Re: [Gems-users] GEMS Simulator Simulation time


Date: Wed, 2 Nov 2005 18:07:49 -0600
From: Hrishikesh Das <hrishikesh.das@xxxxxxxxx>
Subject: Re: [Gems-users] GEMS Simulator Simulation time
Also the ruby stats seem a little strange.

a simple 'ls' command results in 13 million instructions to be executed. running a benchmark for 1 min causes 30 million instructions:

instruction_executed: 30670883 [ 30670880 1 1 1 ]
cycles_per_instruction: 3.59862 [ 0.899655 2.75932e+07 2.75932e+07 2.75932e+07 ]
misses_per_thousand_instructions: 0.939653 [ 0.939654 0 0 0 ]


So is it doing the simulation correctly ? or doing some extra work ?

I did recompile the simics with -02 options and without the debugging options.


Hrishikesh.


P.S. I have a amd64 3700+ san diego at home, will there be a significant increase in performance with that. ?? because I have to install linux , so is it worth the extra effort in that direction.


On 11/2/05, Hrishikesh Das <hrishikesh.das@xxxxxxxxx> wrote:
Hi,

Thanks to all of you for the suggestions.

I am using top to monitor the cpu usage. its 100% i Think.

top - 16:53:24 up 21 days,  6:58,  3 users,  load average: 0.10, 0.40, 0.65
Tasks:  75 total,   3 running,  72 sleeping,   0 stopped,   0 zombie
Cpu(s): 50.1% us,  0.3% sy,  0.0% ni, 49.6% id,  0.0% wa,  0.0% hi,  0.0% si
Mem:   2073480k total,  2055880k used,    17600k free,    78764k buffers
Swap:  2040244k total,        0k used,  2040244k free,  1008036k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
31447 hd35      25   0  888m 753m 5808 R 99.9 37.2  37:34.45 simics-common
 2897 root      16   0  7624 3040 2496 S  0.3  0.1  37:18.22 sge_execd


also, I dont know if its swapping a lot, because I have to ssh to this remote machine, But it seems to have 2 GB of RAM, also its a 3.4GHz p4 machine.

Now the other suggestions about the optimizations (O2), can someone tell me how to do that ?
Also is reducing the memory latency, the same thing as suggested by Dan with the ruby/config/rubyconfig.default
s file ?


Thanks,
Hrishikesh


On 11/2/05, Mike Marty <mikem@xxxxxxxxxxx > wrote:
As Dan said, slowdown is significant with timing models.  When Ruby is
used for timing on top of Simics, we see a slowdown of 10x or so.

Things to watch out for:

-- Make sure your host machine is not paging.  Simics should be cranking
away at 100% of the CPU time

-- Make sure that Opal and Ruby are compiled with optimizations.  O2 or
better.

-- Try setting the memory latency of Ruby to a very low value to see if
this speeds things up.  Of course you may not want to simulate a fast
memory latency



Mike



> Dear Sirs,
>  I am a student using GEMS in a Parallel computing architecture course. I am
> using SIMICS and a SIMICS suse 7.3 image running on a X86 dual processor
> server.
>  The problem I am facing is that when using Ruby for simulation, the
> simulation is extremely slow. Ex- a simple simulation of Unixbech, that
> takes SIMICS about 1 -2 minutes to emulate, takes SIMICS with Ruby about 10+
> hours. Is this kind of slowdown natural ? Is there anything I can do to
> speed things up ? I need to get these results done for my course work,
>  Any help will be greatly appreciated.
> Thanks,
> Hrishikesh Das.
>  Graduate Student,
> Mississippi State University.
>
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users


[← Prev in Thread] Current Thread [Next in Thread→]