SPEC CPU95 Optimization Flag Descriptions for AMD Results Using Compaq Visual Fortran and KAP Preprocessor 17 Jan 2000 /optimize:n Fortran's general optimization level, per n: 0 disable all optimizations 1 local optimizations and common subexpressions 2 global optimizations such as code motion, strength reduction, lifetime analysis, and code scheduling 3 additional global optimizations that may cost more space, such as loop unrolling and code replication 4 inline expansion 5 loop transformation /Ox Perform optimizations that favor speed over code size. /fast Generate code for fastest run-time performance. Equivalent to /assume:noaccuracy_sensitive /align:records /align:dcommons /math_library:fast /arch:host /tune:host /math_library:fast Require slightly less accuracy and less error checking. /arch:host Generate code that may use instructions available on the X86 processor model being used to compile the code which may not be available on other processor models. /tune:host Generates and schedules code optimized for the processor type in use on the x86 system being used for compilation. /align:dcommons Align the data items of all COMMON data blocks on natural boundaries. /align:records Align components of derived types and fields of records on natural boundaries. /assume:noaccuracy_sensitive Allow expression re-ordering. /unroll:n Specifies how many times inner loops are unrolled, up to n=16. /noautomatic Causes all local variables to be statically allocated. /nofpconstant Causes a single-precision constant assigned to a double-precision variable to be evaluated in single-precision. /nofltconsistency Allow reordering of floating-point operations and retention of operation results on the register stack. /notransform_loops Supress loop transformation optimizations. /WKscalaropt Run KAP preprocessor with default scalar (uniprocessor) optimizations. /WKnoprocess Suppress all KAP preprocessor optimizations.