Port and integrate the irregexp engine from V8

runtime/vm/raw_object.h

Index: runtime/vm/raw_object.h
diff --git a/runtime/vm/raw_object.h b/runtime/vm/raw_object.h
index 19d0ef40b48b2aa3d5b3fc7dda0a4577ecf00707..5ad4d229d28783fadcfef876e3c676f2a675bec6 100644
--- a/runtime/vm/raw_object.h
+++ b/runtime/vm/raw_object.h
@@ -639,6 +639,7 @@ class RawFunction : public RawObject {
     kMethodExtractor,  // converts method into implicit closure on the receiver.
     kNoSuchMethodDispatcher,  // invokes noSuchMethod.
     kInvokeFieldDispatcher,  // invokes a field as a closure.
+    kIrregexpFunction,   // represents a generated irregexp matcher function.
   };
 
   enum AsyncModifier {
@@ -657,6 +658,7 @@ class RawFunction : public RawObject {
   RawString* name_;
   RawObject* owner_;  // Class or patch class or mixin class
                       // where this function is defined.
+  RawJSRegExp* regexp_;  // The corresponding regexp for irregexp functions.

[Florian Schneider, 2014/10/01 17:04:14]

Maybe re-use the owner field for this? It is not used otherwise, since those
functions are not member methods of a specific class.

[Vyacheslav Egorov (Google), 2014/10/01 20:13:21]

Kill this field by being more explicit about compilation instead of using lazy
compilation stub for this.

[jgruber1, 2014/10/03 18:59:52]

As discussed, the regexp object pointer is now stored in data_.

[jgruber1, 2014/10/03 18:59:53]

As discussed, the regexp object pointer is now stored in data_.

   RawAbstractType* result_type_;
   RawArray* parameter_types_;
   RawArray* parameter_names_;
@@ -684,6 +686,7 @@ class RawFunction : public RawObject {
   uint32_t kind_tag_;  // See Function::KindTagBits.
   uint16_t optimized_instruction_count_;
   uint16_t optimized_call_site_count_;
+  intptr_t regexp_specialization_;

[Florian Schneider, 2014/10/01 17:04:14]

As mentioned, try using the data_ field for that. It is a kind-specific field
that can be used for different purposes depending on function kind.

[Vyacheslav Egorov (Google), 2014/10/01 20:13:21]

Ditto.

[jgruber1, 2014/10/03 18:59:52]

As discussed, changed to int16_t and moved 3 lines up to avoid increasing the
size of RawFunction.

[jgruber1, 2014/10/03 18:59:53]

As discussed, changed to int16_t and moved 3 lines up to avoid increasing the
size of RawFunction.

 };
 
 
@@ -1776,8 +1779,16 @@ class RawJSRegExp : public RawInstance {
   RawSmi* data_length_;
   RawSmi* num_bracket_expressions_;
   RawString* pattern_;  // Pattern to be used for matching.
+  RawFunction* one_byte_function_;
+  RawFunction* two_byte_function_;
+  RawFunction* external_one_byte_function_;
+  RawFunction* external_two_byte_function_;
+  RawString* one_byte_subject_;

[Vyacheslav Egorov (Google), 2014/10/01 20:13:21]

This four subject fields smell like memory leak to me. Please kill them with
fire.

[jgruber1, 2014/10/03 18:59:53]

Done.

+  RawString* two_byte_subject_;
+  RawString* external_one_byte_subject_;
+  RawString* external_two_byte_subject_;
   RawObject** to() {
-    return reinterpret_cast<RawObject**>(&ptr()->pattern_);
+    return reinterpret_cast<RawObject**>(&ptr()->external_two_byte_subject_);
   }
 
   // A bitfield with two fields: