cat /proc/cpuinfo. Your best friend anytime you feel like cursing your
stupid old Pentium 2 for not having SSE.
Now, normally, if you use an SSE2 instruction on a processor without
SSE2, you get the highly popular Illegal instruction error message (and
the SIGILL signal). Since this does not happen with Orange Box SRCDS
(henceforth OBSRCDS), there are two ways to explain this:
1. At runtime, depending on what the CPUID instruction tells OBSRCDS
(you can read up on CPUID in the Intel and AMD processor manuals), a
mathematics function set is chosen to reflect hardware capabilities (no
SSE2? Pick the FPU. No FPU? Use the non-hand-optimized C/C++ functions
and hope that due to some magic, they do work). However, since probably
every QA/testing machine at Valve has SSE2, the engine never uses the
more "primitive" math functions, so the bug which probably exists in
them can't be found. (The way around this: comment out the
CPUID/choosing code and force it to use the hand-optimized FPU or the
compiler-optimized C/C++ floating point functions. However, I don't
think we have access to them. The other option would be to hack the
operating system to not report SSE2 support when OBSRCDS issues CPUID,
but that's way over my head and I'm not sure this place is populated
with kernel hackers.)
2. It's not SSE2 at all.
I guess we just gotta wait for Guillaume to answer if his processor can
do SSE2, and then we can jump to conclusions.
~~ Ondra
P.S. Depending on the compiler and the flags passed on to it, C/C++
floating point arithmetic for Intel machines is either turned into FPU
or SSE2 instructions, but I'm guessing that Valve has a few
hand-optimized routines.
P.P.S. The previous SRCDS (for Source SDK Base) did go SIGILL on my
Pentium 2, which has no SSE. Not sure if OBSRCDS does that too (I don't
plan hosting TF2 until I get the Orange Box, so I'm postponing the
server download), but if not, it means it chooses the math functions as
described in point 1. of my extremely long rant.
On 26.09.07 4:50 Uhr, Guillaume Parent wrote:
The processor is a AMD 2700+, I'm not sure if that supports SSE2 or not.
Probably not.
FWIW -- We do NOT have SSE2 support on our server... we have SSE, but
that's it... It sure sounds like if the server is lacking SSE2, for
now, SOL...
----- Original Message ----
From: Guillaume Parent <[EMAIL PROTECTED]>
To: hlds_linux@list.valvesoftware.com
Sent: Tuesday, September 25, 2007 9:09:37 PM
Subject: Re: [hlds_linux] TF2 32 slots???
Well at least I'm not the only one having the problem it seems.
*Waves at fellow Bad Bone Weight server admin*
I just want to be able to run the darn TF2 server in Linux! GWAH!
(Still
having Bad Bone Weight errors -- I wish someone from Valve would at
least
say something about this..)
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of George
Witczak
Sent: Tuesday, September 25, 2007 06:14 PM
To: hlds_linux@list.valvesoftware.com
Subject: [hlds_linux] TF2 32 slots???
Ok someone please fill me in on how to get this going? Below is one
of the
IP's and the only 2 servers in game monitor playing 32
slots......and of
course they are full!
70.87.162.59:27015
http://www.game-monitor.com/GameServer/64.182.92.161:27016/THE_TEMPEST_2_247
_DUSTBOWL_DALLAS_TX_-_WDFNews.com_Clan_p.html
http://www.game-monitor.com/GameServer/70.87.162.59:27015/3PointGames.com_TF
2_32_Player_DustBowl__1.html
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux
