Hi,
In GUI tool(amplxe-gui), I can get "GPU usage", "CPU and GPU Active", "GPU Active" and etc. on CPU/GPU Concurrency mode.

However, I can't get these information on command line tool(amplxe-cl).

How to use amplxe-cl to get "GPU usage", "CPU and GPU Active", "GPU Active" and etc. ?

here is my command and result:

./amplxe-cl -report summary -r /root/intel/amplxe/projects/test/r041cgc

CPU
---
Parameter          r041cgc
-----------------  -------------------------------------
Name               Intel(R) Processor code named Skylake
Frequency          3000000203
Logical CPU Count  8

GPU
---
Parameter            r041cgc
-------------------  -------------------------------------------------
Name                 Display controller: Intel Corporation Device 5912
Vendor               Intel Corporation
EU Count             24
Max EU Thread Count  7
Max Core Frequency   1000000000

Summary
-------
Elapsed Time:             6.432
Paused Time:              0.0
CPU Time:                 6.198
Average CPU Utilization:  1.191
CPI Rate:                 0.596

Average DRAM Bandwidth
----------------------
Package    Total, GB/sec:Self
---------  ------------------
package_0               1.226

Event summary
-------------
Hardware Event Type                 Hardware Event Count:Self  Hardware Event Sample Count:Self  Events Per Sample
----------------------------------  -------------------------  --------------------------------  -----------------
INST_RETIRED.ANY                                  34812000000                             11604  3000000
CPU_CLK_UNHALTED.THREAD                           20742000000                              6914  3000000
CPU_CLK_UNHALTED.REF_TSC                          18594000000                              6198  3000000
CPU_CLK_UNHALTED.REF_XCLK                           148804464                              1488  100003
CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE                  147704431                              1477  100003

Uncore Event summary
--------------------
Uncore Event Type         Uncore Event Count:Self
------------------------  -----------------------
UNC_IMC_DRAM_DATA_READS                  89380439
UNC_IMC_DRAM_DATA_WRITES                 33807073

GPU Usage
---------
GPU Engine        GPU Time:Self
----------------  -------------
Render and GPGPU          0.198
amplxe: Executing actions 100 % done

vtune version: 2019 (build 570779)

I have a trial version vtune_amplifier_2019.1.0.579888. A server is on RHEL 7.4. I want to install command-line tools.

1) It complained about not enough space in /opt, though it is not true:

$ df -h /opt/
Filesystem                 Size  Used Avail Use% Mounted on
xxx   50G   24G   27G  47% /opt
2) I tried to do silent installation. Config file:

ACCEPT_EULA=accept
CONTINUE_WITH_OPTIONAL_ERROR=yes
PSET_INSTALL_DIR=/path/to/intel
CONTINUE_WITH_INSTALLDIR_OVERWRITE=yes
PSET_MODE=install
CLUSTER_INSTALL_AUTOMOUNT=yes
AMPLIFIER_SAMPLING_DRIVER_INSTALL_TYPE=kit
AMPLIFIER_DRIVER_ACCESS_GROUP=vtune
AMPLIFIER_DRIVER_PERMISSIONS=666
AMPLIFIER_LOAD_DRIVER=yes
AMPLIFIER_C_COMPILER=/usr/bin/gcc
AMPLIFIER_KERNEL_SRC_DIR=/lib/modules/3.10.0-693.21.1.el7.x86_64/build
AMPLIFIER_MAKE_COMMAND=/usr/bin/make
AMPLIFIER_INSTALL_BOOT_SCRIPT=yes
ACTIVATION_TYPE=license_file
ACTIVATION_LICENSE_FILE=/path/to/xxx.lic
AMPLIFIER_DRIVER_PER_USER_MODE=no
COMPONENTS=;intel-vtune-amplifier-2019-cli-common__noarch;intel-vtune-amplifier-2019-common__noarch;intel-vtune-amplifier-2019-cli__x86_64;intel-vtune-amplifier-2019-cli-32bit__i486;intel-vtune-amplifier-2019-collector-32linux__i486;intel-vtune-amplifier-2019-collector-64linux__x86_64;intel-vtune-amplifier-2019-doc__noarch;intel-vtune-amplifier-2019-sep__noarch;intel-vtune-amplifier-2019-target__noarch;intel-vtune-amplifier-2019-gui-common__noarch;intel-vtune-amplifier-2019-gui__x86_64;intel-vtune-amplifier-2019-vpp-server__x86_64;intel-vtune-amplifier-2019-common-pset

