Development Reference Guides

Contents

Intel's valarray Implementation

The
Intel® oneAPI
DPC++/C++
Compiler
provides a high performance implementation of specialized one-dimensional valarray operations for the C++ standard STL valarray container.
The standard C++ valarray template consists of array/vector operations for high performance computing. These operations are designed to exploit high performance hardware features such as parallelism and achieve performance benefits.
Intel's valarray implementation uses the Intel® Integrated Performance Primitives (Intel® IPP), which is part of the product. Select IPP when you install the product.
The valarray implementation consists of a replacement header,
<valarray>,
that provides a specialized, high-performance implementation for the following operators and types:
Operator
Type
abs, acos, acosh, asin, asinh, atan, atan2, atanh, cbrt, cdfnorm, ceil, cos, cosh, erf, erfc, erfinv, exp, expm1, floor, hypot, inv, invcbrt, invsqrt, ln, log, log10, log1p, nearbyint, pow, pow2o3, pow3o2, powx, rint, round, sin, sinh, sqrt, tan, tanh, trunk
float, double
add, conj, div, mul, mulbyconj, mul, sub
Ipp32fc, Ipp64fc
addition, subtraction, division, multiplication
float, double
bitwise or, and, xor
(all unsigned) char, short, int
min, max, sum
signed or short/signed int, float, double

Use valarray in Source Code

This content is specific to C++; it does not apply to
DPC++
.
Intel's valarray implementation allows you to declare huge arrays for parallel processing. Improved implementation of valarray is tied up with calling the IPP libraries that are part of Intel® Integrated Performance Primitives (Intel® IPP).
To use valarrays in your source code, include the valarray header file,
<valarray>
. The
<valarray>
header file is located in the path
<installdir>/perf_header
.
The following example shows a valarray addition operation (+) specialized through use of Intel's implementation of valarray:
#include <valarray> void test( ) { std::valarray<float> vi(N), va(N); … vi = vi + va; //array addition … }
To use the static merged library containing all CPU-specific optimized versions of the library code, you need to call the
ippStaticInit
function first, before any IPP calls. This ensures automatic dispatch to the appropriate version of the library code for Intel® processor and the generic version of the library code for non-Intel processors at runtime. If you do not call
ippStaticInit
first, the merged library will use the generic instance of the code. If you are using the dynamic version of the libraries, you do not need to call
ippStaticInit
.

Compiling valarray Source Code

To compile your valarray source code, the compiler option,
/Quse-intel-optimized-headers
(for Windows*) or
-use-intel-optimized-headers
(for Linux*), is used to include the required valarray header file and all the necessary IPP library files.
The following examples illustrate how to compile and link a program to include the Intel valarray replacement header file and link with the Intel® IPP libraries. Refer to the Intel® IPP documentation for details.
In the following examples, "merged" libraries refers to using a static library that contains all the CPU-specific variants of the library code.

Linux* OS Examples

The following command line performs a one-step compilation for a system based on Intel® 64 architecture, running Linux OS:
icpx -use-intel-optimized-headers source.cpp
The following command lines perform separate compile and link steps for a system based on Intel® 64 architecture, running Linux OS:
so (dynamic):
icpx -use-intel-optimized-headers -c source.cpp
icpx source.o -use-intel-optimized-headers -shared-intel
Merged (static):
icpx -use-intel-optimized-headers -c source.cpp
icpx source.o -use-intel-optimized-headers

Windows* OS Examples

The following command line performs a one-step compilation for a system based on IA-32 architecture, running Windows OS:
icx /Quse-intel-optimized-headers source.cpp
The following command lines perform separate compile and link steps for a system based on IA-32 architecture, running Windows OS:
DLL (dynamic):
icx /Quse-intel-optimized-headers /c source.cpp
icx source.obj /Quse-intel-optimized-headers
Merged (static):
icx /Quse-intel-optimized-headers /Qipp-link:static /c source.cpp icx source.obj /Quse-intel-optimized-headers /Qipp-link:static
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.
Notice revision #20201201

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.