[Mlir-commits] [mlir] [MLIR][SCFToGPU] Guard operands before AffineApplyOp::create to avoid crash (PR #167959)

[Shashi Shankar]

================
@@ -0,0 +1,31 @@
+// RUN: mlir-opt %s --convert-parallel-loops-to-gpu | FileCheck %s
+
+// Goal: exercise the per-dim index computation
+//        newIndex = hardware_id * step + lowerBound
+// and ensure we see a gpu.launch and an affine.apply (no crash).
+
+module {
+  func.func @two_dim_parallel_mapped() {
----------------
shashforge wrote:

@linuxlonelyeagle  Done. I moved the tests into mlir/test/Conversion/SCFToGPU/parallel_loop.mlir

https://github.com/llvm/llvm-project/pull/167959