OpenACC Tutorial - Profiling: Difference between revisions

What is so important about hotspots in the code ?  

Amdahl's law says that "Parallelizing the most time-consuming routines (i.e. the hotspots) will have the most impact".

== Build the Sample Code == <!--T:10-->

For this example we will use code from the [https://github.com/calculquebec/cq-formation-openacc repositories]. Download the package and change to the '''cpp''' or '''f90''' directory. The object of this exercise is to compile and link the code, obtain an executable, and then profile it.

}}

After the executable is created, we are going to profile that code.

{{Callout

|title=<translate><!--T:6-->

}}

=== PGPROF Profiler  ===

[[File:Pgprof new0.png|thumbnail|300px|Starting a new PGPROF session|left  ]]

=== NVIDIA NVPROF Command Line Profiler  ===

NVIDIA also provides a command line version called NVPROF, similar to GPU prof

{{Command

|nvprof --cpu-profiling on ./cgi.x  

======== Data collected at 100Hz frequency

}}

== Compiler Feedback  ==

Before working on the routine, we need to understand what the compiler is actually doing by asking ourselves the following questions:

CXXFLAGS=-fast -Minfo=all,intensity,ccff LDFLAGS=${CXXFLAGS}

* Rebuild

{{Command

|make

pgc++ CXXFLAGS=-fast -Minfo=all,intensity,ccff LDFLAGS=-fast main.o -o cg.x -fast

}}

== Computational Intensity  ==

Computational Intensity of a loop is a measure of how much work is being done compared to memory operations.

== Understanding the code  ==

Let's look closely at the following code:

<syntaxhighlight lang="cpp" line highlight="1,5,10,12">

for(int i=0;i<num_rows;i++) {

}

</syntaxhighlight>  

Given the code above, we search for data dependencies:

* Does one loop iteration affect other loop iterations?

[[OpenACC Tutorial - Adding directives|Onward to the next unit: Adding directives]]<br>

[[OpenACC Tutorial|Back to the lesson plan]]