Analyzers
Talk to fellow users of Intel Analyzer tools (Intel VTune™ Profiler, Intel Advisor)

VTune Amplifier XE 2015 native analysis on Xeon PHI with zero elapsed time

Andreas_T_1
Beginner
319 Views

I am trying to use VTune Amplifier XE 2015 to analyze application on Xeon Phi in the native mode.

Without VTune, the application on Xeon Phi runs normally (follows script: run.sh).
  #!/bin/sh
  export LD_LIBRARY_PATH=/tmp
  /home/atrianta/clstmocrtrain

With VTune, the elapsed time is zero but VTune runs the application normally.

command : /opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-cl -v -collect=advanced-hotspots -target-system=mic-native:0 -- /home/atrianta/run.sh

results :

amplxe: Using target: mic-native:0
TOTAL time:  48.7338 secs  = 0.812229 mins
amplxe: Using result path `/home/atrianta/r113ah'
amplxe: Executing actions  8 % Clearing the database                           
amplxe: The database has been cleared, elapsed time is 0.126 seconds.
amplxe: Executing actions 16 % Updating precomputed scalar metrics             
amplxe: Raw data has been loaded to the database, elapsed time is 0.146 seconds.
amplxe: Executing actions 16 % Processing profile metrics and debug information
amplxe: Data transformations have been finished, elapsed time is 0.000 seconds.
amplxe: Executing actions 16 % Resolving module symbols                        
amplxe: Locating file `/lib64/ld-2.12.so' on the remote system
amplxe: Locating file `/usr/lib/debug/lib/modules/2.6.32-573.18.1.el6.x86_64/vmlinux' on the remote system
amplxe: Locating file `/lib64/libc-2.12.so' on the remote system
amplxe: Locating file `/usr/lib64/libcrypto.so.1.0.1e' on the remote system
amplxe: Locating file `/opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-runss' on the remote system
amplxe: Warning: Cannot locate file `/lib64/ld-2.12.so'.
amplxe: Executing actions 17 % Resolving information for `ssh'                 
amplxe: Warning: Cannot locate debugging symbols for file `/tmp/amplxe-tmp-atrianta/modules.mic-native_0/ssh/fb2032ff0cc20cd193fb35dcf39ca6a2/ssh'.
amplxe: Locating file `/boot/vmlinuz-2.6.32-573.18.1.el6.x86_64' on the remote system
amplxe: Warning: Cannot locate file `/lib64/libc-2.12.so'.
amplxe: Executing actions 18 % Resolving information for `libc-2.12.so'        
amplxe: Warning: Cannot locate file `/usr/lib64/libcrypto.so.1.0.1e'.
amplxe: Executing actions 19 % Resolving information for `libcrypto.so.1.0.1e'
amplxe: Warning: Cannot locate file `/opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-runss'.
amplxe: Executing actions 19 % Resolving information for `amplxe-runss'        
amplxe: Locating file `vmlinux-2.6.32-573.18.1.el6.x86_64' on the remote system
amplxe: Locating file `/boot/vmlinuz' on the remote system
amplxe: Warning: Cannot locate file `/usr/lib/debug/lib/modules/2.6.32-573.18.1.el6.x86_64/vmlinux'.
amplxe: Executing actions 24 % Resolving information for `vmlinux'             
amplxe: Symbol resolution has been finished, elapsed time is 4.027 seconds.
amplxe: Executing actions 26 % Processing profile metrics and debug information
amplxe: Deferred data transformations have been finished, elapsed time is 0.012 seconds.
amplxe: Executing actions 27 % Setting data model parameters                   
amplxe: Data model parameters have been set, elapsed time is 0.006 seconds.
amplxe: Executing actions 35 % Updating precomputed scalar metrics             
amplxe: Precomputing frequently used data has been finished, elapsed time is 0.032 seconds.
amplxe: Executing actions 41 % Saving the result                               
amplxe: Redundant overtime data has been discarded, elapsed time is 0.004 seconds.
amplxe: Raw collector data has been discarded, elapsed time is 0.000 seconds.
amplxe: Executing actions 50 % Saving the result                               
amplxe: Finalizing the result took 4.626 seconds.
amplxe: Executing actions 50 % Generating a report                             

General Exploration Metrics
---------------------------
Parameter  r113ah
---------  ------

