FPCW

[Lyndon_L_]

Hi,

   I got a running-time error "forrtl: error (65): floating invalid" when the application is run with debug mode ( compiled with -g -fpe0) and am wondering where the exception (division by zero) exactly happened in the program. I read some articles on fpcw, but do not know how to use it. This application is compiled with ifort in Linux.

    Thanks.

Cheers,

Lyndon.

[Heinz_B_Intel]

Hello Lyndon you should add -traceback option. Your application will display a stack trace including source line information (as long as you compile with -g) when the exeception occurs. An alternatve would be running your application within a debugger like GDB or Intel Debugger IDB. Search user manual for 'traceback' to find more information. Heinz

[TimP]

As you didn't include -mia32 in your compile options, you have no assurance (with ifort versions of recent years) that you are running in x87 mode. Are you assuming we can guess which compiler you use? The suggestion from Heinz looks more relevant. You may also wish to add -check in your compilation.

[Lyndon_L_]

Hi, I am using Intel 11.1.046 compiler and have included flags -traceback while compiling. The printed source file seems not right when the program crashed. When it said that the crashing point is in the sub-routine A, I printed the outputs for all variables and there is no any problems with the data. So, I suspect that the error "floating point invalid" should happen somewhere in the programs other than the sub-routine A. Is there a way to exactly diagnose where the error "floating point invalid" happened? Cheers, Lyndon.