Re: pre-2.3.7-1 fails compile

Ingo Molnar (mingo@chiara.csoma.elte.hu)
Mon, 14 Jun 1999 17:36:36 +0200 (CEST)

Andrea, we've already fixed most pagecache issues. You can find the
swapping fixes in pre-2.3.7-2-dangerous, and some other fixes are in the
pipeline as well. (most notably named mappings and there were some bugs in
buffer.c) As you probably know, this all is work in progress, so please be
a bit patient before complaining. Swapping (and everything else) is now
stable here.

On Mon, 14 Jun 1999, Andrea Arcangeli wrote:

> Then flushpage may start all the I/O before going to sleep, it may be far
> more smart simply starting I/O for _all_ dirty buffers and then going to
> wait for I/O completation. [...]

this makes a difference only if blocksize != PAGE_SIZE. We have mainly
concentrated on making the 4k blocksize stuff go like smoke - 1k can and
will be speeded up later on, but it will probably never perform as well as
4k ext2fs.

some read() scalability numbers on a 4-CPU box, using 4k ext2fs:

1 process reading a 60k file:
READ: 270.75 MB/sec

2 processes reading the same 60k file:
READ: 258.19 MB/sec
READ: 258.12 MB/sec

3 processes reading the same 60k file:
READ: 256.52 MB/sec
READ: 256.54 MB/sec
READ: 256.57 MB/sec

4 processes reading the same 60k file:
READ: 252.95 MB/sec
READ: 252.93 MB/sec
READ: 252.76 MB/sec
READ: 252.87 MB/sec

we scale almost perfectly.

as a comparison, the very same read() workload done on the very same
user-space buffer, simulated via memcpy:

MEMCPY: 277.73 MB/sec
MEMCPY: 277.68 MB/sec
MEMCPY: 277.72 MB/sec
MEMCPY: 277.65 MB/sec

this shows that we are quite close to the physical limit already. And this
is only the beginning ;)

-- mingo

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/