XED error when running --benchmarks with multiple workloads

499 views
Skip to first unread message

GA

unread,
Feb 6, 2014, 5:42:09 AM2/6/14
to snip...@googlegroups.com

Hi Wim / Trevor

When i run  multiple workloads using --benchmarks  for eg :

./run-sniper --benchmarks=splash2-fft-test-2,splash2-barnes-test-2,splash2-fmm-test-2

I get the following error

[SNIPER] Setting instrumentation mode to DETAILED
sniper: sift_reader.cc:589: const Sift::StaticInstruction* Sift::Reader::decodeInstruction(uint64_t, uint8_t): Assertion `result == XED_ERROR_NONE' failed.
Pin app terminated abnormally due to signal 6.
[SIFT_RECORDER] [SIFT_RECORDER] [SIFT_RECORDER] sift_writer.cc:sift_writer.cc290: :sift_writer.cc:290290: void Sift::Writer::InstructionCount(uint32_t): Assertion `respRec.Other.type == RecOtherSyncResponse: void Sift::Writer::InstructionCount(uint32_t)' failed.: Assertion `void Sift::Writer::InstructionCount(uint32_t)respRec.Other.type == RecOtherSyncResponse
' failed.: Assertion `
respRec.Other.type == RecOtherSyncResponse' failed.
[SNIPER] End
[SNIPER] Elapsed time: 9.28 seconds
Pin app terminated abnormally due to signal 6.
Pin app terminated abnormally due to signal 6.
Pin app terminated abnormally due to signal 6.


Also if try to use -p option for one workload  ie  ./run-sniper -p splash2-barnes -i test -n 2 then simulation completes.


Can you please let me know why this error is happening ?


Thanks
Gayathri

Trevor E. Carlson

unread,
Feb 6, 2014, 6:03:32 AM2/6/14
to snip...@googlegroups.com, rambli...@gmail.com
Gayathri,

    If your splash benchmarks aren't compiled correctly (or run into other issues), the SIFT feeder (another process that runs the application in a Pin tool) might crash or fail to run the application. This will cause an error to show up in Sniper like the one you are seeing. Take a look at the benchmark-app*-run*.log files to see what is going wrong.

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.

GA

unread,
Feb 6, 2014, 6:24:38 AM2/6/14
to snip...@googlegroups.com, rambli...@gmail.com

Hi Trevor,

If Splash2 aren't compiled correctly, howcome -p option works for single benchmarks ?

for single benchmarks with -p option it works fine. Only with --benchmarks option it gives error.

A:Source/pin/pin/image.cpp:RTN_Name:RTN_Name:1699: assertion failed: RTN_Valid(x)


Gayathri

Trevor E. Carlson

unread,
Feb 6, 2014, 7:22:26 AM2/6/14
to snip...@googlegroups.com, rambli...@gmail.com
Gayathri,

There was a discussion about this on the Sniper mailing list a few months ago [1]. This problem could be caused by Pin or the experimental --viz-aso feature that we added to Sniper recently. If you are using --viz-aso, could you try simulating without it. Otherwise, which version of Pin are you using? You might want to try using a newer (or older) version to see if that helps. Also, which OS are you using, and are you running in 32-bit or 64-bit mode? Right now we are not sure if this is a Pin issue, or an issue with the Sniper pin tools, so any other feedback would definitely be helpful, especially if it could help us reproduce this problem.

Trevor

GA

unread,
Feb 6, 2014, 12:22:55 PM2/6/14
to snip...@googlegroups.com, rambli...@gmail.com


Hi Trevor,

1. I am NOT using  --viz option

( I am running command ./run-sniper --benchmarks=splash2-fft-test-2 from $SNIPER_ROOT/benchmarks folder)

(Also If i use ./run-sniper -p splash2-fft -i test -n 2 , the simulation works perfectly fine)

2.  This error happens only in sniper 5.3. My Sniper 5.2 is running fine with --benchmarks

3. I am using Ubuntu 12.04 Linux and X86_64 architecture

4. I tried using both pinplay1.1 with pin 2.13_61206 and also pinplay-1.2-pin-2.13-62732-gcc.4.4.7-linux. Both of them result in the same error.

sniper: sift_reader.cc:589: const Sift::StaticInstruction* Sift::Reader::decodeInstruction(uint64_t, uint8_t): Assertion `result == XED_ERROR_NONE' failed.
Pin app terminated abnormally due to signal 6.


benchmark_app0_log0 gives the following information :


A:Source/pin/pin/image.cpp:RTN_Name:RTN_Name:1699: assertion failed: RTN_Valid(x)

################################################################################
## STACK TRACE
################################################################################
addr2line -C -f -e "/mnt/srishtistr0/home/gayathri/sniper-5.3/sift/recorder/sift_recorder" 0x7fc64f116839 0x7fc64f1173d0 0x7fc64f1a7ca8 0x7fc64f02e8fa 0x7fc64f02eb06 0x7fc64f1d07d2 0x030604641 0x03064bdd3 0x0305fe556 0x0305ff683 0x03060058b 0x030692f30 0x03067d84e 0x030711c90
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
??
??:0
Pin 2.13
Copyright (c) 2003-2013, Intel Corporation. All rights reserved.
@CHARM-VERSION: $Rev: 62728 $
@CHARM-BUILDER: BUILDER
@CHARM-COMPILER: gcc 4.4.7
@CHARM-TARGET: ia32e
@CHARM-CFLAGS:  __OPTIMIZE__=1  __NO_INLINE__=__NO_INLINE__
[SPLASH] [----------    End of output    ----------]
[SPLASH] Done.


