FreeBSD Port: setiathome-3.03_7

[Peter Schultz]

Kris Kennaway wrote:
> On Sun, Apr 20, 2003 at 05:13:52PM +0200, Dirk Meyer wrote:
> 
>>Hallo Peter Schultz,
>>
>>
>>>I have been given the source code for setiathome-3.08 to build an
>>>updated version for FreeBSD.  Since I'm running 4.8-STABLE, I'd like to
>>>add support for all versions of FreeBSD i386 2.2.5-4.x, as has been done
>>>with the current binary. I'm sure there must be numerous ways to do 
>>>this, and I'm hoping someone can advise me.
>>
>>If you want to support FreeBSD 2.2.x you must create a.out binaries.
>>Best way to do this is on a FreeBSD 2.2.x System.
>>modern FreeBSD will run will be able to run this stil.
>>Not sure how good cross-compiling for a.out works,
>>but I can test on 2.2.8 here.
> 
> 
> Support for a.out binaries on 5.0 requires a kernel compatibility
> option.  A native 4.x/5.0 ELF version (as well as the 2.x a.out
> version) would be better.
> 
I'll just build 4.8 and 5.0 versions then and see how that gets by.

> 
>># file /usr/local/sbin/setiathome
>>/usr/local/sbin/setiathome: FreeBSD/i386 compact demand paged executable
>>
>>It would be nice to have some versions optimzed ...
> 
> 
> Yes.
> 
$file setiathome-3.08.i386-unknown-freebsd4.8
setiathome-3.08.i386-unknown-freebsd4.8: ELF 32-bit LSB executable, 
Intel 80386, version 1 (FreeBSD), for FreeBSD 4.8, statically linked, 
stripped

I agree, this application is more for platform advocacy than it is 
anything else these days.  I use it to burn in machines because it helps 
peg the proc and memory.

Yesterday I built it statically with optimized libc and libm.  For libc 
I used `-O3 -pipe', because it wouldn't compile with the flags I used 
for libm and setiathome: `-O3 -fno-exceptions -fno-rtti -funroll-loops 
-fomit-frame-pointer -fno-strict-aliasing -fno-common -pipe'.  Since I 
normally use `-O -pipe', I'm hoping there are people here who can give 
me optimization suggestions.  Would making it i686 only help?

So, I got a report back from a guy who tested it.  He normally runs the 
linux binary and said that it's still about a half hour faster.  If 
you've run setiathome much then you know a half hour difference is good 
because the work units vary.  I gave him a reference unit used to test 
the clients.  Anyway, one would have to use the reference unit on both 
to get a fair time trial.

Anyone interested in testing please e-mail me.

Thanks,
Pete...