[mlir] Slightly optimize bytecode op numbering

If the bytecode encoding supports properties, then the dictionary
attribute is always the raw dictionary attribute of the operation,
regardless of what it contains. Otherwise, get the dictionary attribute
from the op: if the op does not have properties, then it returns the raw
dictionary, otherwise it returns the combined inherent and discardable
attributes.

Author: Mogball

mlir/lib/Bytecode/Writer/IRNumbering.cpp

@@ -429,17 +429,18 @@ void IRNumberingState::number(Operation &op) {
   // Prior to a version with native property encoding, or when properties are
   // not used, we need to number also the merged dictionary containing both the
   // inherent and discardable attribute.
-  if (config.getDesiredBytecodeVersion() <
-          bytecode::kNativePropertiesEncoding ||
-      !op.getPropertiesStorage()) {
+  DictionaryAttr dictAttr;
+  if (config.getDesiredBytecodeVersion() >= bytecode::kNativePropertiesEncoding)
+    dictAttr = op.getRawDictionaryAttrs();
+  else
     dictAttr = op.getAttrDictionary();
-  }
   if (!dictAttr.empty())
     number(dictAttr);
 
   // Visit the operation properties (if any) to make sure referenced attributes
   // are numbered.
-  if (config.getDesiredBytecodeVersion() >= bytecode::kNativePropertiesEncoding &&
+  if (config.getDesiredBytecodeVersion() >=
+          bytecode::kNativePropertiesEncoding &&
       op.getPropertiesStorageSize()) {
     if (op.isRegistered()) {
       // Operation that have properties *must* implement this interface.