What is the Difference

[newport_j]

I am very new to using the Intel VTUNE Amplifier XE and I have one question after using the getting started tutorials for Linux.

What is the difference between the exectables:

tachyon_analyze_locks

tachyon_find_hotspots

It is the same program compiled the same way yet you have two different executables instead of one. Why?
It seems that one executbale could be used for both analyses.

Newport_j

[Rob5]

The Tachyon example contains an executable for use with Locks and Waits analysis and another specific to the Hotsposts analysis. This follows the flow of the Intel VTune Amplifier XE 2011 Getting Started Tutorials (PDF) which walks through each of the examples using workflow methodology to resolve the issues in the respective Tachyon source code examples.

- Rob

[newport_j]

Yes, but why have two executables? I am assuming the source code for each is the same. So why not have one executbale and do the analysis for hotspots and locks on the same excutable?

What I am saying is for my code please tell me how I should compile for hotspots analysis vs locks analysis. When I compile my program, I only have one executable. There must be different options or flags set for each type of analysis.

Is there a unique way in the Makefile that allows compilati0on for hotspots analysis and a unique way for compilation for locks analysis.?

Thanks for any help in advance.

Newport_j

[Rob5]

I am not certain why there are two executable instead of one for the sample labs. Perhaps someone else on the forum can chime in on this.

In regards to compiler settings, it may depend on what you are attempting to accomplish, your environment, codebase, etc. The on-line help provides a good explanation with recommendations for compiler settings here as they specifically relate to Intel VTune Amplifier XE for Linux*. Additionally there are important notes / comments at the bottom of the help page that you may also find of interest.

- Rob

[robert-reed]

You know the trouble with "assume"? The fact is, the codes are NOT identical: one uses "analyze_locks.cpp" and the other "find_hotspots.cpp." The former has an OpenMP critical section, which is probably what they're using for the locks-and-waits demo. A quick glance at the hot-spot version suggested that it might not even be threaded (doesn't have the parallel_for that's in the analyze_locks version).