facing problem while running cpu2006 benchmarks

[Kapil Dev]

Hi,

I am trying to run cpu2006 benchmarks using sniper, but I am not successful yet. I am using following command:
$SNIPER_PATH/sniper-4.1/run-sniper -c gainestown -- runspec --config=mytest.cfg --size=test --noreportable --tune=base --iterations=1 bzip2

Is it the right way to run? I run the above command from $BENCHMARK_PATH/cpu2006/config directory, because "runspec" is located in the config directory.

By the way, I use following command to run the same benchmark outside sniper and it runs.
cd $BENCHMARK_PATH/cpu2006/config
and then, run,

runspec --config=mytest.cfg --size=test --noreportable --tune=base --iterations=1 bzip2

If someone is able to run cpu2006 benchmarks, manually, (i.e. one benchmark at a time) or in an integrated fashion, please let me know. I would appreciate if you could tell me the steps you followed to run a sample cpu2006 benchmark.

Thanks,
Kapil

[Wim Heirman]

Hi all,

We're working on properly integrating SPEC CPU2006 into our
benchmarks. We have a preliminary version ready but it could use some
testing. The patch at [1] should add everything you need (onto an
up-to-date checkout of our benchmarks repository).

Steps to get this working:
- download the patch at [1] and apply it to your benchmarks directory:
$ patch -p1 < path/to/benchmarks_cpu2006.patch
- copy over the required files from your SPEC CPU2006 DVD using the
script provided
$ ./install_cpu2006 PATH_TO_CPU2006
- build everything
$ make
- run Sniper with a cpu2006 benchmark (the ipctrace is to get some
visual feedback on whether Sniper is making progress)
$ ./run-sniper -pcpu2006-bwaves -sipctrace

Hope this helps, all feedback is welcome.

Regards,
Wim

[1] http://snipersim.org/documents/benchmarks_cpu2006.patch

> --
> --
> --
> You received this message because you are subscribed to the Google
> Groups "Sniper simulator" group.
> To post to this group, send email to snip...@googlegroups.com
> To unsubscribe from this group, send email to
> snipersim+...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/snipersim?hl=en
>
> ---
> You received this message because you are subscribed to the Google Groups
> "Sniper simulator" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to snipersim+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Kapil Dev]

Hi Wim,

Thanks a lot for providing the detailed steps and the patch file.

I followed the below mentioned steps. I am able to run up to following step: $ ./install_cpu2006 PATH_TO_CPU2006

But, I am getting errors while I am building it using make command . An excerpt from errors is  as follows;

/usr/bin/gcc -c -o rtl.o -DSPEC_CPU -DNDEBUG -I.   -O3 -fno-strict-aliasing       -DSPEC_CPU_LP64  -I/home/sreda/workspace/    x86simulator/sniper/benchmarks/tools/hooks -I/home/sreda/workspace/x86simulator/sniper/sniper-4.1/include        rtl.c
 In file included from rtl.c:188:0:
 machmode.def:79:1: warning: left shift count >= width of type [enabled by default]
 machmode.def:79:1: error: initializer element is not constant
 machmode.def:79:1: error: (near initialization for ‘mode_mask_array[4]’)
 machmode.def:89:1: warning: left shift count >= width of type [enabled by default]
 machmode.def:89:1: error: initializer element is not constant
 machmode.def:89:1: error: (near initialization for ‘mode_mask_array[10]’)

I am not sure why I am getting these errors.

Thanks,
Kapil

[Wim Heirman]

Hi Kapil,

Which OS, architecture (32/64-bit) and GCC version are you on?

Regards,
Wim

On 13/03/2013 22:06, Kapil Dev wrote:
> Hi Wim,
>
> Thanks a lot for providing the detailed steps and the patch file.
>
> I followed the below mentioned steps. I am able to run up to following
> step: $ ./install_cpu2006 PATH_TO_CPU2006
>
> But, I am getting errors while I am building it using make command . An
> excerpt from errors is as follows;
>
> /usr/bin/gcc -c -o rtl.o -DSPEC_CPU -DNDEBUG -I. -O3
> -fno-strict-aliasing -DSPEC_CPU_LP64 -I/home/sreda/workspace/
> x86simulator/sniper/benchmarks/tools/hooks
> -I/home/sreda/workspace/x86simulator/sniper/sniper-4.1/include rtl.c
> In file included from rtl.c:188:0:
> machmode.def:79:1: warning: left shift count >= width of type [enabled
> by default]
> machmode.def:79:1: error: initializer element is not constant

> machmode.def:79:1: error: (near initialization for �mode_mask_array[4]�)

> machmode.def:89:1: warning: left shift count >= width of type [enabled
> by default]
> machmode.def:89:1: error: initializer element is not constant

