The difference between uninitialized variables according to the compute node.

[youn__kihang]

Hello All,

I'm writing a forum to inquire.
When performing MPI operation including a specific node, the program result is different from the usual result.
As a result of analysis at the source code level, it was found that uninitialized variables were created differently (a certain random small/large value) when a specific node was included and when not included.
However, if the node is not included in the calculation, the same random value will be always genetrated.

What I am curious about is whether the initial value of an uninitialized variable can be changed according to certain conditions. For example, library, intel version, etc.
*Resolving the random initial value issue does not guarantee that the above results will solve other problems.

Thank you all.

[AbhishekD_Intel · Accepted solution]

Hi Kihang,

I didn't get the whole use-case you have wanted to tell. But I want to add that we cannot predict the value of the uninitialized variables. There could be the scenario that you are getting the same value at each run on a specified node, but it may not be the same on all systems.

Due to this unpredictable nature of uninitialized variables, we cannot say for sure that the initial value of the uninitialized variable will depend on a certain condition or environment.

Warm Regards,

Abhishek

[AbhishekD_Intel]

Hi Kihang,

Please give us an update on the details.

Thank You,

Abhishek

[AbhishekD_Intel]

Hi,

We are assuming that the provided details have solved your issue and will no longer monitor this thread.

Please post a new thread if you have more issues.

Thank you.