Simple elapsed-time timers: begin, end, print.

The elapsed time is obtained by reading the timebase 
register and converting to seconds just before printing.
This keeps overhead down to ~40 nsec for 32-bit, ~20 nsec
for 64-bit mode.  Initialization is done on the first
call to Tbeg.  All you have to do is call with matching
labels for begin/end pairs; then print at the end.

--------------------------
C syntax:

void Tbeg(char *);
void Tend(char *);
void Tprt();

--------------------------
C example:

Tbeg("main");

Tbeg("do_work");
do_work(...);
Tend("do_work");

Tend("main");
Tprt();

--------------------------
Fortran example
  call tbeg('main')

  call tbeg('do_work')
  call do_work(...)
  call tend('do_work')

  call tend('main')
  call tprt()

