You can prevent parallelization on a loop-by-loop basis using the
no_parallel directive or pragma.
The form of this directive and pragma is shown i
n Table 6-4 “Form
of no_parallel directive
and pragma”.
Table 6-4 Form
of no_parallel directive
and pragma
| Language | Form |
|---|
| Fortran | C$DIR NO_PARALLEL |
| C | #pragma _CNX no_parallel |
Use these directives to prevent parallelization of the loop
that immediately follows them. Only parallelization is inhibited;
all other loop optimizations are still applied.
no_parallel
The following Fortran example illustrates the use of no_parallel:
DO I = 1, 1000
C$DIR NO_PARALLEL
DO J = 1, 1000
A(I,J) = B(I,J)
ENDDO
ENDDO |
In this example, parallelization of the J
loop is prevented. The I loop can
still be parallelized.
The +Onoautopar compiler
option is available to disable automatic parallelization but allows
parallelization of directive-specified loops. Refer to Chapter 7 “Controlling optimization”, and Chapter 9 “Parallel programming
techniques”, for more information
on +Onoautopar.
Printable version
