[llvm-branch-commits] [flang] [mlir] [flang][OpenMP] Lower target in_reduction for host fallback (PR #199967)

[Sairudra More via llvm-branch-commits]

================
@@ -8222,6 +8226,61 @@ convertOmpTarget(Operation &opInst, llvm::IRBuilderBase &builder,
   bool isOffloadEntry =
       isTargetDevice || !ompBuilder->Config.TargetTriples.empty();
 
+  // Validate and resolve in_reduction clauses on omp.target. We currently
+  // only support the non-offload host-fallback path: the per-task private
+  // pointer is obtained by calling __kmpc_task_reduction_get_th_data inside
+  // the to-be-outlined target task body. Threading that pointer through the
+  // device kernel argument list is left as follow-up work.
+  SmallVector<llvm::Value *> inRedOrigPtrs;
+  if (!targetOp.getInReductionVars().empty()) {
+    if (isTargetDevice || isOffloadEntry)
----------------
Saieiei wrote:

I pushed a separate prototype commit for this direction. The target body now uses the matching map_entries block argument instead of a separate in_reduction block argument, while the in_reduction operands remain as host-side metadata. I kept it as a separate commit so it can be reviewed/rolled back independently if this direction does not hold up.


https://github.com/llvm/llvm-project/pull/199967