[Mlir-commits] [mlir] [OpenMP][mlir] Add Groupprivate op in omp dialect. (PR #162704)

[llvmlistbot at llvm.org]

================
@@ -6030,7 +6030,7 @@ static bool isTargetDeviceOp(Operation *op) {
   // by taking it in as an operand, so we must always lower these in
   // some manner or result in an ICE (whether they end up in a no-op
   // or otherwise).
-  if (mlir::isa<omp::ThreadprivateOp>(op))
+  if (mlir::isa<omp::ThreadprivateOp, omp::GroupprivateOp>(op))
----------------
skc7 wrote:

In `convertOmpGroupprivate()`, `device_type` attribute and `host/device` compilation flag are checked, to decide whether to do allocation or use the global as-is is decided. So, it is not required to check for the same here.

https://github.com/llvm/llvm-project/pull/162704