I solved the problem.
The problem was that the run_job script did not properly link the spec
files into run directory. I wasn't able to figure out why the problem
occurs in my linux system (CentOS 5.3, x86-64). To fix the problem, I just
changed the run_job script as follows.
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
375 # exec bash -c "pushd $run_dir; ln -sf
../../.skel/{global,$jobspec}/* .; ln -sf ../../.skel/ user/* .; rm -f
'*'; ln -s ../../.skel/start.simics $run_dir/start.simics; popd"
376 exec bash -c "pushd $run_dir; ln -sf
../../../.skel/{global,$jobspec}/* .; ln -sf ../../../ .skel/user/* .;
rm -f '*'; ln -s ../../../.skel/start.simics $run_dir/start.simics; popd"
385 # exec bash -c "ln -s ../../.skel/$flexuslib
$run_dir/$flexuslib"
386 exec bash -c "ln -s ../../../.skel/$flexuslib
$run_dir/$flexuslib"
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Best regards,
Jeongseob
2014-06-30 10:46 GMT+09:00 J Ahn <[email protected]>:
> Hi Cansu,
>
> Thank you for your e-mail.
> I was just wondering how to create the phases by using the script provided
> with Flexus even if Javier's script can generate the phase.
>
> My second question is creating flexpoints by using the script. To create
> flexpoints for a Phase-0 of cassandra application, I used the run_job
> script as follows. I did not find the source of error. Have you experienced
> similar error? (FYI, my system is Linux 2.6.18 with CentOS 5.8)
>
>
> -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> [jeongseob@ilsan flexus-4.1]$ run_job -postprocess
> "/home/jeongseob/flexus-4.1/scripts/postprocess_ckptgen.sh flexpoint 80
> cassandra" -cfg trace -run flexpoint -local -ckpt-gen CMP.L2Shared.Trace
> cassandra/1cpu
> Setting up cassandra/1cpu (0:0):
> 000:000/home/jeongseob/flexus_workspace/results/trace-CMP.L2Shared.Trace-101700
>
>
> >/home/jeongseob/flexus_workspace/results/trace-CMP.L2Shared.Trace-101700/cassandra/1cpu/000_000
>
> sh: ./go.sh: No such file or directory
>
> </home/jeongseob/flexus_workspace/results/trace-CMP.L2Shared.Trace-101700/cassandra/1cpu/000_000
>
> -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Best regards,
> Jeongseob Ahn
>
>
> 2014-06-30 5:51 GMT+09:00 Cansu Kaynak <[email protected]>:
>
> Dear Jeongseob,
>>
>> Thanks for pointing this out. We will add the instructions into the
>> guide.
>>
>> The easiest way to create phases is the way Javier described in his
>> e-mail, which is to write a short script that will write a Simics
>> checkpoint into the corresponding phase directory for each phase.
>>
>> Script:
>>
>> @iter=1
>>
>> @def ccc_hap_callback(user_arg, cpu, arg):
>> global iter
>>
>> file="/full_path_to_phases_folder/phase_00"+`iter`+"/simics/phase_00"+`iter`
>> SIM_write_configuration_to_file(file)
>> iter=iter+1
>> SIM_hap_add_callback_obj_index("Core_Cycle_Count", conf.cpu0, 0,
>> ccc_hap_callback, None, SIM_cycle_count(conf.cpu0)+2000000000)
>>
>> @SIM_hap_add_callback_obj_index("Core_Cycle_Count", conf.cpu0, 0,
>> ccc_hap_callback, None, SIM_cycle_count(conf.cpu0)+2000000000)
>>
>> Place this piece of code into a file and then run it inside Simics by
>> using the command "run-command-file full_path_to_script_file".
>> This script will start writing checkpoints under the directories
>> phase_000, phase_001, phase_0002, etc. It will write a checkpoint each 2
>> billion cycles (1 second of execution time).
>>
>> --
>> Cansu
>>
>> On 23 Jun 2014, at 19:01, J Ahn <[email protected]> wrote:
>>
>> Hi Picorel,
>>
>> I also faced with the same issue creating phases. As mentioned the
>> CloudSuite guide, there is a script to automatically generate phases for
>> CloudSuite. It seems that run_job.rc.tcl and run_job script provide the
>> useful features including phases, flexpoints and other stuffs. However, I
>> cannot find how to create the phases using the run_job provided with Flexus
>> 4.1 release.
>>
>> I tried to create phases as follows.
>>
>> jeongseob@ilsan:~/flexus-4.1$ run_job -run phase NONE cassandra/1cpu
>> Setting up cassandra/1cpu (0-9:0): 009:000
>>
>> >/home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/000_000
>> sh: ./go.sh: No such file or directory
>>
>> </home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/000_000
>> Interactive run, ran this job:
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/000_000
>> Will not run other jobs:
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/001_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/002_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/003_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/004_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/005_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/006_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/007_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/008_000
>>
>>
>> /home/jeongseob/flexus_workspace/results/interactive-interactive/cassandra/1cpu/009_000
>>
>> Could you give me one example to create phases and flexpoints through
>> the run_job command?
>>
>> - Jeongseob
>>
>>
>>
>>
>
