Re: [PATCH, og8] Add OpenACC 2.6 `serial' construct support

2018-12-20 Thread Maciej W. Rozycki
Hi Chung-Lin, > My only issue is should "serial" really be promoted to such a visible > construct > in the middle-end? It's just a special case of parallel, and user debug errors > can be dealt with specifically. I don't see much value of it being preserved > past the front-ends into gimplify/omp-

Re: [PATCH, og8] Add OpenACC 2.6 `serial' construct support

2018-12-17 Thread Chung-Lin Tang
Hi Maciej, I don't think there's anything wrong with the patch itself; if the testcases pass, then it should implement the functionality correctly. My only issue is should "serial" really be promoted to such a visible construct in the middle-end? It's just a special case of parallel, and user deb

[PATCH, og8] Add OpenACC 2.6 `serial' construct support

2018-12-16 Thread Maciej W. Rozycki
The `serial' construct is equivalent to a `parallel' construct with clauses `num_gangs(1) num_workers(1) vector_length(1)' implied. Naturally these clauses are therefore not supported with the `serial' construct. All the remaining clauses accepted with `parallel' are also accepted with `seria