[flang-commits] [flang] [mlir] [flang][mlir][OpenMP] Add linear modifier (val, ref, uval) (PR #187142)

[via flang-commits]

================
@@ -8,7 +8,7 @@ subroutine do_simd
 !CHECK: %{{.*}} = arith.constant 1 : i32
 !CHECK: %[[IV_STEP:.*]] = arith.constant 1 : i32
 !CHECK: omp.wsloop {
-!CHECK: omp.simd linear(%[[X]]#0 : !fir.ref<i32> = %[[CONST]] : i32, %[[I]]#0 : !fir.ref<i32> = %[[IV_STEP]] : i32) private(@_QFdo_simdEi_private_i32 {{.*}} -> %arg0 : !fir.ref<i32>) {
+!CHECK: omp.simd linear(val(%[[X]]#0 : !fir.ref<i32> = %[[CONST]] : i32), val(%[[I]]#0 : !fir.ref<i32> = %[[IV_STEP]] : i32)) private(@_QFdo_simdEi_private_i32 {{.*}} -> %arg0 : !fir.ref<i32>) {
----------------
NimishMishra wrote:

I am confused by this check. The !RUN of this file does not have an explicit `-fopenmp-version=52`, but the generated MLIR still has `val()` attribute. Are we generating v5.2 specific MLIR irrespective of the OpenMP version? 

https://github.com/llvm/llvm-project/pull/187142