tf_to_gpu kernel reproducer

powderluv · powderluv · commit 6c3f36ce11b3 · 2020-09-23T22:43:30.000-07:00
diff --git a/LLVM-IR-E2E/tf_abs/mlir_reproducer_pleaseremovehostname-e6b4553d-2112-5b00898d0d240.mlir b/LLVM-IR-E2E/tf_abs/mlir_reproducer_pleaseremovehostname-e6b4553d-2112-5b00898d0d240.mlir
@@ -0,0 +1,10 @@
+// configuration: -pass-pipeline='func(xla-legalize-tf{allow-partial-conversion=false device-type=INVALID_DEVICE_TYPE legalize-chlo=true use-tf2xla-fallback=false}, materialize-broadcast, unfuse-batch-norm), unknown<mlir::mhlo::{anonymous}::HloLegalizeToLhlo>{results-escape-function=true}, func(buffer-placement, unknown<{anonymous}::CopyRemovalPass>), shape-to-descriptors, canonicalize, func(unknown<mlir::{anonymous}::LhloLegalizeToLinalgPass>, unknown<mlir::lmhlo::{anonymous}::LhloFuseLinalgPass>{tile-sizes= use-parallel-loops=true}, convert-linalg-to-parallel-loops, canonicalize, cse, unknown<xla::mlir_gpu::{anonymous}::FuseInnerParallelLoopsPass>, cse, unknown<xla::mlir_gpu::{anonymous}::StoreForwardingPass>, unknown<xla::mlir_gpu::{anonymous}::DeadTempBufferRemovalPass>, canonicalize, cse, unknown<xla::mlir_gpu::{anonymous}::MapParallelLoopsPass>), convert-parallel-loops-to-gpu, func(canonicalize, cse, unknown<mlir::mhlo::{anonymous}::LegalizeTrigonometricToApproximationPass>, unknown<xla::mlir_gpu::{anonymous}::MoveScalarComputationsIntoGpuLaunchPass>), gpu-kernel-outlining, func(unknown<xla::mlir_gpu::{anonymous}::RewriteKernelSignaturePass>), lower-affine, convert-scf-to-std'
+// note: verifyPasses=true
+
+
+module {
+  func @abs(%arg0: tensor<*xf32>) -> tensor<*xf32> {
+    %0 = "tf.Abs"(%arg0) : (tensor<*xf32>) -> tensor<*xf32>
+    return %0 : tensor<*xf32>
+  }
+}