Although CONTINUE_WITH_OPTIONAL_ERROR is yes, it complained about xorg:

Missing optional prerequisites

-- Xorg X11 server is not found

3) Tried option --cli-mode:

#  ./install.sh --cli-mode -s d4install.cfg
--------------------------------------------------------------------------------
Please select at least one component before you continue.
-------------------------------------------------------------------------------- 

4) Then I installed xorg. Still the same:

#  ./install.sh -s d4install.cfg
--------------------------------------------------------------------------------
Please select at least one component before you continue.
-------------------------------------------------------------------------------- 

5) Tried interactive installation. The same result.

Hi Intel 

I am Running a simple enough python code ; while doing Vtune hotspot analysis I found many functions could not get linked to source code. while some were.

while looking at the collection log I found warning that "Cannot locate debugging information for file /usr/bin/python2.7".

Since my most work revolves around Python 2.7 and 3.x ; getting debug symbols will be very helpful in optimizing my apps.

can you help me how /where can I get debug symbols for Python 2.7 and 3.x.?  

I am running it on Ubuntu 16.04 server with Linux fmxxx 4.10.0-42-generic #46~16.04.1-Ubuntu SMP Mon Dec 4 15:57:59 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

thanks

Ankit

Hi,

I use QuickWin to provide graphical out during execution in order to accelerate development and find errors.This is a invaluable tool for e.g. physical problems. I now want to profile the code and wonder if there are any special precautions to be taken at implementation of Vtune Amplifier. Is Vtune 2019 stable or should I rely on earlier versions? Graphics will be removed during production runs, so maybe profiling should be performed with graphics removed giving a simpler profiling output?

I am very grateful for suggestions how to proceed. I have understood that there are different ways of attaching Vtune to the executable. Any suggestions here?

With thanks in beforehand

Anders S

Hi,

I am using Vtune amplifier_2017_for_systems.0.2.480304. This was installed as part of Intel Systems Studio 2017.

Whenever I try to profile a PID running in the system, I am getting this error

"Error: [Instrumentation engine]: Source/pin/injector_nonmac/auxvector.cpp: CopyAux:291: Unexpected AUX VEC Type 26

Collection failed.

Internal error. "

This software is running on l_ubuntu 16.04 OS.

I have another machine which has same OS and ISS 2017, IN that Vtune amplifier works perfectly without any errors.

Is there any fix for this error?

VTune 2019 U1 fails to analyze the data collected by a JIT virtual machine.  The application's profiling machinery used to work in several versions of VTune since at least 2017, but for some reason now VTune 2019 U1 fails to finalize the data after the application exits normally.  More interestingly, VTune 2018 and 2017 also fail the same way now.  This behavior happens consistently on multiple physical machines.  Restarting VTune and opening the result causes VTune to try to finalize the result, and again VTune crashes.  Different versions of VTune now crash the same way, irrespective of which version is doing the collection.

I had recently updated the VTune SDK, rolling back to a previous version did not fix the problem on multiple VTune versions --- even the ones in which the setup did work.

The crash report notes a multitude of other errors: "we only support MD5 and SHA1 now" (?!), failure to open files, inability to deal with clock tick events, and so on.

Does anyone know what's going on here?  That VTune 2019 U1 fails, well ok that may be a problem.  But why would older versions of VTune also fail, seemingly just because VTune 2019 U1 is installed?

Hi ,

in the sommer 2018 I have installed VTune_Amplifier_2018_update2_setup and my operating system is Windows 10.

At the beginning VTune works fine and all was OK. 

For two weeks now I cannot profile with VTune. I can start the tool, but if I want to start a profile session I always get the error message: 

"Amplifier cannot detect local machine configuration"

I removed the installation and installed it again but the error message has remained.

Best regards, 

Heidi

.

Is it possible to produce output similar to perf script from a trace collected with Intel PT (with timestamps) in Linux command line?

