[flang-commits] [flang] [Flang][MLIR][OpenMP] Add MLIR lowering support for taskloop clauses. (PR #165851)

[Kaviya Rajendiran via flang-commits]

================
@@ -70,3 +92,106 @@ subroutine omp_taskloop_private
 ! CHECK:         }
   !$omp end taskloop
 end subroutine omp_taskloop_private
+
+!===============================================================================
+! `allocate` clause
+!===============================================================================
+
+! CHECK-LABEL:  func.func @_QPtaskloop_allocate
+! CHECK:           %[[ALLOCA_I:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFtaskloop_allocateEi"}
+! CHECK:           %[[DECL_I:.*]]:2 = hlfir.declare %[[ALLOCA_I]] {uniq_name = "_QFtaskloop_allocateEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK:           %[[ALLOCA_X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFtaskloop_allocateEx"}
+! CHECK:           %[[DECL_X:.*]]:2 = hlfir.declare %[[ALLOCA_X]] {uniq_name = "_QFtaskloop_allocateEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+subroutine taskloop_allocate()
+   use omp_lib
----------------
kaviya2510 wrote:

omp_high_bw_mem_alloc is available through the omp_lib module, so I used omp_lib.

https://github.com/llvm/llvm-project/pull/165851