Parallel Programming Guide for HP-UX Systems

Parallel synchronization

Synchronization tools

Chapter 8 141

The form of the wait functions is shown in Table 8-6.

where (in Fortran)

barrier

is a barrier variable of the indicated type and

nthr

the number of threads calling the routine.

where (in C/C++)

barrier_p

is a pointer of the indicated type and

nthr

is a pointer

referencing the number of threads calling the routine.

You can use a barrier variable in multiple calls to the

wait function, if you ensure that two such barriers are

not simultaneously active. You must also verify that

nthr

reﬂects the correct number of threads.

sync_routine

Among the most basic optimizations performed by the HP compilers is

code motion. This optimization moves code across routine calls. If the

routine call is to a synchronization function that the compiler cannot

identify as such, and the code moved must execute on a certain side of it,

this movement may result in wrong answers.

The compiler is aware of all synchronization functions and does not move

code across them when they appear directly in code. However, if the

synchronization function is hidden in a user-deﬁned routine, the

compiler has no way of knowing about it and may move code across it.

Any time you call synchronization functions indirectly using your own

routines, you must identify your routines with a sync_routine directive

or pragma.

Table 8-6 Form of wait functions

Language Form

Fortran INTEGER FUNCTION WAIT_BARRIER(barrier,nthr)

C/C++ int wait_barrier(barrier_t *barrier_p,const int *nthr);