Omit Hardware to Support the
no_global_work_offsetAttribute in
parallel_for Kernels
no_global_work_offset
Attribute in
parallel_for
KernelsThe
[[intel::no_global_work_offset(1)]]
kernel attribute instructs the
Intel® oneAPI
to omit generating hardware required to support global work offsets. This may improve area and/or throughput for all device kernels that are invoked using
DPC++/C++
Compilerparallel_for
without the
workItemOffset
parameter and that do not use
ndrange
objects with offsets.