Developer Guide

Contents

Loops

The
Intel® oneAPI
DPC++/C++
Compiler
attempts to maximize the occupancy of the datapath of a loop within a DPC++ task kernel by executing iterations in a pipeline parallel method. The following sections provide guidelines and describe techniques for writing loops in DPC++ task kernels such that the
Intel® oneAPI
DPC++/C++
Compiler
can best extract pipeline parallelism from these loops.

Product and Performance Information

1

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