Rewriting generic-mode kernel with a customized state machine. [OMP131]

This optimization remark indicates that a generic-mode kernel on the device was specialized for the given target region. When offloading in generic-mode, a state machine is required to schedule the work between the parallel worker threads. This optimization specializes the state machine in cases where there is a known number of parallel regions inside the kernel. A much simpler state machine can be used if it is known that there is no nested parallelism and the number of regions to schedule is a static amount.

Examples

This optimization should occur on any generic-mode kernel that has visibility on all parallel regions, but cannot be moved to SPMD-mode and has no nested parallelism.

#pragma omp declare target
int TID;
#pragma omp end declare target

void foo() {
#pragma omp target
{
 TID = omp_get_thread_num();
 #pragma omp parallel
 {
   work();
 }
}
}

$ clang++ -fopenmp -fopenmp-targets=nvptx64 -O2 -Rpass=openmp-opt omp131.cpp
omp131.cpp:8:1: remark: Rewriting generic-mode kernel with a customized state machine. [OMP131]
#pragma omp target
^

Diagnostic Scope

OpenMP target offloading optimization remark.