> case NE_EXPR:
> - if (!flag_enable_cilk)
> + /* NE_EXPR is only allowed for Cilk Plus loops. */
> + if (flag_enable_cilk
> + && gimple_omp_for_kind (for_stmt) == GF_OMP_FOR_KIND_CILKSIMD)
> + break;
> + else
> gcc_unreachable ();
> - /* NE_EXPR is technically not allowed in OpenMP, but it is
> - allowed in Cilk Plus, which generates OMP_SIMD constructs. */
> - break;
Something I'm going to bring up wrt the main cilk+ patches, but checking
flag_enable_cilk is useless. One can only get KIND_CILKSIMD if cilk is
enabled, and the test itself is very inexpensive.
Better as
if (gimple_omp_for_kind (for_stmt) == GF_OMP_FOR_KIND_CILKSIMD)
{
// if you really thing we need to check it...
gcc_checking_assert (flag_enable_cilk);
break;
}
gcc_unreachable ();
Or, best as
gcc_checking_assert (flag_enable_cilk);
gcc_assert (gimple_omp_for_kind (for_stmt) == GF_OMP_FOR_KIND_CILKSIMD);
break;
since I believe the front end should have already checked this grammar.
> + bool cilkplus_block = false;
> + if (flag_enable_cilk)
> + {
> + if ((branch_ctx
> + && gimple_code (branch_ctx) == GIMPLE_OMP_FOR
> + && gimple_omp_for_kind (branch_ctx) & GF_OMP_FOR_KIND_CILKSIMD)
> + || (gimple_code (label_ctx) == GIMPLE_OMP_FOR
> + && gimple_omp_for_kind (label_ctx) & GF_OMP_FOR_KIND_CILKSIMD))
> + cilkplus_block = true;
> + }
The & tests here aren't right. Recall that CILKSIMD == 3 << 3, so that covers
DISTRIBUTE and normal SIMD as well. The only & tests should have been for SIMD
when one wants both SIMD and CILKSIMD.
r~
