Intel-Specific Pragmas

The Intel-specific C++ compiler pragmas described in the Intel-Specific Pragma reference are listed below.

Some pragmas are available for both Intel® and non-Intel microprocessors but they may perform additional optimizations for Intel® microprocessors than they perform for non-Intel microprocessors.

Click on the pragmas for a more detailed description.

Pragma	Description
alloc_section	allocates variable in specified section
cilk grainsize	specifies the grain size for one `cilk_for` loop
distribute_point	instructs the compiler to prefer loop distribution at the location indicated
inline	instructs the compiler that the user prefers that the calls in question be inlined
intel_omp_task	for Intel legacy tasking, specifies a unit of work, potentially executed by a different thread
intel_omp_taskq	for Intel legacy tasking, specifies a unit of work, potentially executed by a different thread
ivdep	instructs the compiler to ignore assumed vector dependencies
loop_count	indicates what the loop count is likely to be
nofusion	Prevents a loop from fusing with adjacent loops
novector	specifies that the loop should never be vectorized
offload	causes the statement following the pragma to execute on the target; this pragma only applies to Intel® MIC Architecture.
offload_attribute	specifies that all functions and variables declared subsequent to the pragma are available on the coprocessor; this pragma only applies to Intel® MIC Architecture.
offload_transfer	initiates asynchronous data transfer, or initiates and completes synchronous data transfer; this pragma only applies to Intel® MIC Architecture.
offload_wait	specifies a wait for a previously initiated asynchronous activity; this pragma only applies to Intel® MIC Architecture.
omp atomic	ensures that a specific memory location is updated atomically, which prevents the possibility of multiple, simultaneous reading and writing of threads
omp task	defines a task region
omp taskyield	enables or disables optimizations for specific functions; provides some degree of compatibility with Microsoft's implementation of optimize pragma
omp taskwait	specifies a wait on the completion of child tasks generated since the beginning of the current task
optimize	enables [disables] optimizations for code after this pragma till another optimize pragma or end of the translation unit is encountered
optimization_level	indicates change of optimization from current line of the compilation unit
optimization_parameter	tells the compiler to generate code specialized for a particular processor, at the function level, similar to the -m (/arch) option
parallel/noparallel	facilitates auto-parallelization of an immediately following loop; using keyword [always] forces the compiler to auto-parallelize; noparallel pragma prevents auto-parallelization
prefetch/noprefetch	asserts that the data prefetches are generated or not generated for some memory references; this pragma only applies to Intel® MIC Architecture.
simd	instructs the compiler to vectorize the loop below
unroll/nounroll	instructs the compiler the number of times to unroll/not to unroll a loop
unroll_and_jam/nounroll_and_jam	instructs the compiler to partially unroll higher loops and jam the resulting loops back together. Specifying the nounroll_and_jam pragma prevents unrolling and jamming of loops.
unused	describes variables that are unused (warnings not generated)
vector	indicates to the compiler that the loop should be vectorized according to the arguments: `always/aligned/unaligned/nontemporal/temporal`