@tlemo Thanks for the quick review. I added another commit that just does small refactoring. 89521cc

@tlemo Thanks for the quick review. I added another commit that just does small refactoring. 89521cc

Looks fine to me, but what is the rationale for moving getPredicateFlagName to GridReduction?

@tlemo Thanks for the quick review. I added another commit that just does small refactoring. 89521cc

Looks fine to me, but what is the rationale for moving getPredicateFlagName to GridReduction?

That's because it's only used for GridReuction. Here's just one example usage. T2pred is the flag variable in this case.

  bool T2pred;
  // Allocate global tensor float T7[( T0.size[0] * 2 )];
  // Allocate global tensor int64_t T8[T0.size[0]];
  T2pred = reduction::gridReduce< false, true, false, false, false, true > ( T2[ 0 ], block_result, reduction_add_float, &T7[0], T8, reinterpret_cast<float*>(shared_mem));

Fix is good!