> machmode.def:89:1: error: (near initialization for �mode_mask_array[10]�)

> On 11 March 2013 21:08, Kapil Dev <kapi...@brown.edu <javascript:>>

> <javascript:>

> > To unsubscribe from this group, send email to

> > snipersim+...@googlegroups.com <javascript:>

> > For more options, visit this group at
> > http://groups.google.com/group/snipersim?hl=en
> <http://groups.google.com/group/snipersim?hl=en>
> >
> > ---
> > You received this message because you are subscribed to the
> Google Groups
> > "Sniper simulator" group.
> > To unsubscribe from this group and stop receiving emails from it,
> send an

> > email to snipersim+...@googlegroups.com <javascript:>.

> > For more options, visit https://groups.google.com/groups/opt_out

> <https://groups.google.com/groups/opt_out>.

> >
> >
>
> --
> --
> --
> You received this message because you are subscribed to the Google
> Groups "Sniper simulator" group.
> To post to this group, send email to snip...@googlegroups.com
> To unsubscribe from this group, send email to
> snipersim+...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/snipersim?hl=en
>
> ---
> You received this message because you are subscribed to the Google
> Groups "Sniper simulator" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to snipersim+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

--
dr. ir. Wim Heirman,
ELIS Department, Ghent University, Belgium
Phone: +32-9-264.34.56
E-mail: wim.h...@elis.UGent.be
http://www.elis.UGent.be/~wheirman/

[Kapil Dev]

Hi Wim,

I am using Ubuntu OS (3.2.0-35-generic-pae #55-Ubuntu), with i686 architecture (32-bit). The gcc version is 4.6.3 (gcc version 4.6.3 (Ubuntu/Linaro 4.6.3-1ubuntu5) )

Thanks,
Kapil

[Kenzo Van Craeynest]

Hi,

I didn't check, but I wouldn't be surprised if spec2006 doesn't compile immediately with the very recent version of gcc that you are using. Have you tried to use an older version?

Kenzo

[Kapil Dev]

Kenzo: Just want to tell you that I was able to compile these benchmarks outside sniper with my current version of GCC. So, I don't know why I would not be able to compile them inside sniper.

Wim: I would like to ask you which gcc version did you use during your compilation? Also, what will I have to do for compiling these benchmarks with gcc version 4.6.3. It seems like we get the compilation error when it is compiling the "gcc" benchmark inside cpu2006; some benchmarks, which appear before gcc benchmark during compilation, seem to have compiled properly.

Any insight in this regard is appreciated.

Thanks,

Kapil

[Wim Heirman]

Kapil,

Our CPU2006 integration was hardcoded for 64-bit platforms. You'll
need to make a few changes to the makefiles to get it to work for
32-bit:

* in install_cpu2006, use the -i386 version rather than the -AMD64
version of the cpu2006tools package (line 29)
* get rid of the -DSPEC_CPU_LP64 flag in
cpu2006/CPU2006/403.gcc/src/Makefile.spec, line 44

There may be other fixes needed (I haven't gotten round to compiling
all benchmarks yet), comparison of the Makefile.spec files in our
distribution with those generated by your local compilation should
tell you what has to be changed. If you could post an overview of the
other required changes here that would be much appreciated, so we can
include them in our next distribution of the benchmarks.

Regards,
Wim

[Kapil Dev]

Wim,

Thanks again for the help.

I am now able to compile the gcc  benchmark. And, as you rightly guessed, every benchmark needs special attention for its successful compilation. Just for the convenience of others, one needs to have gfortran installed to compile the benchmarks written in fortran, e.g. bwaves.

Now, I am working on resolving the following error that I am getting during 416.gamess compilation
PATH_TO_BENCHMARKS/benchmarks/cpu2006/bin/specperl: 1: PATH_TO_BENCHMARKS/benchmarks/cpu2006/bin/specperl: Syntax error: "(" unexpected
make[2]: *** [aldeci.fppized.f] Error 2

If you have resolved this kind of error before, please let me know.

Thanks,
Kapil

[Kapil Dev]

Wim,

The good news is that I am able to compile and run  some of the CPU2006 benchmarks inside sniper with the great help from you. Thanks a lot, Wim!

For now, I removed those benchmarks that are giving compilation-error ("syntax error.., perspec...") from the "Makefile" itself. So, if I could somehow resolve this error, then I would be able to compile and run all benchmarks. Please let me know if someone has resolved this error.