Collection and Platform Info
----------------------------
Parameter                 r113ah                    
------------------------  --------------------------
Application Command Line  /home/atrianta/run.sh     
Operating System          2.6.32-573.18.1.el6.x86_64
Computer Name             cedar                     
Result Size               1568205                   

CPU
---
Parameter          r113ah                    
-----------------  --------------------------
Name               Intel Xeon Phi coprocessor
Logical CPU Count  8                         

Summary
-------
Elapsed Time:  0.000

Event summary
-------------
Hardware Event Type  Hardware Event Count:Self  Hardware Event Sample Count:Self  Events Per Sample
-------------------  -------------------------  --------------------------------  -----------------
cycles                                  432000                               108  4000             
amplxe: Executing actions 100 % done 

However, VTune analyzes without a problem /tmp/matrix.mic (from the example here: https://software.intel.com/en-us/articles/how-to-analyze-xeon-phi-coprocessor-applications-using-intel-vtune-amplifier-xe-2015).

Can anyone help me figure out this problem?

To be more specific,

ldd command for my programm on PHI resulted on this :

    linux-vdso.so.1 =>  (0x00007fffde5ff000)
    libz.so.1 => /usr/lib64/libz.so.1 (0x00007fe086917000)
    libimf.so => not found
    libsvml.so => not found
    libirng.so => not found
    libintlc.so.5 => not found
    libm.so.6 => /lib64/libm.so.6 (0x00007fe0866e8000)
    libiomp5.so => not found
    libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00007fe0863ee000)
    libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fe0861dc000)
    libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fe085fbf000)
    libc.so.6 => /lib64/libc.so.6 (0x00007fe085c67000)
    libdl.so.2 => /lib64/libdl.so.2 (0x00007fe085a63000)
    /lib64/ld-linux-k1om.so.2 (0x00007fe086b2b000)

So I had copied from /opt/intel/composer_xe_2015.1.133/compiler/lib/mic/ folder the "not found" libraries to /tmp/ of PHI (as I did with libiomp5.so for the matrix.mic example).

IF IT HELPS, I also ran VTune after i set the following variables :
  export AMPLXE_LOG_LEVEL=TRACE
  export AMPLXE_LOG_DIR=./log

