> From what I've seen so far the bigger problem than contention in the
> lwlocks itself, is the spinlock protecting the lwlocks...
Postgres 9.3.1 also reports spindelay, it seems that there is no contention on spinlocks.
PID 21121 lwlock 0: shacq 0 exacq 33 blk 1 spindelay 0
PID 21121 lwlock 33: shacq 7602 exacq 14688 blk 4381 spindelay 0
PID 21121 lwlock 34: shacq 7826 exacq 15113 blk 3786 spindelay 0
PID 21121 lwlock 35: shacq 7792 exacq 15110 blk 3356 spindelay 0
PID 21121 lwlock 36: shacq 7803 exacq 15125 blk 3075 spindelay 0
PID 21121 lwlock 37: shacq 7822 exacq 15177 blk 2756 spindelay 0
PID 21121 lwlock 38: shacq 7694 exacq 14863 blk 2513 spindelay 0
PID 21121 lwlock 39: shacq 7914 exacq 15320 blk 2400 spindelay 0
PID 21121 lwlock 40: shacq 7855 exacq 15203 blk 2220 spindelay 0
PID 21121 lwlock 41: shacq 7942 exacq 15363 blk 1996 spindelay 0
PID 21121 lwlock 42: shacq 7828 exacq 15115 blk 1872 spindelay 0
PID 21121 lwlock 43: shacq 7820 exacq 15159 blk 1833 spindelay 0
PID 21121 lwlock 44: shacq 7709 exacq 14916 blk 1590 spindelay 0
PID 21121 lwlock 45: shacq 7831 exacq 15134 blk 1619 spindelay 0
PID 21121 lwlock 46: shacq 7744 exacq 14989 blk 1559 spindelay 0
PID 21121 lwlock 47: shacq 7808 exacq 15111 blk 1473 spindelay 0
PID 21121 lwlock 48: shacq 7729 exacq 14929 blk 1381 spindelay 0