RE: spin_count and cpu usage

  • From: Tanel Poder <tanel.poder.003@xxxxxxx>
  • To: ax.mount@xxxxxxxxx, "'Stephane Faroult'" <sfaroult@xxxxxxxxxxxx>
  • Date: Tue, 05 Sep 2006 17:41:37 +0800

Well, thanks to faster CPUs the time to try to get a latch (spin) has
reduced, but also the time to release the latch (to complete the work that
requires latch protection) has gone down as well. 
2nd question - CPU MHz (precisely the core clock speed) is just one factor
in your system. And in a large SMP the core clock speed has small
significance. Rather the memory bus speed, memory bus architecture and
algorithms, CPU cache size and how well the CPU cache is managed can be even
more important factors than just CPU core clock speed.
If most of the time in a CPU is spent on stalling on main memory accesses,
it doesn't really help if you just clock the CPU speed up even more -
without "clocking up" memory access speed...


From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of amonte
Sent: Tuesday, September 05, 2006 06:10
To: Stephane Faroult
Cc: tim@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: Re: spin_count and cpu usage

I didnt miss the "Fen Shui" part :-P
I was just wondering if a parameter default value configured 15 years ago
still applies for modern CPUs.
Regarding my second question anyone know? Getting a CPU twice as fast than
the old one would we be able to reduce CPU time by alf?


Other related posts: