Programming Guide

Contents

Optimization Flags for CPU Architectures

In offline compilation mode, optimization flags can be used to produce code aimed to run better on a specific CPU architecture. Those are passed via the
-Xsycl-target-backend dpcpp
option:
dpcpp -fsycl-targets=spir64_x86_64-unknown-linux-sycldevice -Xsycl-target-backend=spir64_x86_64-unknown-linux-sycldevice “<CPU optimization flags>” a.cpp b.cpp -o app.out
Supported CPU optimization flags are:
-march=<instruction_set_arch> Set target instruction set architecture: 'sse42' for Intel(R) Streaming SIMD Extensions 4.2 'avx2' for Intel(R) Advanced Vector Extensions 2 'avx512' for Intel(R) Advanced Vector Extensions 512
The set of supported optimization flags may be changed in future releases.

Product and Performance Information

1

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