Example perf commands:

perf record -e intel_pt// ls
perf script --ns

Output:

ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffbcec2 dl_main (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffc6090 _dl_new_object (/lib/x86_64-linux-gnu/ld-2.27.so)
ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffc60bc _dl_new_object (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffd92f0 strlen (/lib/x86_64-linux-gnu/ld-2.27.so)
ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffd9329 strlen (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffc60c1 _dl_new_object (/lib/x86_64-linux-gnu/ld-2.27.so)
ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffc60fc _dl_new_object (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffbaec0 calloc@plt (/lib/x86_64-linux-gnu/ld-2.27.so)
ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffbaec0 calloc@plt (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffd55e0 calloc (/lib/x86_64-linux-gnu/ld-2.27.so)
ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffd55f5 calloc (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffd55fc calloc (/lib/x86_64-linux-gnu/ld-2.27.so)
ls 10068 [001] 385286.629487960: 1 branches: 7fc98ffd55fc calloc (/lib/x86_64-linux-gnu/ld-2.27.so) =>     7fc98ffbaea0 malloc@plt (/lib/x86_64-linux-gnu/ld-2.27.so)

Reproduce command:

amplxe-cl -quiet -collect-with runsa -knob event-config=CPU_CLK_UNHALTED.THREAD,CORE_POWER.LVL0_TURBO_LICENSE,CORE_POWER.LVL1_TURBO_LICENSE,CORE_POWER.LVL2_TURBO_LICENSE -data-limit=0 -finalization-mode=none -result-dir="/export/users/linliu/vtune" -target-tmp-dir="/export/users/linliu/xxx" -no-summary -- ls

Vtune still writes temporary files to /tmp directory instead of the directory I specified. Temporary files of vtune on our benchmarks are too large to be stored in the root directory. As a result, it always throws error "no space left on device".

Anybody knows about this problem? Much appreciated!

Hi team, I got a vtune crash shows as follow, anyone can help?

amplxe-cl -collect gpu-hotspots $(which python) run.py

================

amplxe: Collection stopped.
amplxe: Using result path `/tmp/r002gh'
amplxe: Executing actions  0 %
amplxe: Warning: The result contains a lot of raw data. Finalization may take a long time to complete.
amplxe: Executing actions 12 % Loading '17787-17793.0.trace' file              Intel(R) VTune(TM) Amplifier 2019 (build 579888) feedback tool
Copyright (C) 2009-2018 Intel Corporation. All rights reserved.

Intel(R) VTune(TM) Amplifier 2019 Update 1; 579888 experienced an unexpected error.

Please send a problem report:
amplxe-feedback --send-crash-report "/tmp/amplxe-log-root/2018-12-17-14-53-30-722803.amplxe-cl/crash_info.txt"

Report data may be used to improve product stability. Our apologies for the inconvenience and thank you for your assistance.

Assertion failed: reinterpret_cast<const tpss_u8_t*>(binaryFileName)-binary_file_names_buf == binary_file_names_buf_size - 1
file: vcs/tpssplug2/src/gpu/tp_compute_cbk_utils.cpp:63
Collecting modules information...
Collecting system information...
[1]    17707 abort (core dumped)  amplxe-cl -collect gpu-hotspots $(which python)

When attempting to use VTune to do software profiling of my application, I receive the following error and the application freezes (though does not appear to crash).  It seems that the assertion is coming from inside vtunes sampler somewhere?

Assertion Failed: runtime_gen_2:165: thread_state.runtime_level == tpss_le_out_of_runtime : thread_state.runtime_level == 2 . Please contact the technical support.

I need to run a DPDK application under vtune amplifier. It needs to be run with super user privileges. This application is run using a script as it needs some environment variables to be set.

Usually I run this script by switching the user to root.

How can I do this?

Things that I've tried:
+ Run vtune amplifier as root, but this fails to launch any application saying 'License check failed'.
+ Set suid bit on the script, but this fails to run the application.