but it stacks for my program on :
/opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-cl -v -collect=advanced-hotspots -target-system=mic-native:0 -- /home/atrianta/run.sh
  amplxe: Using target: mic-native:0
  TOTAL time:  48.723 secs  = 0.812051 mins
  amplxe: Using result path `/home/atrianta/r118ah'
  amplxe: Executing actions  8 % Clearing the database                           
  amplxe: The database has been cleared, elapsed time is 0.127 seconds.
  amplxe: Executing actions 16 % Updating precomputed scalar metrics             
  amplxe: Raw data has been loaded to the database, elapsed time is 0.157 seconds.
  amplxe: Executing actions 16 % Processing profile metrics and debug information
  amplxe: Data transformations have been finished, elapsed time is 0.000 seconds.
  amplxe: Executing actions 16 % Resolving module symbols                        
  amplxe: Locating file `/lib64/libpthread-2.12.so' on the remote system
  amplxe: Locating file `/lib64/libc-2.12.so' on the remote system
  amplxe: Locating file `/usr/lib64/libcrypto.so.1.0.1e' on the remote system
  amplxe: Locating file `/lib64/libselinux.so.1' on the remote system
  amplxe: Locating file `/usr/lib/debug/lib/modules/2.6.32-573.18.1.el6.x86_64/vmlinux' on the remote system
  amplxe: Warning: Cannot locate file `/lib64/libpthread-2.12.so'.
  amplxe: Executing actions 17 % Resolving information for `libpthread-2.12.so'  
  amplxe: Warning: Cannot locate file `/lib64/libc-2.12.so'.
  amplxe: Executing actions 17 % Resolving information for `libc-2.12.so'        
  amplxe: Warning: Cannot locate file `/usr/lib64/libcrypto.so.1.0.1e'.
  amplxe: Executing actions 18 % Resolving information for `libcrypto.so.1.0.1e'
  amplxe: Warning: Cannot locate file `/lib64/libselinux.so.1'.
  amplxe: Executing actions 19 % Resolving information for `libselinux.so.1'     
  amplxe: Locating file `/boot/vmlinuz-2.6.32-573.18.1.el6.x86_64' on the remote system
  amplxe: Locating file `vmlinux-2.6.32-573.18.1.el6.x86_64' on the remote system
  amplxe: Locating file `/boot/vmlinuz' on the remote system
  amplxe: Warning: Cannot locate file `/usr/lib/debug/lib/modules/2.6.32-573.18.1.el6.x86_64/vmlinux'.
  amplxe: Executing actions 19 % Resolving information for `vmlinux'
and for matrix.mic on :   
/opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-cl -v -collect=advanced-hotspots -target-system=mic-native:0 -- /home/atrianta/run1
  amplxe: Using target: mic-native:0
  Addr of buf1 = 0x7fd62df50010
  Offs of buf1 = 0x7fd62df50180
  Addr of buf2 = 0x7fd626ecf010
  Offs of buf2 = 0x7fd626ecf1c0
  Addr of buf3 = 0x7fd61fe4e010
  Offs of buf3 = 0x7fd61fe4e100
  Addr of buf4 = 0x7fd618dcd010
  Offs of buf4 = 0x7fd618dcd140
  Threads #: 228 OpenMP threads
  Matrix size: 3840
  Using multiply kernel: multiply1
  Execution time = 32.422 seconds
  amplxe: Using result path `/home/atrianta/r117ah'
  amplxe: Executing actions  8 % Clearing the database                           
  amplxe: The database has been cleared, elapsed time is 0.125 seconds.
  amplxe: Executing actions 16 % Updating precomputed scalar metrics             
  amplxe: Raw data has been loaded to the database, elapsed time is 0.161 seconds.
  amplxe: Executing actions 16 % Processing profile metrics and debug information
  amplxe: Data transformations have been finished, elapsed time is 0.000 seconds.
  amplxe: Executing actions 16 % Resolving module symbols                        
  amplxe: Locating file `/lib64/libpthread-2.12.so' on the remote system
  amplxe: Locating file `/lib64/libkrb5.so.3.3' on the remote system
  amplxe: Locating file `/usr/lib64/libcrypto.so.1.0.1e' on the remote system
  amplxe: Locating file `/lib64/ld-2.12.so' on the remote system
  amplxe: Locating file `/lib64/libc-2.12.so' on the remote system
  amplxe: Locating file `/opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-runss' on the remote system
  amplxe: Locating file `/usr/lib/debug/lib/modules/2.6.32-573.18.1.el6.x86_64/vmlinux' on the remote system

when I ran with -no-auto-finalize flag the 2 above commands collection finishes normally.

Logs are in log.zip attachment.

0 Kudos
4 Replies
Andreas_T_1
Beginner
319 Views

I just added flag -knob collection-detail=stack-sampling and now it's ok.

full command : /opt/intel/vtune_amplifier_xe_2015.1.1.380310/bin64/amplxe-cl -v -collect=advanced-hotspots -knob collection-detail=stack-sampling -target-system=mic-native:0 -- /home/atrianta/run.sh

I don't know why nut it ran...

0 Kudos
Dmitry_P_Intel1
Employee
319 Views

Hello,

The point why  -knob collection-detail=stack-sampling helped is that in this case another driver (and actually collection flow) works. But this is available only for advanced-hotspots..

Could you please provide the following output:

On host:

<VTune_install_dir>/bin64/sep -version -mic

and on target:

>lsmod | grep sep

Also I would highly recommend to update VTune to 2016 U4 - since 2015 version we had a bunch of fixes etc.

Thanks & Regards, Dmitry

0 Kudos
Andreas_T_1
Beginner
319 Views

><VTune_install_dir>/bin64/sep -version -mic
Sampling Enabling Product version: 3.15 (private) built by patbbinn on Oct 26 2014 01:38:16
SEP User Mode Version: 3.15.5

>lsmod | grep sep
sep3_15                45337  0

 

0 Kudos
Dmitry_P_Intel1
Employee
319 Views

Hello,

It seems that in early versions  "sep -version -mic" did not show info about available cards and drivers that are installed there etc.

Is it still possible to upgrade the VTune to VTune to 2016 U4 to see if it works and get output from "sep" command on this.

Thanks & Regards, Dmitry

 

0 Kudos
Reply