[lkml]   [2006]   [Jan]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH/RFC] Shared page tables
On Monday 23 January 2006 18:19, Dave McCracken wrote:

> My apologies. I do have a small test program and intended to clean it up
> to send to Robin, but got sidetracked (it's fugly at the moment). I'll see
> about getting it a bit more presentable.

I created a test case that mmaps() a 20 GB anonymous, shared region, then
forks off 128 child processes that touch the 20 GB region. The results are
pretty dramatic:

2.6.15 2.6.15-shpt
--------- ----------------
Total time: 182 s 12.2 s
Fork time: 30,2 s 10.4 s
pte space: 5 GB 44 MB

Here total time is the amount of time for all 128 threads to fork and touch
all 20 GB of data and fork time is the elapsed time for the parent to do all
of the forks. pte space is the amount of space reported in /proc/meminfo
to hold the page tables required to run the test program. This is on an
8-core, 4 socket Opteron running at 2.2 GHZ with 32 GB of RAM.

Of course, it would be more dramatic with a real DB application, but that is
going to take a bit longer to get running, perhaps a couple of months by the
time all is said and done.

Now I am off to figure out how Andi's mmap() randomization patch interacts
with all of this stuff.

Ray Bryant
AMD Performance Labs Austin, Tx
512-602-0038 (o) 512-507-7807 (c)

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2006-01-25 00:47    [W:0.096 / U:0.896 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site