Add kGenerateToken HLO instruction.

The new HLO instruction serves two purposes. (1) It generates a new token value. This is the only way to create tokens. (2) The operation is variadic, taking zero or more token operands. The operation acts as a join of its operands.

I considered initially using a kConstant constant as a method to create new tokens, but this ran into problems because of expectations in backends regarding constants and their materialization.

This CL enables creation of generate-token instructions, but the new instruction is not supported yet in any backend.

PiperOrigin-RevId: 199836205

1 files changed, 11 insertions, 0 deletions

diff --git a/tensorflow/compiler/xla/service/shape_inference.cc b/tensorflow/compiler/xla/service/shape_inference.cc
index d624f548b1..fdc7f41759 100644
--- a/tensorflow/compiler/xla/service/shape_inference.cc
+++ b/tensorflow/compiler/xla/service/shape_inference.cc
@@ -463,6 +463,17 @@ StatusOr<Shape> InferWindowOutputShape(const Shape& base_shape,
   return ShapeUtil::MakeShape(element_type, new_dimensions);
 }
 
+/* static */ StatusOr<Shape> ShapeInference::InferTokenShape(
+    tensorflow::gtl::ArraySlice<const Shape*> arg_shapes) {
+  for (const Shape* arg_shape : arg_shapes) {
+    if (arg_shape->element_type() != TOKEN) {
+      return InvalidArgument(
+          "Operands of token instructions must be TOKEN types.");
+    }
+  }
+  return ShapeUtil::MakeTokenShape();
+}
+
 /* static */ StatusOr<Shape> ShapeInference::InferConvertShape(
     const Shape& operand_shape, PrimitiveType new_element_type) {
   auto old_element_type = operand_shape.element_type();