--rwEMma7ioTxnRzrJ
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

First of all,
thank you all for your answers!

* Christos Zoulas writes:
> In article <20050115171001.GA1137@enterprise.starfleet>,
> Lasse Kliemann  <lasse-list-current-users-netbsd-2004@plastictree.net> wr=
ote:

> >> sounds like your processes are being swapped to disk very often
> >
> >dmesg talks about 224 MB available memory, but I am having trouble match=
ing=20
> >this with the output of top:
> >
> >Memory: 101M Act, 56M Inact, 1260K Wired, 39M Exec, 57M File, 12M Free
> >Swap: 512M Total, 9532K Used, 503M Free
> >
> >I do not believe that Firefox is doing a lot of swapping, because the LE=
D for=20
> >the harddisk is quiet most of the time, and obviously there is very few=
=20
> >swapspace used.
>=20
> Does top show pagedaemon running?

ps ax shows such a process:

   13 ?      DK   0:01.97 [pagedaemon]

> Does vmstat 1 show paging?

I guess so. The output is like this:

 procs    memory      page                       disks      faults      cpu
 r b w    avm    fre  flt  re  pi   po   fr   sr m0 w0 c0   in   sy  cs us =
sy id
 1 0 0 156824   1108  422   6   0    1    2    9  0  5  0  111  589  77 14 =
 3 83
 1 0 0 156972    964   44   0   0    0    0    0  0  4  0  114  253  52 91 =
 9  0
 1 0 0 157004    932   16   0   0    0    0    0  0  0  0  108  195  38 100=
 0  0
 1 0 0 157176    820   39   0   0    0    0    0  0  3  0  108  359  58 100=
 0  0
 1 0 0 130804   2544 1264 14117   0    0  584 14589  0  1  0  103 2137 770 =
69 29  2
 1 0 0 130956   2392   69   0   0    0    0    0  0  2  0  104  531 110 94 =
 3  3
 1 0 0 130968   2380   31   0   0    0    0    0  0  0  0  101  194  93 99 =
 1  0
 2 0 0 130976   2372  381   0   0    0    0    0  0  1  0  165  553 103 97 =
 3  0
 1 0 0 131984   1364  310   0   1    0    0    0  0  1  0  101 1436 606 91 =
 8  1
 2 0 0 131988   1360   10   0   0    0    0    0  0 15  0  228  855 199 52 =
 8 40
 1 0 0 131988   1360    8   0   0    0    0    0  0  0  0  153  513 154 60 =
 4 36
 0 0 0 131988   1360   13   0   0    0    0    0  0  0  0  146  996 269 70 =
 4 26
 2 0 0 132100   1248   35   0   0    0    0    0  0  0  0  276 1510 416 90 =
10  0
 1 0 0 132172   1296    7   0   0    0    0    0  0  0  0  114  551 146 54 =
 3 43
 1 0 0 132172   1296   73   0   0    0    0    0  0  1  0  117 1173 290 97 =
 3  0
 1 0 0 132392   1076   65   0   0    0    0    0  0  1  0  112  391 138 94 =
 4  2
 1 0 0 133012    456  148   0   0    0    0    0  0 11  0  141  371  89 88 =
 9  3
 1 0 0 132956    512   20   0   0    0    0    0  0  2  0  140  481  92 91 =
 9  0
 procs    memory      page                       disks      faults      cpu
 r b w    avm    fre  flt  re  pi   po   fr   sr m0 w0 c0   in   sy  cs us =
sy id
 1 0 0 133004    464   18   0   0    0    0    0  0  0  0  128  331  87 96 =
 4  0
 1 0 0 133108    360   32   0   0    0    0    0  0  2  0  109   87  34 95 =
 2  3
 1 0 0 132200   1176   60   0   0    0  257  233  0  4  0  105   64  28 99 =
 1  0
 1 0 0 132300   1076   30   0   0    0    0    0  0  0  0  101   91  26 99 =
 1  0
 1 0 0 132936    440   93   0   0    0    0    0  0 42  0  151  237  71 88 =
 6  6
 1 0 0 132036   1364   30   0   0    0  272  236  0 14  0  116  213  63 94 =
 4  2
 1 0 0 132256   1144   29   0   0    0    0    0  0 15  0  112  205  72 99 =
 1  0
 1 0 0 132328   1136   10   0   0    0    0    0  0  3  0  123  226  57 98 =
 2  0
 1 0 0 132344   1124   19   0   0    0    0    0  0  0  0  108  296  56 99 =
 1  0
 1 0 0 132400   1068   86   0   0    0    0    0  0  0  0   99 1984 604 83 =
17  0
 0 0 0 132400   1068    9   0   0    0    0    0  0  0  0   98  342  85  8 =
 5 87
 0 0 0 132400   1068    8   0   0    0    0    0  0  0  0  102  123  35  2 =
 2 96
 1 0 0 132400   1068    6   0   0    0    0    0  0  1  0  108  131  37 34 =
 2 64
 1 0 0 132400   1068    7   0   0    0    0    0  0  4  0  105   69  35 100=
 0  0


> If it is paging, you can give processes more memory by running down=20
> vm.execmax and vm.filemax.

You mean decreasing these values? Well, I've put them down to 15 now withou=
t=20
any noticible effect. I've also tried increasing them with a similar outcom=
e.

> >
> >Another idea of mine is that it is connected with the LongRun support. I=
've=20
> >read that it can be configured via sysctl. Anyone knows an example to tu=
ne for=20
> >more performance this way? Energy consumption is no problem (as long as =
the=20
> >notebook does not become too hot), because it is almost always connected=
 to a=20
> >power socket.
>=20
> sysctl -a | grep longrun
>=20
> But that should not matter either.

I switched to mode 3 like suggested in another post, but it did not have an=
y=20
noticible effect.

I tried to collect some more details on the behavior of Firefox. First, the=
=20
_very_ slow graphics output (black rectangles being drawn first, before a m=
enu is=20
displayed) obviously is due to the fact that I run the application with=20
untrusted X access. Running it with full access partly gets me rid of this.=
 The=20
black rectangles are gone, but the user interface still responds with long=
=20
delays only (slower than Linux).

Even simple operations like placing the cursor in the address bar by pressi=
ng=20
Ctrl-L takes slightly below a second. A real problem is the loading of=20
webpages. For example, the first time loading www.netbsd.org takes 14 secon=
ds=20
from the moment i press enter to the moment the page is completely displaye=
d. =20
The second time I load it, it is still 7 seconds. w3m-img however takes bel=
ow 2=20
seconds (so it is not my slow internet connection). It is also much faster =
on=20
Linux, on the same internet connection. (Unfortunately, I do not have the L=
inux=20
notebook here right now to take exact times there.)

During the loading of the pages, a major part of time is spent with=20
"Transferring from ..." or "Read ...".


Thanks for your patience!

--=20
Lasse Kliemann
      private homepage: http://plastictree.net
   NO software patents: http://swpat.ffii.org
do NOT use M$ products: http://plastictree.net/articles/noms

--rwEMma7ioTxnRzrJ
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (NetBSD)

iD8DBQFB6sn/1gObwed86AkRAp7tAKCra50O77TK56cut/3M/aYhz81zEgCcCkjG
UClz84dYmx/12wzv6F/vCRQ=
=7OWu
-----END PGP SIGNATURE-----

--rwEMma7ioTxnRzrJ--