[PATCH, v2, OpenMP 5.0] Implement relaxation of implicit map vs. existing device mappings (for mainline trunk)

[Jakub Jelinek]

On Sat, Nov 06, 2021 at 12:51:59AM +0800, Chung-Lin Tang wrote:
>  static int
>  get_kind (bool short_mapkind, void *kinds, int idx)
>  {
> -  return short_mapkind ? ((unsigned short *) kinds)[idx]
> -		       : ((unsigned char *) kinds)[idx];
> +  int val = (short_mapkind
> +	     ? ((unsigned short *) kinds)[idx]
> +	     : ((unsigned char *) kinds)[idx]);
> +
> +  if (short_mapkind && GOMP_MAP_IMPLICIT_P (val))
> +    val &= ~GOMP_MAP_IMPLICIT;
> +  return val;
> +}

It doesn't make sense to test it twice.  I'd do:
  if (!short_mapkind)
    return ((unsigned char *) kinds)[idx];
  int val = ((unsigned short *) kinds)[idx];
  if (GOMP_MAP_IMPLICIT_P (val))
    val &= ~GOMP_MAP_IMPLICIT;
  return val;

> +
> +
> +static bool
> +get_implicit (bool short_mapkind, void *kinds, int idx)
> +{
> +  int val = (short_mapkind
> +	     ? ((unsigned short *) kinds)[idx]
> +	     : ((unsigned char *) kinds)[idx]);
> +
> +  return short_mapkind && GOMP_MAP_IMPLICIT_P (val);
>  }

and here even simpler, no need to read kinds at all:
  if (!short_mapkind)
    return false;
  int val = ((unsigned short *) kinds)[idx];
  return GOMP_MAP_IMPLICIT_P (val);
?

Otherwise LGTM.

	Jakub