Re: opannotate -s -i example AND duplicated ophelp output on ARM Cortex-A7 MPCore

On 11/12/2015 10:30 AM, Panagiotis Mousouliotis wrote:
> Hello all,
> 
> I'm using the latest version of oprofile (1.1.0).
> 
> I would like to get source code details for a specific symbol using "opannotate -s -i <symbol>", but it seems that I can't get it right.
> 
> e.g.
> 
> $  opreport -g -l -t 1
> Using /home/pmousoul/Downloads/opencv3/opencv/build/bin/oprofile_data/samples/ for samples directory.
> 
> WARNING! Some of the events were throttled. Throttling occurs when
> the initial sample rate is too high, causing an excessive number of
> interrupts.  Decrease the sampling frequency. Check the directory
> /home/pmousoul/Downloads/opencv3/opencv/build/bin/oprofile_data/samples/current/stats/throttled
> for the throttled event names.
> 
> CPU: Intel Ivy Bridge microarchitecture, speed 2700 MHz (estimated)
> Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask of 0x00 (No unit mask) count 90000
> warning: some functions compiled without debug information may have incorrect source line attributions
> samples  %        linenr info                 image name               symbol name
> 2338141  61.0359  fast.cpp:56                 libopencv_features2d.so.3.0.0 void cv::FAST_t<16>(cv::_InputArray const&, std::vector<cv::KeyPoint, std::allocator<cv::KeyPoint> >&, int, bool)
> 761503   19.8786  fast_score.cpp:121          libopencv_features2d.so.3.0.0 int cv::cornerScore<16>(unsigned char const*, int const*, int)
> 346500    9.0452  (no location information)   libz.so.1.2.8            /lib/x86_64-linux-gnu/libz.so.1.2.8
> 305026    7.9625  (no location information)   libpng12.so.0.50.0       /lib/x86_64-linux-gnu/libpng12.so.0.50.0
> 
> 
> $ opannotate -s -i "void cv::FAST_t<16>(cv::_InputArray const&, std::vector<cv::KeyPoint, std::allocator<cv::KeyPoint> >&, int, bool)"
> 
> Doing the above I get the following output
> 
> $ opannotate -s -i "void cv::FAST_t<16>(cv::_InputArray const&, std::vector<cv::KeyPoint, std::allocator<cv::KeyPoint> >&, int, bool)"
> Using /home/pmousoul/Downloads/opencv3/opencv/build/bin/oprofile_data/samples/ for session-dir
> 
> WARNING! Some of the events were throttled. Throttling occurs when
> the initial sample rate is too high, causing an excessive number of
> interrupts.  Decrease the sampling frequency. Check the directory
> /home/pmousoul/Downloads/opencv3/opencv/build/bin/oprofile_data/samples/current/stats/throttled
> for the throttled event names.
> 
> /*
>  * Command line: opannotate -s -i void cv::FAST_t<16>(cv::_InputArray const&, std::vector<cv::KeyPoint, std::allocator<cv::KeyPoint> >&, int, bool)
>  *
>  * Interpretation of command line:
>  * Output annotated source file with samples
>  * Output all files
>  *
>  * CPU: Intel Ivy Bridge microarchitecture, speed 2700 MHz (estimated)
>  * Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask of 0x00 (No unit mask) count 90000
>  */
> 
> 
> What I'm I doing wrong?

Hi,

You need to tell opannotate where the source files are. You might take a look the tutorial at http://people.redhat.com/wcohen/Fedora12OprofileTutorial.txt for an example on how to use opannotate.

> As for the ophelp duplicated output on ARM Cortex-A7 MPCore:
> 
> 
> 
> $ ophelp
> oprofile: available events for CPU type "ARM Cortex-A7"
> 
...

> There are a lot of duplicates on the above ophelp output - is this normal?

In the armv7-ca7 file there is:

include:arm/armv7-common
include:arm/armv7-ca5

The arm/armv7-ca7 also has:

include:arm/armv7-common

Thanks for pointing that out.  A fix has been checked into the upstream oprofile git repository.

-Will

------------------------------------------------------------------------------