----- Original Message ----- From: "Stefan `Sec` Zehl" <[email protected]>
To: "Half-Life dedicated Linux server mailing list" 
<[email protected]>
Sent: Friday, March 21, 2014 1:49 PM
Subject: Re: [hlds_linux] Running TF2/CSGO Servers on FreeBSD


Hi,

On Fri, Mar 21, 2014 at 13:06 -0000, Steven Hartland wrote:
Not invariant tsc thats unusual, are you running a really old CPU?

According to dmesg it's:

| CPU: AMD Athlon(tm) 64 X2 Dual Core Processor 5600+ (2900.03-MHz K8-class CPU)
| FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs | cpu0 (BSP): APIC ID: 0 | cpu1 (AP): APIC ID: 1
I'm not sure if that qualifies as "really old".

Ahh AMD, unfortunately their TSC doesnt reset to zero with RESET IPI before
Bulldozer.

I've put the benchmark dl:
http://blog.multiplay.co.uk/dropzone/bench_clock_realtime.c

My output is:

mcp:~>./bench_clock_realtime 1000000 | sort -rnk1
clock realtime micro-benchmark.  1000000 syscall iterations.
Avg. us/call    Elapsed         Name
1.557101        1.557101        gettimeofday(2)
1.531963        1.531963        clock_gettime(2/CLOCK_REALTIME)
1.531301        1.531301        clock_gettime(2/CLOCK_MONOTONIC)
0.209740        0.209740        clock_gettime(2/CLOCK_REALTIME_FAST)
0.188042        0.188042        clock_gettime(2/CLOCK_MONOTONIC_FAST)
0.183686        0.183686        clock_gettime(2/CLOCK_SECOND)
0.178539        0.178539        time(3)
Value from time(3): 1395409387
Last value from gettimeofday(2): 1395409387.633267      Equal: 0
Last value from clock_gettime(2/CLOCK_SECOND): 1395409389.000000000
Equal: 999999
Last value from clock_gettime(2/CLOCK_REALTIME_FAST):
1395409389.553192896     Equal: 999789
Last value from clock_gettime(2/CLOCK_REALTIME): 1395409389.343852343
Equal: 0
Last value from clock_gettime(2/CLOCK_MONOTONIC_FAST): 75099.751341116
Equal: 999811
Last value from clock_gettime(2/CLOCK_MONOTONIC): 75099.563772993
Equal: 0

That looks more like your FreeBSD-8 output than the FreeBSD-HEAD one. Do
you think updating to -CURRENT might be worth a try?

Unfortunately that likely won't help the key difference is likely I have
SMP consistent TSC in my test machine which is an old Intel Q6600.

If you switch from ACPI-fast to TSC-low you may well see a benefit but
you also may get strange results due to clock skews.

It's at least worth testing though as it will confirm your looking in
the right area.

   Regards
   Steve



_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux

Reply via email to