On Mon, 29 Oct 2012, Payne, Benjamin wrote:
Thanks for the advice Nilay. I found the following:
Line 173 of convert.py is the error message, so I go back to the top of
the function and see that "toFrequency" is being called. On line 119,
the toFrequency function is looking for "THz" or "GHz" or "MHz". Thus my
command '2.8Ghz' was using the wrong case.
build/X86/gem5.opt configs/example/se.py -c bench.lex --clock='2.8GHz'
works as desired, in that the processor is closer to what I want to
simulate: an Intel Core i7-930 @ 2.8GHz.
When I use 2.8GHz, the reported CPU time is 0.06 seconds, and just 0.01
seconds when I use 10GHz. That is as expected since
0.06/(10/2.8)=0.0168.
Now I am back to my original question: how to most accurately simulate a
real x86 processor for comparison?
From cat /proc/cpuinfo, the "cache size" is reported as 6144KB. The
specifications sheet
http://ark.intel.com/products/41447/Intel-Core-i7-930-Processor-8M-Cache-2_80-GHz-4_80-GTs-Intel-QPI
reports that "Intel Smart Cache" is 8MB.
How would I give this information to gem5?
Reading the documentation provided on gem5.org is a highly recommended. In
particular, I believe the information on the page
http://gem5.org/Running_gem5 should be useful for you.
[Suggestion: a script which can be run on a real system (ie it queries
/proc/cpuinfo and /proc/meminfo) and gives the command line arguments
for an equivalent gem5 simulation.]
You have all the code available. Change it in whatever way you like.
--
Nilay
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
