On a 32Mbyte box running a fair bit of NFS page colouring causes
enough fragmentation that the NFS performance goes to pieces, and
on non NFS I can't measure any difference for normal use.
Im sure page colouring makes sense on a box with 512Mb of RAM and
with fragmentation control, but not on anything that much
smaller. This is also an SMP box with low memory bandwidths - the
sort that might gain most
Time some kernel builds with the following combinations:
1) Vanilla-2.1.62
2) Vanilla-2.1.62 + page coloring patches
3) VGER from before I merged in the page coloring stuff
(this is to get a measurement with the "need_resched hack"
on SMP)
4) Most recent VGER (page coloring + "need_resched hack")
On non-SMP I was getting 7% (consistantly) faster kernel builds with
the page coloring stuff on sparc64, and this is on boxes where memory
bandwidth is good enough that it shouldn't have shown up that much.
Later,
David S. Miller
davem@dm.cobaltmicro.com