[Mlir-commits] [flang] [mlir] [llvm][mlir][OpenMP] Support translation for linear clause in omp.wsloop and omp.simd (PR #139386)

[Tom Eccles]

================
@@ -695,6 +786,48 @@ llvm.func @simd_simple(%lb : i64, %ub : i64, %step : i64, %arg0: !llvm.ptr) {
 
 // -----
 
+// CHECK-LABEL: @simd_linear
+
+// CHECK: entry:
+// CHECK: %[[LINEAR_VAR:.*]] = alloca i32, align 4
+// CHECK: %[[LINEAR_RESULT:.*]] = alloca i32, align 4
+// CHECK: br label %omp.simd.region
+
+// CHECK: omp.simd.region:
+// CHECK: br label %omp_loop.preheader
+
+// CHECK: omp_loop.preheader:
+// CHECK: %[[ORIG_VAL:.*]] = load i32, ptr {{.*}}, align 4
+// CHECK: store i32 %[[ORIG_VAL]], ptr %[[LINEAR_VAR]], align 4
+// CHECK: br label %omp_loop.header
+
+// CHECK: omp_loop.body:
+// CHECK: %[[LINEAR_LOAD:.*]] = load i32, ptr %[[LINEAR_VAR]], align 4
----------------
tblah wrote:

Please could you test that these loads and stores do get the llvm access group metadata

https://github.com/llvm/llvm-project/pull/139386