Make HloDataFlowAnalysis updatable after transforming the HLO graph.

Updating is possible if operands/uses or computation roots change in
the graph. Updating is not possible if instructions are deleted or if
new instructions are added.

Specific changes:
* Add verification methods for asserting invariants and checking the
  analysis after updating.
* Always add phi values at while instructions. Previously these were
  added only if the phi had different inputs. The advantage of using
  phi's unconditionally is that the set of values is fixed for a
  module. Updates due to changing operands/uses in the graph do not
  create new values.
* Store values in a vector rather than a map. With unconditional phi
  values, the number of HloValues is fixed so the values can be held
  in a vector with stable references to elements.

PiperOrigin-RevId: 164778750

1 files changed, 3 insertions, 3 deletions

diff --git a/tensorflow/compiler/xla/service/hlo_alias_analysis.cc b/tensorflow/compiler/xla/service/hlo_alias_analysis.cc
index 0c03d72752..ad2c1ba773 100644
--- a/tensorflow/compiler/xla/service/hlo_alias_analysis.cc
+++ b/tensorflow/compiler/xla/service/hlo_alias_analysis.cc
@@ -44,10 +44,10 @@ void HloAliasAnalysis::InitializeBufferSets() {
   std::unordered_map<HloValue::Id, const HloBuffer*> value_to_buffer;
 
   // Initially define a buffer for every HloValue in the module.
-  for (const HloValue* value : dataflow_analysis_->values()) {
+  for (const HloValue& value : dataflow_analysis_->values()) {
     HloBuffer* buffer = NewHloBuffer();
-    buffer->AddValue(*value);
-    value_to_buffer[value->id()] = buffer;
+    buffer->AddValue(value);
+    value_to_buffer[value.id()] = buffer;
   }
 
   // Construct the Instruction buffer set to contain the HloBuffers for each