Lab 8: Thursday April 24, 2014

The code for this lab can be found in $UWHPSC/labs/lab8.

Lab started with a demo of using compiler flags and gdb to debug Fortran code, using the code in $UWHPSC/labs/lab8/demo.

Running it without any compiler flags gives no error by there is a NaN value in the results:

$ cd $UWHPSC/labs/lab8
$ make run

 The max value of y is   0.99973335466585400
 x(501) is    0.0000000000000000
 y(501) is                        NaN

Running it with compiler flags to catch floating point exceptions:

$ make debug
make: *** [debug] Floating point exception: 8

Once it’s compiled with the flags specified in the Makefile for the debug target, the debugger gdb can be used to run the code and figure out where it died:

$ gdb ./a.out
(gdb) run

Program received signal SIGFPE, Arithmetic exception.
0x0000000000400a6d in demo () at demo.f90:12
12          y(j) = sin(x(j)) / x(j)

(gdb) p j
$1 = 501

(gdb) p x(j)
$2 = 0

Many commands are available in gdb, see for example this documentation.

Note: On the Mac with the Mavericks OS, gdb has been replaced by lldb. See http://lldb.llvm.org/index.html for more information.

The page GDB TO LLDB COMMAND MAP gives a good summary of both gdb and lldb commands and the relation between them.

Debugging compile-time errors

The code in $UWHPSC/labs/lab8/problem1 does not compile. See if you can find and fix all the errors. See $UWHPSC/labs/lab8/problem1b for a corrected version (and see the README.txt file in that directory for some comments).

Debugging run-time errors

The code $UWHPSC/labs/lab8/problem2/array1.f90 does not run properly. See if you can find and fix all the errors. See $UWHPSC/labs/lab8/problem2/array1b.f90 for a corrected version and use diff to see the differences.