Thanks,

Kapil

[fatemeh]

hi
I have a problem with runnin cpu2006 benchmark
I use the following command, but I dont know why it says this warning and error?
./run-sniper -c gainestown --benchmarks=cpu2006-milc-test-1 -d cpu2006/result/
[SNIPER] Start
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Sniper using SIFT/trace-driven frontend
[SNIPER] Running full application in DETAILED mode
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Enabling performance models
[SNIPER] Setting instrumentation mode to DETAILED
[SNIPER] WARNING: No SIFT connections made yet. Waiting...
[SNIPER] ERROR: Could not establish SIFT connection, aborting! Check benchmark-app*.log for errors.
[SNIPER] End
[SNIPER] Elapsed time: 60.30 seconds

what should I do to resolve this problem?
thanks
fatemeh

[fatemeh]

the content of benchmark-app*.log is:
[CPU2006] [========== Running benchmark milc ==========]
[CPU2006] Running ' /home/hw/sniper-5.2/record-trace --routine-tracing -o /tmp/tmpZx6f57/run_benchmarks -e 1 -r 1 -s 0 --roi --  /tmp/tmpzT3dnl/milc_base.sniper-x86_64 < su3imp.in ':
[CPU2006] [---------- Beginning of output ----------]
SU3 with improved KS action
Microcanonical simulation with refreshing
MIMD version 6
Machine = Scalar processor, with 1 nodes
R algorithm
type 0 for no prompts  or 1 for prompts
nflavors 2
nx 10
ny 10
nz 10
nt 2
iseed 1234
LAYOUT = Hypercubes, options = EVENFIRST,
Made lattice
Made nn gathers
Made 3nn gathers
Finished setup


warms 0
trajecs 1
traj_between_meas 1
beta 5.6
mass 0.125
u0 1
microcanonical_time_step 0.2
steps_per_trajectory 5
max_cg_iterations 100
error_per_site 0.125
error_for_propagator 0.125
fresh
forget
unit gauge configuration loaded
CHECK PLAQ: 3.000000e+00 3.000000e+00
Unitarity checked.  Max deviation 0.00e+00
Symanzik 1x1 + 1x2 + 1x1x1 action
loop coefficients: nloop rep loop_coeff  multiplicity
                    0 0      1.000000e+00     6
                    1 0      -5.000000e-02     12
                    2 0      0.000000e+00     16
"O(a^2): couplings(pi)=0, Naik term, No O(a^2) errors, tadpole weights"
path coefficients: npath  path_coeff  multiplicity
                    0      6.250000e-01     8
                    1      -4.166667e-02     8
                    2      -6.250000e-02     48
                    3      1.562500e-02     192
                    4      -2.604167e-03     384
                    5      -6.250000e-02     48
WARMUPS COMPLETED


thanks
fatemeh

[Trevor E. Carlson]

Fatemeh,

    This line is the most important from the Sniper output provided:

[SNIPER] ERROR: Could not establish SIFT connection, aborting! Check benchmark-app*.log for errors.

    It means that the application front-end did not properly open a connection to the timing back-end. That could mean that the benchmark wasn't compiled correctly, or that the front-end crashed for some reason. Could you copy the output of the benchmark-app*.log file in an e-mail so we can see what the problem is?

Thanks,
Trevor

--
--
--
You received this message because you are subscribed to the Google
Groups "Sniper simulator" group.
To post to this group, send email to snip...@googlegroups.com
To unsubscribe from this group, send email to
snipersim+...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/snipersim?hl=en
 
---
You received this message because you are subscribed to the Google Groups "Sniper simulator" group.
To unsubscribe from this group and stop receiving emails from it, send an email to snipersim+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

-- 
----------------------------------
Trevor E. Carlson
Ph.D Student, Ghent University
trevor....@elis.ugent.be
http://www.elis.ugent.be/~tcarlson
http://snipersim.org

[Wim Heirman]

Trevor,

It looks like there is no ROI marker in this benchmark? Or if there is (I don't think we added one though) it's more than 60 seconds of fastforwarding beyond the start.

Did our public CPU2006 integration link with hooks_noroi?

-Wim

[fatemeh]

Trevor,

thanks
I copied it before. exactly after my question.

Fatemeh

[Trevor E. Carlson]

Fatemeh,

Could you try adding the '--no-roi' option to the run-sniper command? It looks like the benchmark might not be properly triggering ROI in Sniper.

Trevor

[fatemeh]

Trevor,

I added --no-roi to the command, the benchmark seems to be running, but it takes too long more than an hour, and it still runs.
thanks

Fatemeh

[Trevor E. Carlson]

Fatemeh,

You might also want to consider the SPEC CPU2006 pinballs that we released for Sniper [1]. You can will find both the whole-program pinballs (that you can use to generate your own representative PinPoint pinballs) and SimPoint-derived representative region pinballs. Using representative regions might prove to be beneficial as it could reduce overall simulation time (depending on your research direction, of course). You can find more information about generating your own representative pinballs here [2], and you can find the Pinplay version of Pin here [3].

