Intel® oneAPI DPC++ FPGA Optimization Guide

Developer Guide

Contents

Disable Hardware Kernel Invocation Queue (
-Xsno-hardware-kernel-invocation-queue
)

To direct the
Intel® oneAPI
DPC++/C++
 Compiler
 to reduce kernel area use by removing kernel invocation queue in DPC++ kernel, include the
-Xsno-hardware-kernel-invocation-queue
 option in your
dpcpp
 command.
Example
dpcpp -fintelfpga -Xshardware -Xsno-hardware-kernel-invocation-queue <source_file>.cpp
Using this option may result in longer kernel execution time as the kernel invocation queue allows DPC++ runtime environment to queue kernel launches in accelerator so that the accelerator can start execution on the next invocation as soon as previous invocation of the same kernel is complete.
Use the
-Xsno-hardware-kernel-invocation-queue
 option only when your kernel execution time is much greater than the system and DPC++ runtime environment overhead hidden by the kernel invocation queue (20-100us), or if the
Intel® oneAPI
DPC++/C++
 Compiler
 has a difficulty fitting your kernel.
Refer to Utilizing Hardware Kernel Invocation Queue for more information about how to utilize the kernel invocation queue.

Product and Performance Information

1

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