cache BoxedString*'s for a module's constant AST_TYPE::Str nodes. emit the...

cache BoxedString*'s for a module's constant AST_TYPE::Str nodes.  emit the pointers to these BoxedString's in the generated IR

src/codegen/ast_interpreter.cpp

@@ -1013,7 +1013,7 @@ Value ASTInterpreter::visit_set(AST_Set* node) {
 
 Value ASTInterpreter::visit_str(AST_Str* node) {
     if (node->str_type == AST_Str::STR) {
-        return boxString(node->str_data);
+        return source_info->parent_module->getStringConstant(node->str_data);
     } else if (node->str_type == AST_Str::UNICODE) {
         return decodeUTF8StringPtr(&node->str_data);
     } else {

src/codegen/codegen.h

@@ -72,7 +72,7 @@ struct GlobalState {
     llvm::Type* llvm_str_type_ptr;
     llvm::Type* frame_info_type;
     llvm::Type* llvm_clfunction_type_ptr, *llvm_closure_type_ptr, *llvm_generator_type_ptr;
-    llvm::Type* llvm_module_type_ptr, *llvm_bool_type_ptr;
+    llvm::Type* llvm_module_type_ptr, *llvm_bool_type_ptr, *llvm_boxed_string_type_ptr;
     llvm::Type* llvm_excinfo_type;
     llvm::Type* i1, *i8, *i8_ptr, *i32, *i64, *void_, *double_;
     llvm::Type* vector_ptr;

src/codegen/irgen/irgenerator.cpp

@@ -1013,7 +1013,12 @@ private:
 
     CompilerVariable* evalStr(AST_Str* node, UnwindInfo unw_info) {
         if (node->str_type == AST_Str::STR) {
-            return makeStr(&node->str_data);
+            llvm::Value* rtn
+                = embedConstantPtr(irstate->getSourceInfo()->parent_module->getStringConstant(node->str_data),
+                                   g.llvm_boxed_string_type_ptr);
+
+            return new ConcreteCompilerVariable(STR, rtn, true);
+
         } else if (node->str_type == AST_Str::UNICODE) {
             return makeUnicode(emitter, &node->str_data);
         } else {

src/codegen/runtime_hooks.cpp

@@ -153,6 +153,9 @@ void initGlobalFuncs(GlobalState& g) {
     g.frame_info_type = g.stdlib_module->getTypeByName("struct.pyston::FrameInfo");
     assert(g.frame_info_type);
 
+    g.llvm_boxed_string_type_ptr = g.stdlib_module->getTypeByName("class.pyston::BoxedString")->getPointerTo();
+    assert(g.llvm_boxed_string_type_ptr);
+
 #define GET(N) g.funcs.N = getFunc((void*)N, STRINGIFY(N))
 
     g.funcs.printf = addFunc((void*)printf, g.i8_ptr, true);

src/runtime/types.cpp

@@ -390,6 +390,25 @@ std::string BoxedModule::name() {
     }
 }
 
+Box* BoxedModule::getStringConstant(const std::string& ast_str) {
+    auto idx_iter = str_const_index.find(ast_str);
+    if (idx_iter != str_const_index.end())
+        return str_constants[idx_iter->second];
+
+    Box* box = boxString(ast_str);
+    str_const_index[ast_str] = str_constants.size();
+    str_constants.push_back(box);
+    return box;
+}
+
+extern "C" void moduleGCHandler(GCVisitor* v, Box* b) {
+    boxGCHandler(v, b);
+
+    BoxedModule* d = (BoxedModule*)b;
+
+    v->visitRange((void* const*)&d->str_constants[0], (void* const*)&d->str_constants[d->str_constants.size()]);
+}
+
 extern "C" Box* boxCLFunction(CLFunction* f, BoxedClosure* closure, bool isGenerator,
                               std::initializer_list<Box*> defaults) {
     if (closure)
@@ -1859,8 +1878,8 @@ void setupRuntime() {
     function_cls->simple_destructor = builtin_function_or_method_cls->simple_destructor = functionDtor;
 
 
-    module_cls = new BoxedHeapClass(object_cls, NULL, offsetof(BoxedModule, attrs), 0, sizeof(BoxedModule), false,
-                                    new BoxedString("module"));
+    module_cls = new BoxedHeapClass(object_cls, &moduleGCHandler, offsetof(BoxedModule, attrs), 0, sizeof(BoxedModule),
+                                    false, new BoxedString("module"));
     member_cls
         = new BoxedHeapClass(object_cls, NULL, 0, 0, sizeof(BoxedMemberDescriptor), false, new BoxedString("member"));
     capifunc_cls

src/runtime/types.h

@@ -529,6 +529,11 @@ public:
     BoxedModule(const std::string& name, const std::string& fn, const char* doc = NULL);
     std::string name();
 
+    Box* getStringConstant(const std::string& ast_str);
+
+    llvm::StringMap<int> str_const_index;
+    std::vector<Box*> str_constants;
+
     DEFAULT_CLASS(module_cls);
 };