A couple points with regard to the run-job script:

If you run an out-of-order simulator like UniFlex.OoO, you need to also
include the "-ma" flag.

In general, Flexus supports creating flexpoints with Trace simulators.
You seem to be trying to write a flexpoint (the "-fxpt-out" option) with
an out-of-order simulator.  This may work correctly, but is not behavior
we exercise or test.  If you intended to read the flexpoint instead, use
the "-fxpt" option.

Stephen


On Thu, 23 Jul 2009, Evangelos Vlachos wrote:

>
> Hi Eugen,
>
> Let me try answering some of your questions.
>
> The application you are going to run on the target (simulated) machine has
> to have the SIMICS_magic_instructions in them. Check the structure of the
> flexus-test-app to see the sequence of SIMICS_magic_instructions there.
> These instructions are driving the whole process of checkpoint generation.
> There is some more detailed description here (it's the same for x86 and v9
> systems) https://sos.ece.cmu.edu/pipermail/simflex/2008-October/000443.html.
>
>
> I see you run directly the "run-job" script. This is not a good idea. You
> have to follow the steps mentioned in the guide.
> 'prepare-test-app' will boot the system and run your application until the
> warmup breakpoint. 'create-initial-flexpoints' will run until the second
> iteration of the parallel section and generate a flexpoint.
> 'create-flexpoints' will generate the rest of the flexpoints. Finally,
> 'run-timing-jobs' is used to run the actual experiments.
>
> I'm not sure what exactly the run-job script does with the parameters you
> provided, but in this part of the process you surely don't need a timing
> simulator, such as UniFlex.OoO.
>
> Start again with a benchmark that does have the magic breakpoints and follow
> the steps described in the guide.
>
> Let us know if you have any other question.
>
> Evangelos
>
> -----Original Message-----
> From: simflex-bounces at ece.cmu.edu [mailto:simflex-bounces at ece.cmu.edu] 
> On
> Behalf Of Leontie Eugen
> Sent: Thursday, July 23, 2009 9:33 PM
> To: simflex at ece.cmu.edu
> Subject: [Simflex] running my own benchmarks with Flexus
>
>
> After successfully following the steps in this post :
> [Simflex] Help for installing ultrasparc over linux
> Evangelos Vlachos evlachos at ece.cmu.edu
> Tue Jun 16 18:24:22 EDT 2009
>
> I tried running the following job :
> [eugen at localhost flexus-test-app]$ ./run-job -fxpt-out flexpoint_000 -cfg
> OoO UniFlex.OoO
>
> It did not go smooth ... I am hoping someone can help me out. There is no
> benchmark application running on the target machine at this point. But from
> my understanding that should not be a problem.
>
>
> Starting initial flexpoint creation of flexus_test with UniFlex.OoO in
> /home/eugen/work/simics/flexus/Flexus-3.0.0/flexus-test-app/runs/eugen-UniFl
> ex.OoO-23Jul09-165202
> Launching job
> Opening debug output file: debug.out
> Opening debug output file: stats.out
> Successfully parsed debug configurations from debug.cfg
> Initializing Flexus::ConfigurationManager...done
> Initializing Flexus::ComponentManager...done
> Entered init_local
> 1 <startup.cpp:110> {0}- Initializing Flexus.
>
> Flexus (C) 2006 The SimFlex Project
> Eric Chung, Michael Ferdman, Brian Gold, Nikos Hardavellas, Jangwook Kim,
> Ippokratis Pandis, Minglong Shao, Jared Smolens, Stephen Somogyi,
> Evangelos Vlachos, Thomas Wenisch, Roland Wunderlich
> Anastassia Ailamaki, Babak Falsafi and James C. Hoe.
>
> Flexus Simics simulator - Built as UniFlex.OoO v3.0
>
> 2 <ComponentManager.cpp:81> {0}- Instantiating system with a width factor
> of: 1
> 3 <uFetch.hpp:76> {0}- ufetch port FetchSnoopOut is not wired
> 4 <Cache.hpp:82> {0}- L1d port BackSideOut_Prefetch is not wired
> 5 <Cache.hpp:82> {0}- L2 port BackSideOut_Snoop is not wired
> 6 <Cache.hpp:82> {0}- L2 port BackSideOut_Prefetch is not wired
> 7 <wiring.cpp:89> {0}-  initializing Parameters...
> Warning: The 'flexus-UniFlex.OoO-v9-iface-gcc' module unexpectedly defined
> the 'Flexus' class
> Warning: The 'flexus-UniFlex.OoO-v9-iface-gcc' module unexpectedly defined
> the 'SimicsInterface' class
> 8 <flexus.cpp:377> {0}- Set stat interval to : 50000
> 9 <flexus.cpp:382> {0}- Set region interval to : 50000
> Opening debug output file: trace.out
> Successfully parsed debug configurations from iteration.trace.cfg
> WARNING: There is no parameter named "-L2:allow_offchip_stream_fetch"
> WARNING: There is no parameter named "-L2:bypass_bus"
> type run to start simulation
> simics> run
> 10 <ComponentManager.cpp:96> {0}- Initalizing components...
> 11 <FetchAddressGenerateImpl.cpp:90> {0}- Thread[0.0] connected to cpu0
> Initial PC: v:000532eb4
> 12 <mai_api.hpp:386> (<undefined>[<undefined>]) {0}- Assertion failed:
> ((!(false))) : <undefined>
> Abort (SIGABRT) in main thread
> The simulation state has been corrupted. Simulation cannot continue.
> Please restart Simics.
> [cpu0] v:0x0000000000532eb4 p:0x00000132eb4  lduba [%i3 + %g0] 0x15, %i0  #
> ASI_PHYS_BYPASS_EC_WITH_EBIT
>
>
>
> So far I managed to get Linux on the Simics sunfire-bagle configuration. I
> automatically login, and start running a benchmark ( bitcount ), a snapshot.
> ( part of what prepare-test-app is suppose to do).
>
> I need some assistance getting to the next steps:
> - do I need to insert SIMICS_magic_instructions into the benchmark code?
> - do I need a special compiler that will allow me to cross compile for
> v9_Linux that knows about magic_instructions ?
>
> I have a hard time (based on the Getting starting guide) understanding the
> relation of what happens on the target system running the simulation and
> these steps :
>   prepare-test-app
>   create-initial-flexpoints
>   create-flexpoints
>   run-timing-jobs
> and
>   run-job
>
> I am sure these are obvious for someone familiar with flexus, but they do
> not come out easily from the "Getting Started Guide".
> Thank you in advance for your help,
> Eugen
>
>
>
> _______________________________________________
> SimFlex mailing list
> SimFlex at ece.cmu.edu
> https://sos.ece.cmu.edu/mailman/listinfo/simflex
> SimFlex web page: http://www.ece.cmu.edu/~simflex
>
> _______________________________________________
> SimFlex mailing list
> SimFlex at ece.cmu.edu
> https://sos.ece.cmu.edu/mailman/listinfo/simflex
> SimFlex web page: http://www.ece.cmu.edu/~simflex
>

Reply via email to