Forums » General

Downtime

«12
Jul 20, 2003 raybondo link
The client.
The server runs on x86 FreeBSD.

If the program temporarily stores pointers into ints then get them back out, you get zeros in the upper 32bits and that's bad if your pointer is above the 4gig address.
Jul 20, 2003 Cmdr. Freeman link
Is the server a Linux binary under emulation or an actual FreeBSD binary?


What revision of FreeBSD is on the server? Do you guys dabble in -CURRENT anywhere?
Jul 20, 2003 raybondo link
Heh, the game would better benefit with Altivec, SSE, 3DNow! optimizations before it would benefit from 64bit math.
Converting it to 64bit is as simple as recompiling it and making sure we don't make any 32bit int or 32bit pointer assumptions. Visual Studio 7's compiler has an option to find those for us. Using SIMD stuff is a bit more complex andn would require some reworking of some parts of the game.
Jul 21, 2003 incarnate link
The server might benefit from SSE, but we've had some issues with its use on our non-coppermine P3s. Maybe someday if we can afford new hardware.

The server-side programs are all developed and built natively on FreeBSD, which is our choice of server OS. I have 5.1-STABLE on one machine (a donated dual P3 600 that will probably become our webserver soon), built with the new SCHED_ULE non-spinlock scheduler and various other manual tweaks. That's the only machine where we're going out on a limb, for experimental purposes. Everything else runs 4.x. I'm more inclined to stick with 4.x on production machines for the moment (especially since they're all single CPU), until 5.2 or further maturation of the 5.x tree.
Jul 21, 2003 alienb1212 link
uh, you probably had more than "issues" as the non-coppermine P3's don't have SSE.......0_o
Jul 21, 2003 a1k0n link
Mar 14 18:31:44 majikthise /kernel: CPU: Pentium III/Pentium III Xeon/Celeron (551.25-MHz 686-class CPU)
Mar 14 18:31:44 majikthise /kernel: Origin = "GenuineIntel" Id = 0x673 Stepping = 3
Mar 14 18:31:44 majikthise /kernel: Features=0x387f9ff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,PN,MMX,FXSR,SSE>

shrug. CPUID says it has SSE. I compiled the server with gcc3's SSE stuff, and it bombs.
Jul 22, 2003 alienb1212 link
CPUID is making a rather general guess of what the processor is... Xeon...? i dont think so..:p
Jul 22, 2003 raybondo link
When did SSE1 come out? When did SSE2 come out?
Jul 22, 2003 alienb1212 link
Hmm....SSE1 is kinda old, 97 i think, SSE2 is more like 01


actual release dates are later than that i believe
Jul 24, 2003 incarnate link
According to all the specs I can find on Intel's site, the entire P3 line (including the older, non-cumine chips) have SSE. I have no idea why it doesn't work. There were some SSE-specific bugs in the earlier chips, maybe the kernel disables it? I dunno.
Jul 24, 2003 alienb1212 link
hmm....well technically ALL p4's have hyper-threading capability.....they actually do, and intel (used) to say they do, but it is disabled.... maybe thats the case with the non-cmine p3's also....