Hope this helps !

Gayathri

Wim Heirman

unread,
Feb 7, 2014, 3:32:29 AM2/7/14
to snip...@googlegroups.com, rambli...@gmail.com
Hi Gayathri,

A:Source/pin/pin/image.cpp:RTN_Name:RTN_Name:1699: assertion failed: RTN_Valid(x)

Thanks, this is what we needed to see. Can you see if the attached patch helps?

Regards,
Wim


sniper-3e34d31.patch

GA

unread,
Feb 7, 2014, 4:45:39 AM2/7/14
to snip...@googlegroups.com, rambli...@gmail.com


Hi Wim,


Thanks. Now it works fine.

Gayathri

leas63

unread,
Feb 20, 2014, 3:59:32 PM2/20/14
to snip...@googlegroups.com, rambli...@gmail.com
Hello,

Can you please explain how  to apply the patch? I encountered the same problem.

Thank you!

Prashanth Thinakaran

unread,
Apr 6, 2014, 3:53:50 AM4/6/14
to snip...@googlegroups.com
Can you please tell me how to patch the given file ?  Even I am stuck with the same problem.

Wen Zong

unread,
Apr 12, 2014, 3:16:55 AM4/12/14
to snip...@googlegroups.com
You can open the patch file. Two files needs to be modified, the pin/trace_rtn.cc and sift/recorder/trace_rtn.cc.
Change the line started with - to the next line started with + in the two files.

mike anastasakis

unread,
May 23, 2014, 4:18:53 AM5/23/14
to snip...@googlegroups.com
I have changed the lines on those two files described on the patch but I still get the same problem as described.
I only have the problem when i use the command --benchmarks . Could you help me? I describe below the results from the run.
This is the result I get.

sniper: sift_reader.cc:589: const Sift::StaticInstruction* Sift::Reader::decodeInstruction(uint64_t, uint8_t): Assertion `result == XED_ERROR_NONE' failed. Pin app terminated abnormally due to signal 6.

and this is my  log file.

[SPLASH] Benchmarks to run: fft

[SPLASH] [========== Running benchmark fft ==========]
[SPLASH] Setting up run directory: /tmp/tmpUEqS9t
[SPLASH] Running '/home/mike/Desktop/sniper/sniper-5.3/record-trace  --routine-tracing -o /tmp/tmpqNbUZv/run_benchmarks -e 1 -r 1 -s 0 --roi --  /home/mike/Desktop/sniper/sniper-5.3/benchmarks/splash2/splash2/codes/kernels/fft/FFT -m10 -p4':
[SPLASH] [---------- Beginning of output ----------]
A:Source/pin/pin/image.cpp:RTN_Name:RTN_Name:1699: assertion failed: RTN_Valid(x)

################################################################################
## STACK TRACE
################################################################################
addr2line -C -f -e "/home/mike/Desktop/sniper/sniper-5.3/sift/recorder/sift_recorder" 0x7ff001a75b19 0x7ff001a766b0 0x7ff001b0d230 0x7ff001a6b194 0x7ff001a6b3a6 0x7ff001b29b02 0x030635d61 0x030680633 0x03062ec66 0x03062fe61 0x030630c0b 0x0306cec20 0x0306b5d3e 0x030757dd0
@CHARM-VERSION: $Rev: 61147 $
@CHARM-BUILDER: BUILDER
@CHARM-COMPILER: gcc 4.4.7
@CHARM-TARGET: ia32e
@CHARM-CFLAGS:  __OPTIMIZE__=1  __NO_INLINE__=__NO_INLINE__
[SPLASH] [----------    End of output    ----------]
[SPLASH] Done.

Trevor E. Carlson

unread,
Jul 1, 2014, 6:07:34 AM7/1/14
to snip...@googlegroups.com, mike anastasakis
Mike,

This problem should be fixed in the most recent version of Sniper, 6.0

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.

Ganesh Venkatesh

unread,
Jul 23, 2014, 9:21:02 PM7/23/14
to snip...@googlegroups.com, anasta...@gmail.com
I ran into a similar error on Sniper 6.0 as well. I think the error there was the RTN_InsHeadOnly was getting called on rtn that have not been opened yet. In particular, the addRtnTracer function in trace_rtn.cc file calls announceRoutine without opening it first. 

void addRtnTracer(TRACE trace)

{

   if (Sim()->getRoutineTracer())

   {

      // At the start of each trace, check to see if this part of the code belongs to the function we think we're in.

      // This will detect longjmps and tail call elimination, and fix up the call stack appropriately.

      RTN rtn = TRACE_Rtn(trace);


      if (RTN_Valid(rtn))

      {

         if (!Sim()->getRoutineTracer()->hasRoutine(RTN_Address(rtn)))

         {

            announceRoutine(rtn);

Wim Heirman

unread,
Jul 29, 2014, 6:04:24 AM7/29/14
to snip...@googlegroups.com, mike anastasakis
Ganesh,

Looks like you're right. The attached patch should fix this.

Regards,
Wim

rtnopen.patch
Reply all
Reply to author
Forward
0 new messages