trouble with -O0 -mp and -mp1

[Intel_C_Intel]

Hi, I have a problem about the floating point precision control options of fortran linux compiler 7.0. When I compiled my code with default option -02 and other optimizations level options -01, -O3 and execute the executable; my executable gives the results in an output file that those results are floating point numbers which are correct according to my calculations. However, when I compile my code with -O0 or when I use -mp or -mp1 options with optimization levels
-O1, -O2 or -O3, the executable now fills the output file with question marks(?). If you can help me in that floationg point control option problems, I would be very glad...

bozbulut

[TimP]

I take it that you're compiling x87 code (no SSE options). You may be taking advantage of the extended range and precision of the 80-bit floating point format. When you remove the optimizations and force each assignment to be narrowed to declared (or default real) precision, you may be incurring over- or under- flow, from exceeding the declared range.