I tried profiling the same C++ code in two different projects to find the CPU Hotspots. The profiling results for the two projects were different. What could be the possible reason for this.(I used 'Local Host' option in 'Where'  and  'Launch Application' option in 'What')

Thanks

When performing event-based hotspot investigation with Amplifier 2019 Update 2, I get the following error:

"Failed to finalize the result

Cannot finalize the result. Error 0x40000003 (Unexpected internal error / invalid_state) -- boost::lock_error: The operation completed successfully"

This occurs on Windows 10 Version 1803 (OS Build 17134.472)

Hi,

I'm trying to understand the "Instructions Retired" column (not the "Retiring"-class under locators) in Microarchitecture Exploration in VTune.

From the bottom-up-view, I've found a function that I've interested in, and I'm looking at a block of assembly. There is no control flow, so I would guess that each of these instructions would successfully retire the same number of times.

However, the value under "instructions retired" vary quite a lot for this sequence of instructions. One explanation might be sampling noise, but I wonder if my assumption that this column should somehow correlate with the number of that particular instruction that retires is broken, or I have the wrong mental model here.

If anyone would shed some light on this, I would be grateful.

hi, 

I always get the following errors on Ubuntu, is there any way to work around this?

amplxe: Error: [Instrumentation Engine]: munmap() is not suitable for probed function insertion.
amplxe: Collection failed.
amplxe: Internal Error

I am using 

Intel(R) VTune(TM) Amplifier 2019 Update 2 (build 584348) Command Line Tool
Copyright (C) 2009-2018 Intel Corporation. All rights reserved

on 

Distributor ID:    Ubuntu
Description:    Ubuntu 16.04.5 LTS
Release:    16.04
Codename:    xenial

Thanks in advance :)

I'd like to try Intel vtune to profile an application, time is short to do this work.

Here are the details of the issue -

I received an Intel vtune software evaluation email with a serial# to register.

I attempted to register using the serial# to allow a trial download of vtune.

When I enter the serial# and click the Register button an error message "Serial number is already registered." is displayed.

Now what?

Thank you in advance for your time,

Stewart

Hey,

I am attempting to profile a threaded version of a Fortran simulation using VTune. The threading was done using OpenMP and the Intel Fortran compiler. After completing data collection I get the following error message during finalization: 

Cannot load data file 'C:\Users\...\r001hs\data.0\3740-14804.0.trace' (createEventInstance: wrong uniqueTid)

This leads to inaccurate results showing wrong total wall time or wrong thread count/histogram.

I managed to do a couple of analyses before but, since this started happening, I don't seem to be able to redo them without this error. Also, I don't get this error when I profile the single thread version. I don't know where to start with troubleshooting this so any tips will be appreciated. 

Software and system information:

• I am using Intel VTune Amplifier GUI 2019 Update 2 Product Build 588069. Analysis type is Hotspots and its running on my local machine. Also, I always run it as admin.
• Compiler: Intel fortran compiler 19.0 update 1 for the Visual Studio 2017 environment. I compile with the following options: 

/nologo /O2 /fpp /Qopenmp /module:"Release\\" /object:"Release\\" /Fd"Release\vc150.pdb" /libs:static /threads /c /Zi /debug:inline-debug-info

• My OS is:  64 bit Windows 10 Enterprise version 10.0.17134 Build 17134.
• My processor is: Intel Core i7-7800X @ 3.50GHz with 12 Logical Processors

Regards,

Ghassan
 

Hi all,

upon analyzing even a simple python script with the VTune Amplifier, I get the following python error and no performance analysis: 

Fatal Python error: ceval: tstate mix-up

I used the analysis set up as described in this manual: https://software.intel.com/en-us/vtune-amplifier-help-python-code-analysis

The python script looks as follows:

import numpy as np

def my_func():
    x = np.array([1,2,3])
    y = np.array([1,2,3])

    for i in range(500):
        x += y

if __name__ == '__main__':
    my_func()

Am I missing something? Any help is appreciated.

Thank you!

System specifications:

Win 10
VTune Amplifier 2019 Update 2
IntelPython 2

It is fairly easy to find the percentage of vectorised floating point operations in an uninstrumented code (no debug symbols) including vector length (AVX or not). How to find if a code is vectorising integer operations? Preferably using HW counters.

More generally, what to look for in a code which is computationally intensive but has relatively little FP ops? Are there suitable templates for initial analysis?