDMA CM