[flang-commits] [flang] [flang][openmp] Fix `not yet implemented intrinsic` message for omp_lib (PR #71101)

[via flang-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-flang-fir-hlfir

Author: Razvan Lupusoru (razvanlupusoru)

<details>
<summary>Changes</summary>

PR#<!-- -->70386 removed hardcoded omp_lib name when checking if it is intrinsic module procedure reference - but instead relied on bind(c) to avoid generating error when the implementation is external to module.

However, this change only worked for HLFIR. Make it work for FIR flow and update the test to exercise that as well.

---
Full diff: https://github.com/llvm/llvm-project/pull/71101.diff


2 Files Affected:

- (modified) flang/lib/Lower/ConvertExpr.cpp (+2-1) 
- (modified) flang/test/Lower/OpenMP/omp-lib-num-threads.f90 (+2) 


``````````diff
diff --git a/flang/lib/Lower/ConvertExpr.cpp b/flang/lib/Lower/ConvertExpr.cpp
index 76d810e9df6dc2d..8c2318632f725b1 100644
--- a/flang/lib/Lower/ConvertExpr.cpp
+++ b/flang/lib/Lower/ConvertExpr.cpp
@@ -2534,7 +2534,8 @@ class ScalarExprLowering {
             procRef.proc().GetSpecificIntrinsic())
       return genIntrinsicRef(procRef, resultType, *intrinsic);
 
-    if (Fortran::lower::isIntrinsicModuleProcRef(procRef))
+    if (Fortran::lower::isIntrinsicModuleProcRef(procRef) &&
+        !Fortran::semantics::IsBindCProcedure(*procRef.proc().GetSymbol()))
       return genIntrinsicRef(procRef, resultType);
 
     if (isStatementFunctionCall(procRef))
diff --git a/flang/test/Lower/OpenMP/omp-lib-num-threads.f90 b/flang/test/Lower/OpenMP/omp-lib-num-threads.f90
index 01c3e93d97bfcc0..e1af7d375dfa5a8 100644
--- a/flang/test/Lower/OpenMP/omp-lib-num-threads.f90
+++ b/flang/test/Lower/OpenMP/omp-lib-num-threads.f90
@@ -1,5 +1,7 @@
 ! RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - 2>&1 | FileCheck %s
 ! RUN: bbc -fopenmp -emit-hlfir -o - %s 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -emit-fir -fopenmp %s -o - 2>&1 | FileCheck %s
+! RUN: bbc -emit-fir -fopenmp -o - %s 2>&1 | FileCheck %s
 !
 ! Test that the calls to omp_lib's omp_get_num_threads and omp_set_num_threads
 ! get lowered even though their implementation is not in the omp_lib module

``````````

</details>


https://github.com/llvm/llvm-project/pull/71101