Enabling Automatic Optimizations

-O1

/O1

Optimizes to favor smaller code size and code locality. In most cases, -O2 (Linux* OS and Mac OS* X) or /O2 (Windows* OS) is recommended over this option.

This optimization disables some optimizations that normally increase code size. This level might improve performance for applications with very large code size, many branches, and execution time not dominated by code within loops. In general, this optimization level does the following:

• Enables global optimization.

• Disables intrinsic recognition and inlining of intrinsics.

IA-64 architecture:

• The option disables software pipelining, loop unrolling, and global code scheduling.

-O2 or -O

/O2

Optimizes for code speed. Since this is the default optimization, if you do not specify an optimization level the compiler will use this optimization level automatically. This is the generally recommended optimization level; however, specifying other compiler options can affect the optimization normally gained using this level.

In general, the resulting code size will be larger than the code size generated using -O1 (Linux and Mac OS X) or /O1 (Windows).

This option enables the following capabilities for performance gain: inlining intrinsic functions, constant propagation, copy propagation, dead-code elimination, global register allocation, global instruction scheduling and control speculation, loop unrolling, optimized code selection, partial redundancy elimination, strength reduction/induction variable simplification, variable renaming, exception handling optimizations, tail recursions, peephole optimizations, structure assignment lowering optimizations, and dead store elimination.

For IA-32 and Intel 64 architectures:

• Enables certain optimizations for speed, such as vectorization.

IA-64 architecture:

• Enables optimizations for speed, including global code scheduling, software pipelining, predication, speculation, and data prefetch.

-O3

/O3

Enables -O2 (Linux and Mac OS X) or /O2 (Windows) optimizations, as well as more aggressive optimizations, including prefetching, scalar replacement, cache blocking, and loop and memory access transformations.

As compared to -O2 (Linux) or /O2 (Windows), the optimizations enabled by this option often result in faster program execution, but can slow down code execution in some cases. Using this option may result in longer compilation times.

This option is recommended for loop-intensive applications that perform substantial floating-point calculations or process large data sets.

-fast

/fast

Provides a single, simple optimization that enables a collection of optimizations that favor run-time performance.

This is a good, general option for increasing performance in many programs.

For IA-32 and Intel 64 architectures, the -xSSSE3 (Linux and Mac OS X) or /QxSSSE3 (Windows) option that is set by this option cannot be overridden by other command line options. If you specify this option along with a different processor-specific option, such as -xSSE2 (Linux) or /QxSSE2 (Windows), the compiler will issue a warning stating the -xSSSE3 or /QxSSSE3 option cannot be overridden; the best strategy for dealing with this restriction is to explicitly specify the options you want to set from the command line.

Caution

Programs compiled with the -xSSSE3 (Linux and Mac OS X) or /QxSSSE3 (Windows) option will detect non-compatible processors and generate an error message during execution.

While this option enables other options quickly, the specific options enabled by this option might change from one compiler release to the next. Be aware of this possible behavior change in the case where you use makefiles.

Linux and Mac OS X

icpc -O2 prog.cpp

Windows

icl /O2 prog.cpp