[Simh] SIMH 3.8-1 IDLE in VMS, interesting CPU usage
Marc Chametzky
marc at bluevine.net
Wed Feb 18 10:55:45 EST 2009
> Also I've noticed the 'old' idle code seems to work better...
I put in the idle code that you suggested, along with a small change:
return sim_set_idle (uptr, val, cptr, desc);
becomes
return sim_set_idle (uptr, val, NULL, desc);
The reason for that change is that as of 3.8-1, sim_set_idle() now takes
cptr, if it's not NULL, and uses it to parse out an integer. When one
does "SET CPU IDLE=VMS" or any other non-numeric value, sim_set_idle()
returns SCPE_ARG and the command is rejected.
Here's the graph for this idle code:
http://www.bluevine.net/~marc/CPUidle2.jpg
It has the same characteristic sawtooth pattern, but the peak CPU
utilization is only around 2 GHz, instead of the full 2.6 GHz available.
(The sawtooth pattern of CPU usage corresponds to watching the "vax"
process in "top", with slowly increasing CPU% over time.)
As another experiment, unrelated to IDLE, I ran PT_VAX.EXE. Running with
unrestricted access to the 2.6 GHz core (AMD Athlon 64 X2 5200+, so it's
actually 2.0 GHz, but ESXi thinks it's 2.6), it came up with 12 VUPs. If
I restrict it down to 1.5 GHz (57%), it runs along at 6 VUPs.
--Marc
More information about the Simh
mailing list