User-Mandated or SIMD Vectorization

Developer Guide and Reference

Download PDF

ID 767251

Date 10/31/2024

Version

Public

Visible to Intel only — GUID: GUID-42986DEF-8710-453A-9DAC-2086EE55F1F5

View Details

User-Mandated or SIMD Vectorization

User-mandated or SIMD vectorization supplements automatic vectorization just like OpenMP parallelization supplements automatic parallelization. The following figure illustrates this relationship. User-mandated vectorization is implemented as a single-instruction-multiple-data (SIMD) feature and is referred to as SIMD vectorization.

NOTE:

The SIMD vectorization feature is available for both Intel® microprocessors and non-Intel microprocessors. Vectorization may call library routines that can result in additional performance gain on Intel® microprocessors than on non-Intel microprocessors.

SIMD vectorization uses the !$OMP SIMD directive to effect loop vectorization.

For example, if the following code appears in example1.f90, the compiler does not automatically vectorize the loop because of the unknown data dependence distance between I and 2*I .

subroutine add(A, N, X)
integer N, X 
real    A(N) 
DO I=X, N
  A(I) = A(I) + A(2*I) 
END DO 
End

The example is compiled with the following command:

ifx example1.f90 -c -nologo -qopt-report -qopt-report-file=stderr

The example gives output like the following, reporting that vectorization did not occur:

Global optimization report for: add_

LOOP BEGIN at exmaple1.f90 (4, 1)
    remark #15344: Loop was not vectorized: vector dependence prevents vectorization
    remark #15346: vector dependence: assumed FLOW dependence between a (5:3) and a (5:3)
    remark #25439: Loop unrolled with remainder by 4
LOOP END

LOOP BEGIN at example1.f90 (4, 1)
<Remainder loop>
LOOP END
=================================================================

Using the previous example, if you know that X is large enough that data A(I) and A(2*I) do not overlap within a reasonable number of iterations you can enforce vectorization of the loop using !$OMP SIMD. If you know that they do not overlap in at least 8 iterations you may additionally specify !$OMP SIMD SIMDLEN(8) to avoid vectorization that is too wide, which might lead to overlap.

You can update your code from the previous example to use !$OMP SIMD:

subroutine add(A, N, X) 
integer N, X 
real    A(N) 
! X may be 8 or more, so on overlap with 8 iterations at least
!$OMP SIMD SIMDLEN(8)
DO I=X, N
   A(I) = A(I) + A(2*I) 
END DO 
End

The example is compiled with the following command:

ifx example1.f90 -c -nologo -qopt-report -qopt-report-file=stderr –qopenp-simd

The example gives output like the following, reporting that vectorization did occur:

Global optimization report for: add_

LOOP BEGIN at example1.f90 (5, 1)
    remark #15301: SIMD LOOP WAS VECTORIZED
    remark #15305: vectorization support: vector length 8
LOOP END

LOOP BEGIN at example1.f90 (5, 1)
<Remainder loop for vectorization>
LOOP END
=================================================================

The difference between using !$OMP SIMD and auto-vectorization hints is that with !$OMP SIMD, the compiler generates a warning when it is unable to vectorize the loop. With auto-vectorization hints, actual vectorization is still under the discretion of the compiler, even when you use the hint !DIR$ VECTOR ALWAYS.

!$OMP SIMD has optional clauses to guide the compiler on how vectorization must proceed. Use these clauses appropriately so that the compiler obtains enough information to generate correct vector code. For more information on the clauses, see the !$OMP SIMD description.

Additional Semantics

Note the following points when using the !$OMP SIMD directive.

A variable may belong to zero or one of the following clauses: private, linear, or reduction.
Within the vector loop, an expression is evaluated as a vector value if it is private, linear, reduction, or it has a sub-expression that is evaluated to a vector value. Otherwise, it is evaluated as a scalar value (that is, broadcast the same value to all iterations). Scalar value does not necessarily mean loop invariant, although that is the most frequently seen usage pattern of scalar value.
A vector value may not be assigned to a scalar L-value. It is an error.
A scalar L-value may not be assigned under a vector condition. It is an error.

Restrictions on Using an !$OMP DECLARE SIMD Declaration

Vectorization depends on two major factors: hardware and the style of source code. When using the vector declaration, the following features are not allowed:

Locks, barriers, atomic construct, critical sections
Computed and assigned GOTO and SELECT CASE, SELECT TYPE, and SELECT RANK constructs (in some cases these may be supported and converted to IF statements)
The GOTO statement, into or out of a function
An ENTRY statement

Non-vector function calls are generally allowed within vector functions but calls to such functions are serialized lane-by-lane and so might perform poorly. Also, for SIMD-enabled functions it is not allowed to have side effects except writes by their arguments. This rule can be violated by non-vector function calls, so be careful executing such calls in SIMD-enabled functions and subroutines.

Formal parameters must be of the following data types:

(un)signed 8, 16, 32, or 64-bit integer
32- or 64-bit floating point
64- or 128-bit complex

Parent topic: Explicit Vector Programming

Select Your Language

Using Intel.com Search

Quick Links

Recent Searches

Advanced Search

Only search in

Developer Guide and Reference

User-Mandated or SIMD Vectorization

Additional Semantics

Restrictions on Using an !$OMP DECLARE SIMD Declaration

See Also