Re: [Gems-users] Different execution instruction numbers between simics and opal

[Dan Gibson <degibson@xxxxxxxx>]

1. If you want a two-cpu system, I suggest you just boot a 2-cpu checkpoint instead. It IS possible in Solaris to deconfigure a CPU, but I don't know how to do it.

2. I can't tell you how many times I've answered the question  "Why does my multiprocessors simulation run for an unexpected number of instructions?" on this list. Read the paper I mentioned. Multiprocessors aren't deterministic.

Regards,
Dan

On Fri, May 28, 2010 at 7:41 AM, shanshuchang <shanshuchang@xxxxxxxxx> wrote:

Thanks for your quickly reply, Dan.

 

--2. I don't think FFT binds itself at all by default, so disabling P2 and P3 might disable FFT threads

 

I did modify the FFT codes to bind the 2 threads into the first 2 cores with command: ./FFT –P2 –a 0 –x 1

 

--3. Do not simply disable CPUs once the system is booted. The OS believes disabled CPUs are still running, and will still involve them in inter-processor interrupts (IPIs), e.g., TLB shootdowns. Disabling even one CPU will fairly rapidly deadlock the system, as the disabled CPU will not respond to IPIs and the system will eventually hang.

 

Yes, that’s true. So If there is any way I can follow if do I want to disables CPUs after the system is running?

 

By the way, I did disable the cpu2 and cpu3 and run the FFT thread with the first two threads.

The command is like this:

 

./simics –stall

>read-configuration /export/workspace/shanshuchang/gems/checkpoint_files/simics-2.x/sol9-4p/redundant_4p/fft-2p.check

>load-module ruby

>ruby0.setparam SIMICS_RUBY_MULTIPLIER 1

>ruby0.init

>magic-break-enable

>cpu2.disable

>cpu3.disable

>c

 

The execution can successfully interrupted by magic instruction after the MAGIC_BREAKPOINT. And the ruby dump data shows the instructions executed:

 

instruction_executed: 4990252 [ 4500056 490194 1 1 ]

simics_cycles_executed: 19924811 [ 4981096 4981713 4981001 4981001 ]

cycles_per_instruction: 3.99266 [ 1.1069 10.1615 4.9811e+06 4.9811e+06 ]

misses_per_thousand_instructions: 1.17529 [ 0.377995 8.4946 0 0 ]

 

As you said, not advancing CPU in opal is the same as disabling the simics, so why they differed so much?

The opal + ruby runs for 10,000,000 but not finished, while the simics+ruby finishes in 450056 insts…

 

I am so confused..

 

Regards

shuchang

_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.

-- 
http://www.cs.wisc.edu/~gibson [esc]:wq!