I have a piece of code that ran in release mode and generated a floating point error in debug mode until I changed the debug mode "floating point model" compiler setting from "fast" to "strict." The code now runs but I am unsure as to whether I have fixed the problem or whether I have just masked it. I noticed some similar posts on the /fp:strict setting and should note that I am using version 11.0. If I've just masked the problem, can anyone suggest a way of looking for the cause - the floating point divide by zero error doesn't occur until much later in the program execution (ie I know that the problem is caused early in execution but doesn't cause a crash until much later).
Thanks as always for your assistance,

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

If the error occurs fordebug compilations, why not run your program from the Visual Studio and when the error pops up 'break' execution and see what iscausing of the error?
Also know where inexecution (loop, time, whatever) the error occurs and add some dumps might be enlightening as to what variable/calculation might be producing the error.
I expect the fast option v strict option is related tostorage of intermediate 'results' (INTEL chime in here).

Leave a Comment

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