[TF:XLA] Split select HLO into array- and tuple-select.

Array select and tuple-select already are handled separately in all backends and HLO passes: Array select is an elementwise operation. The shapes of the to operands have the same dimensions. Tuple select does not define its own output, but instead forwards the true- or false- operand based on a scalar predicate operand.

This CL reflects this by adding a new kTupleSelect HLO. The XLA builder interface stays the same and dispatches based on the operand shapes.

No change in the operation semantics. This CL just splits the existing select operation into two opcodes and preserves the existing semantics.

HLO cost analysis is fixed to handle the two ops appropriately.

PiperOrigin-RevId: 203180342

1 files changed, 2 insertions, 1 deletions

diff --git a/tensorflow/compiler/xla/service/hlo_graph_dumper.cc b/tensorflow/compiler/xla/service/hlo_graph_dumper.cc
index 8856723f67..7a1372f929 100644
--- a/tensorflow/compiler/xla/service/hlo_graph_dumper.cc
+++ b/tensorflow/compiler/xla/service/hlo_graph_dumper.cc
@@ -966,6 +966,7 @@ ColorScheme HloDotDumper::GetInstructionColor(const HloInstruction* instr) {
     case HloOpcode::kRemainder:
     case HloOpcode::kRng:
     case HloOpcode::kRoundNearestAfz:
+    case HloOpcode::kSelect:
     case HloOpcode::kShiftLeft:
     case HloOpcode::kShiftRightArithmetic:
     case HloOpcode::kShiftRightLogical:
@@ -1001,7 +1002,7 @@ ColorScheme HloDotDumper::GetInstructionColor(const HloInstruction* instr) {
     case HloOpcode::kPad:
     case HloOpcode::kReshape:
     case HloOpcode::kReverse:
-    case HloOpcode::kSelect:
+    case HloOpcode::kTupleSelect:
     case HloOpcode::kTranspose:
       // De-emphasize scalar-shaped data movement ops and all data movement ops
       // inside fusion nodes, both of which are essentially free.