Trevor

[1] http://snipersim.org/w/Pinballs

[2] http://snipersim.org/w/Tutorial:HPCA_2013_PinPoints

[3] http://software.intel.com/en-us/articles/program-recordreplay-toolkit

[Ghizlane Mouslih]

Hi,

I'm facing the same problem

yuanqing@ubuntu:~/research/sniper-6.1/benchmarks$ ./run-sniper -c gainestown --no-roi --benchmarks=cpu2006-milc-test-1 -d cpu2006/result/
[app0] readline() on closed filehandle PERLFILE at /home/yuanqing/research/sniper-6.1/benchmarks/cpu2006/run_spec.pl line 80.
[app0] Use of uninitialized value $file in eval "string" at /home/yuanqing/research/sniper-6.1/benchmarks/cpu2006/run_spec.pl line 84.
[app0] Undefined subroutine &Main::invoke called at /home/yuanqing/research/sniper-6.1/benchmarks/cpu2006/run_spec.pl line 99.
[app0] [CPU2006] [========== Running benchmark milc ==========]
[app0] [CPU2006] Running ' /home/yuanqing/research/sniper-6.1/record-trace --routine-tracing -o /tmp/tmpl6qaUE/run_benchmarks -e 1 -r 1 -s 0 -- /tmp/tmpps9_Nf/ ':
[app0] [CPU2006] [---------- Beginning of output ----------]
[SNIPER] Start
[app0] /tmp/tmpps9_Nf/ : Permission denied
[app0] [CPU2006] [---------- End of output ----------]
[app0] [CPU2006] Done.

[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Sniper using SIFT/trace-driven frontend
[SNIPER] Running full application in DETAILED mode
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Enabling performance models
[SNIPER] Setting instrumentation mode to DETAILED
[SNIPER] WARNING: No SIFT connections made yet. Waiting...
[SNIPER] ERROR: Could not establish SIFT connection, aborting! Check benchmark-app*.log for errors.
[SNIPER] End

[SNIPER] Elapsed time: 60.23 seconds

And this is the benchmark-app*.log:

readline() on closed filehandle PERLFILE at /home/yuanqing/research/sniper-6.1/benchmarks/cpu2006/run_spec.pl line 80.
Use of uninitialized value $file in eval "string" at /home/yuanqing/research/sniper-6.1/benchmarks/cpu2006/run_spec.pl line 84.
Undefined subroutine &Main::invoke called at /home/yuanqing/research/sniper-6.1/benchmarks/cpu2006/run_spec.pl line 99.

[CPU2006] [========== Running benchmark milc ==========]

[CPU2006] Running ' /home/yuanqing/research/sniper-6.1/record-trace --routine-tracing -o /tmp/tmpofbqRu/run_benchmarks -e 1 -r 1 -s 0 --roi -- /tmp/tmp4WFyg6/ ':

[CPU2006] [---------- Beginning of output ----------]

/tmp/tmp4WFyg6/ : Permission denied
[CPU2006] [---------- End of output ----------]
[CPU2006] Done.

Please can you help me to know what is the problem ?

[stevenjer...@gmail.com]

Hello,

I am very sorry for this. I am very new and struggling to understand how to apply this solution. Can you elaborate on it a bit more. I have followed your procedure but I just cant get it to work. 

Thank you

[Alex Pollack]

Hi Trevor,

I find it very odd that I'm unable to locate any ROI markers in the source code yet whenever I run one of the '06 SPEC benchmarks, it begins running with "Entering ROI".  I'd also add that I'm not including the --roi flag.  Could you help me understand how this is happening?

Thank you.

Best,

Alex

On Wednesday, February 12, 2014 at 8:12:50 AM UTC-5, Trevor E. Carlson wrote:

Fatemeh,

Could you try adding the '--no-roi' option to the run-sniper command? It looks like the benchmark might not be properly triggering ROI in Sniper.

Trevor

On Feb 12, 2014, at 2:08 PM, fatemeh <rafieenez...@gmail.com> wrote:

Trevor,

thanks
I copied it before. exactly after my question.

Fatemeh

--
--
--
You received this message because you are subscribed to the Google
Groups "Sniper simulator" group.
To post to this group, send email to snip...@googlegroups.com
To unsubscribe from this group, send email to

snip...@googlegroups.com

For more options, visit this group at
http://groups.google.com/group/snipersim?hl=en
 
---
You received this message because you are subscribed to the Google Groups "Sniper simulator" group.

To unsubscribe from this group and stop receiving emails from it, send an email to snip...@googlegroups.com.