| +O3 | Invoke level 3 optimizations. See the section
“+O3” for
more information. |
| +O3 +Oparallel | Invoke level 3 optimizations and cause
the compiler to honor parallelism directives and pragmas from the
Exemplar programming model and to generate parallel code where appropriate.[1] If you compile with +O3
+Oparallel, be sure to also link
with +O3 +Oparallel
(if you link separately). See the section “+O[no]parallel” for more
information. |
| +Odataprefetch | Prefetch data referenced in loops. See the
section “+O[no]dataprefetch”
for more information. |
| +Ofltacc | Disable floating-point optimizations that can
result in numerical differences. See the cc(1) or f77(1) man page
for more information. |
| +Oinfo | Display information on the optimization process.
See the section “+O[no]info” for more information. |
| +Olibcalls | Use low-call-overhead versions of select library
routines. See the cc(1) or the f77(1) man page for more information. |
| +Onoautopar | Request that the compiler parallelize only
those loops with prefer_parallel
or loop_parallel directives or
pragmas. See the section “+O[no]autopar” for more information. |
| +Onolimit | Do not suppress optimizations that significantly
increase compile-time or consume large amounts of memory. See the
section “+O[no]limit”
for more information. |
| +Onoparmsoverlap | Optimize with the assumption that subprogram
arguments do not refer to the same memory. When this option can
be used, it allows the compiler to generate significantly faster
code. See the cc(1) or the f77(1) man page for more information. |
| +Oreport | Display optimization reports. See the section
“+O[no]report[=report_type]” for more information. |
| -Wl,-aarchive_shared | (For use when linking with the compiler driver)
Search the archive version of a library; if the archive version
is not available, search the shared version of the library. |
| -Wl,+FPD | (For use when linking with the compiler driver)
Enable sudden underflow (flush to zero) of denormalized values. |
Printable version
