Cross stack access

Cross stack access

Hi all,

I'm using OpenMP to parallelize a portion of my code, in a way that one subroutine is fine-grained parallelized, and other two subroutines are being called from the main program as a coarse grained parallelization.

Attached is an Intel Inspector screen shot which tells me that I do not have data race, but at least one of my threads access the stack of other thread, i.e., cross stack access.

How it can be resolved ?

Thank you,

Jack. 

4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

It should be noted that the cross stack access is in the fine grained parallel subroutine.

Thank you in advance,

Jack.

Attachments: 

I was seeing something very similar to these in my app also (on Linux) and my best guess was they were false positives, as they seemed to be in local stack variables allocated within the threads.

Thanks John.

This was also my best guess, since I didn't observe any changes in the results of my app.

Best,

Jack.

Leave a Comment

Please sign in to add a comment. Not a member? Join today