+O[no]loop_transform

+Oloop_transformenables transformation of eligible

 

loops for improved cache performance. The most

 

important transformation is the interchange of nested

 

loops to make the inner loop unit stride, resulting in fewer

 

cache misses. +Onoloop_transform disables

 

transformation of eligible loops. The default is

 

+Oloop_transform.

 

NOTE: This option is deprecated and not supported in

 

future releases. Use +inline_level.

+inline_level num

This option controls inlining in fortran. The format for num

 

is N[.n], where num is either an integral value from0

 

to 9 or a value with a single decimal place from 0.0to

 

9.0. For more information on this option, see

 

F90(1)manpage.

+O[no]loop_unroll[=factor]

+Oloop_unrollturns on loop unrolling. factor is the

 

unroll factor that controls the code expansion. The default

 

unroll factor is 4; that is, four copies of the loop body.

 

By experimenting with different factors, you may improve

 

the performance of your program. This option is only

 

effective at optimization level 2 or higher. The default is

 

+Oloop_unroll=4.

+O[no]loop_unroll_jam

+loop_unroll_jamenables loop unrolling and

 

jamming. +Onoloop_unroll_jam (the default) disables

 

both automatic and directive-specified unroll and jam.

 

Loop unrolling and jamming increases register

 

exploitation.

+O[no]moveflop

+Omoveflops allows the optimizer to move conditional

 

floating-point instructions, enabling other optimizations

 

to occur. This option is only effective at optimization level

 

2 or higher.

 

The behavior of floating-point exception handling may

 

be altered by this option.

 

Using +Onomoveflops is recommended if floating-point

 

traps are enabled and you do not want the behavior of

 

floating-point exceptions to be altered by the relocation

 

of floating-point instructions, as when your program uses

 

the ON statement. The default is +Omoveflops.

+Oprefetch

+Omultiprocessortells the compiler to appropriately

 

optimize several different processes on multiprocessor

 

machines. The optimizations are those appropriate for

 

executables and or shared libraries.

 

+Onomultiprocessor, the default, disables the

 

optimization of more than one process running on a

 

multiprocessor machine.

 

NOTE: This option is valid only on the PA-RISC systems.

+O[no]openmp

+Oopenmp allows users to enable the OpenMP directives.

 

+Onoopenmpwill disable the OpenMP directives.

 

+O[no]openmpis accepted at all optimization levels.

 

The default is +Onoopenmp.

+O[no]parallel

+Oparallel causes the compiler to transform eligible

 

loops for parallel execution on multiprocessor machines.

 

This option is effective only at optimization level 3 or

 

higher. This option is available only on HP 9000 systems.

 

If you link separately from the command line and compile

 

the program with the +Oparallel option, you must link

 

with the f90 command and specify the

 

+Oparalleloption to link in the correct runtime support.

 

The +Onoparallel option disables parallelization for

 

the target program. It is the default at all levels of

 

optimization. The +Oparallel option is not available

46 Compiling and linking