Getting application fatal signal 11 or pin application error

[Laura Angélica Rozo Duque]

Hi,

I'm trying to simulate a pthread application with Sniper, however I'm getting two of the following errors whenever I execute the following command:

./run-sniper -n 4 -c gainestown --roi -- ./fib 2

Sometimes I get the following error:

[SNIPER] Start
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Sniper using Pin frontend
[SNIPER] Running pre-ROI region in  CACHE_ONLY mode
[SNIPER] Running application ROI in DETAILED mode
[SNIPER] Running post-ROI region in FAST_FORWARD mode
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Enabling performance models
[SNIPER] Setting instrumentation mode to DETAILED
Requested 2 only 4 available.
TP: 0, 1
fib(2)
[SNIPER] Disabling performance models
[SNIPER] Leaving ROI after 1.05 seconds
[SNIPER] Simulated 0.1M instructions, 1.9M cycles, 0.03 IPC
[SNIPER] Simulation speed 57.0 KIPS (14.2 KIPS / target core - 70217.8ns/instr)
[SNIPER] Sampling: executed 6.24% of simulated time in detailed mode
[SNIPER] Setting instrumentation mode to FAST_FORWARD
Application received fatal signal 11 at eip 0x7fe20e6c3fea
in :0

Application debugging is currently disabled.
To enable it, start Sniper with the --appdebug or --appdebug-wait command line switch.

[SNIPER] End
[SNIPER] Elapsed time: 4.34 seconds

Some other times, the error is the following:

[SNIPER] Start
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Sniper using Pin frontend
[SNIPER] Running pre-ROI region in  CACHE_ONLY mode
[SNIPER] Running application ROI in DETAILED mode
[SNIPER] Running post-ROI region in FAST_FORWARD mode
[SNIPER] --------------------------------------------------------------------------------
[SNIPER] Enabling performance models
[SNIPER] Setting instrumentation mode to DETAILED
Requested 2 only 4 available.
TP: 0, 1
fib(1)
[SNIPER] Disabling performance models
[SNIPER] Leaving ROI after 2.84 seconds
[SNIPER] Simulated 0.2M instructions, 2.0M cycles, 0.11 IPC
[SNIPER] Simulation speed 72.6 KIPS (18.1 KIPS / target core - 55100.8ns/instr)
[SNIPER] Sampling: executed 10.99% of simulated time in detailed mode
[SNIPER] Setting instrumentation mode to FAST_FORWARD
[SNIPER] Internal exception: Exception Code: ACCESS_INVALID_ADDRESS. Exception Address = 0x7fe5c22593cc. Access Type: UNKNOWN. Access Address = 0x000000010
C: Tool (or Pin) caused signal 11 at PC 0x7fe5c22593cc
Pin app terminated abnormally due to signal 11.
[SNIPER] End
[SNIPER] Elapsed time: 6.34 seconds
------------------------------------------------------------
Exception Code: ACCESS_INVALID_ADDRESS. Exception Address = 0x7fe5c22593cc. Access Type: UNKNOWN. Access Address = 0x000000010

Backtrace:
    pin_sim.cc:traceCallback:239
------------------------------------------------------------

I've been debugging the simulation with gdb, and it seems like the first error is generated after I set thread affinity of my threads. The second one, usually occurs whenever I use the pthread create function, after I set the affinity of each of my threads. My application is able to run normally outside sniper. I would appreciate any help that anyone could provide me.

[Trevor Carlson]

Laura,

Would it be possible to post your application so we can take a look at it? My guess is that something weird is happening that Sniper didn’t expect. Also, which OS are you using? It appears that your version of Linux is 64-bit, is that correct?

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/d/optout.

[Juan F. Martínez]

I am having a very siimilar error in Sniper-6.1. I'm trying to simulate CG benchmark (from NPB3.3) and in the end of the simulation Sniper gives me:

Could not find platform independent libraries <prefix>
Could not find platform dependent libraries <exec_prefix>
Consider setting $PYTHONHOME to <prefix>[:<exec_prefix>]
ImportError: No module named site
C:Tool (or Pin) caused signal 11 at PC 0x7f1ff04c6079
[SIFT_RECORDER] Failed to print backtrace.
[SIFT_RECORDER] Internal exception:Exception Code: ACCESS_INVALID_ADDRESS. Exception Address = 0x7f1ff04c6079. Access Type: UNKNOWN. Access Address = 0x000000000

Pin app terminated abnormally due to signal 11.

sniper: sift_reader.cc:172: bool Sift::Reader::Read(Sift::Instruction&): Assertion `!input->fail()' failed.

It seems I have a problem with python as well...
When I simulate the same benchmark on Sniper-6.0 it works!

I'm on OpenSuse 13.2

[Trevor Carlson]

Juan,

This seems to be a sift-recorder front-end issue only. The back end is probably closing the file pipe, which causes the front-end to abort. Nevertheless, is there a sim.out after execution? My guess is that the back-end ended fine and has all of the statistics. If not, which version of NPB are you running, as it appears that you are using the MPI version, is that correct? If so, which version of MPI are you using, etc. Finally, the Sniper configuration details will help us to debug this issue.

Thanks,

Trevor

[Juan F. Martínez]

Trevor,

Yes, I get sim.out as we can expect.

I'm running the NPB3.3 version. Yes I'm running MPI with Intel MPI v4.1
I am using the Sniper's gainestown configuration.

When running with Sniper6.0 I get 4 -- DONE --, one per MPI rank. Thus, I have a doubt. Though I get the message --DONE--, it is only for the first rank that ends, if Sniper breaks at the final of this rank, we don't have all information in sim.out yet, caused by the fact that Sniper closes the rest of file pipes with the other ranks that are running yet. Am I right?

Thanks!
Fran.

[Laura Angélica Rozo Duque]

Hi Trevor,

Sorry for the late response, I just sent the application source code to you and the snipersim group by email. I also provided some instructions about how to install and run it since the code is kind of big.

Regarding the OS, I'm using a 64 bit version of Linux 3.16.0. To me, it seems like something weird is happening whenever I set the affinity or create the pthreads in the application. With my current configuration, I'm creating 2 pthreads.

Thanks!
Laura

[Aaron Myles Landwehr]

Hi Trevor,

Regarding the issue posted by Laura, it appears the problem was caused by the software we are using being compiled with the '-gp' flag (gprof profiling). Once the switches are removed, we no longer receive random crashes. The issue appeared to be more prevalent the more you increased the work in the program / increased thread counts, but it consistently appeared even in programs that did nothing more than print a few statements and compute total runtime (it took quite a bit of runs to receive the error with these types of programs though). As we don't need the profiling flags anyway, we consider the problem resolved from our side.

In any case, I don't know if this particular issue is a known issue or not, but it would be helpful to have a documented list of flags that have compatibility issues with sniper, etc. for the general user base.

Cheers,

Aaron

[SAM]

Hi Trever,

I'm getting the same error on a multicore application. Was wondering if you ever figured out how to solve it...Thanks!

Sam