dynamism_op_counter->GetCell(op->getName().getStringRef().str())
          ->IncrementBy(1);
    }

    return WalkResult::advance();
  });

  if (has_dynamic_op) {
    dynamism_function_counter->GetCell(kDynamicFunctionName)->IncrementBy(1);
  } else {
    dynamism_function_counter->GetCell(kNotDynamicFunctionName)->IncrementBy(1);
  }
}
}  // namespace

std::unique_ptr<mlir::OperationPass<mlir::func::FuncOp>>

}

tensorflow::Status RecordStatusIfError(absl::Status status) {
  if (status.ok()) {
    return absl::OkStatus();
  }

  tf_dialect_to_executor_dialect_status->GetCell(kExportFailed)->IncrementBy(1);
  VLOG(1) << "Failed to export from TF Dialect to TF Executor Dialect. "
          << status;

  constexpr char bridge_subcomponent[] =
      "TFXLA_TF_FUNCTIONAL_TO_EXECUTOR_EXPORT_v2";

        mlir_function_pass_fallback_count->GetCell(kFailure)->IncrementBy(1);
      } else if (pass_state == MlirOptimizationPassState::Enabled) {
        return pass_status;
      }
    } else {
      if (pass_state == MlirOptimizationPassState::FallbackEnabled) {
        mlir_function_pass_fallback_count->GetCell(kSuccess)->IncrementBy(1);
      }
    }

  auto op_name = op->getName();
  auto name_string = op_name.getStringRef().str();
  auto op_legality = OperationLegalityString(op, target);

  mlir_failed_legalization_count->GetCell(name_string, op_legality)
      ->IncrementBy(1);
}

mlir::LogicalResult ApplyPatterns(Operation *op, RewritePatternSet &patterns,
                                  bool legalize_chlo) {
  ConversionTarget target =

      compilation_result);
  if (comp_status.ok()) {
    phase2_bridge_compilation_status->GetCell(kOldBridgeNoMlirSuccess)
        ->IncrementBy(1);
  } else {
    phase2_bridge_compilation_status->GetCell(kOldBridgeNoMlirFailure)
        ->IncrementBy(1);
  }

  return comp_status;
}

absl::Status CompileMLIRTFFunction(
    tpu::MlirToHloArgs mlir_computation,

  void runOnOperation() override;
};

static void IncrementCounterFor(tensorflow::monitoring::Counter<1>* counter,
                                Operation* op) {
  counter->GetCell(op->getName().getStringRef().str())->IncrementBy(1);
}

bool HasBounds(RankedTensorType type) {
  auto encoding = mlir::dyn_cast_or_null<mlir::mhlo::TypeExtensionsAttr>(
      type.getEncoding());
  return (encoding && !encoding.getBounds().empty());
}

  }

  if (quant_specs_.post_training_quantization) {
    tflite_quantizer_usage_stats->GetCell("post_training")->IncrementBy(1);
    RemoveRedundantStats(func);
  } else {
    tflite_quantizer_usage_stats->GetCell("during_training")->IncrementBy(1);
    // Set the quantization parameters for the quantizable input nodes. If this
    // failed, return the function immediately. This is only required for

    rewriter.replaceOp(op, rewriter.create(state)->getResults());

    // TODO: b/290366702 - Temporarily added metrics for debugging.
    mlir_tf_quant_op_count->GetCell(std::string(op->getName().getStringRef()))
        ->IncrementBy(1);
    return success();
  }
};

// This pattern adds qint <-> int Cast to all qint operands and results for UQ
// ops.
class TFUniformQuantizedOpsPattern : public ConversionPattern {
 public:

                                "The number of composite op expanded.", "name");
}

void IncreaseOpExpansionExecuteCounterByOne(const std::string& op_name) {
  tf_core_op_expansion_op_counter->GetCell(op_name)->IncrementBy(1);
}

}  // namespace tensorflow

//===----------------------------------------------------------------------===//
// The pass to decompose unregistered TF ops with the TFR compose function.
//
namespace mlir {

  VLOG(1) << "XlaLocalLaunchOpBase::Compute "
          << Canonicalize(function_.name(), AttrSlice(&function_.attr()));
  xla_launch_counter->GetCell(platform_info_.device_type().type_string())
      ->IncrementBy(1);

  std::vector<const Tensor*> inputs = InputsFromContext(ctx);
  std::vector<XlaCompiler::Argument> xla_compiler_args;
  const XlaCompiler::CompilationResult* compilation_result;