73 lines · plain
1Changes:2* Change the casting code to be const correct. Now, doing this is invalid:3 const Value *V = ...;4 Instruction *I = dyn_cast<Instruction>(V);5 instead, the second line should be:6 const Instruction *I = dyn_cast<Instruction>(V);7 8* Change the casting code to allow casting a reference value thus:9 const Value &V = ...;10 Instruction &I = cast<Instruction>(V);11 12 dyn_cast does not work with references, because it must return a null pointer13 on failure.14 15* Fundamentally change how instructions and other values are represented.16 Before, every llvm container was an instance of the ValueHolder template,17 instantiated for each container type. This ValueHolder was effectively a18 wrapper around a vector of pointers to the sub-objects.19 20 Now, instead of having a vector to pointers of objects, the objects are21 maintained in a doubly linked list of values (ie each Instruction now has22 Next & Previous fields). The containers are now instances of ilist (intrusive23 linked list class), which use the next and previous fields to chain them24 together. The advantage of this implementation is that iterators can be25 formed directly from pointers to the LLVM value, and invalidation is much26 easier to handle.27 28* As part of the above change, dereferencing an iterator (for example:29 BasicBlock::iterator) now produces a reference to the underlying type (same30 example: Instruction&) instead of a pointer to the underlying object. This31 makes it much easier to write nested loops that iterator over things, changing32 this:33 34 for (Function::iterator BI = Func->begin(); BI != Func->end(); ++BI)35 for (BasicBlock::iterator II = (*BI)->begin(); II != (*BI)->end(); ++II)36 (*II)->dump();37 38 into:39 40 for (Function::iterator BI = Func->begin(); BI != Func->end(); ++BI)41 for (BasicBlock::iterator II = BI->begin(); II != BI->end(); ++II)42 II->dump();43 44 which is much more natural and what users expect.45 46* Simplification of #include's: Before, it was necessary for a .cpp file to47 include every .h file that it used. Now things are batched a little bit more48 to make it easier to use. Specifically, the include graph now includes these49 edges:50 Module.h -> Function.h, GlobalVariable.h51 Function.h -> BasicBlock.h, Argument.h52 BasicBlock.h -> Instruction.h53 54 Which means that #including Function.h is usually sufficient for getting the55 lower level #includes.56 57* Printing out a Value* has now changed: Printing a Value* will soon print out58 the address of the value instead of the contents of the Value. To print out59 the contents, you must convert it to a reference with (for example)60 'cout << *I' instead of 'cout << I;'. This conversion is not yet complete,61 but will be eventually. In the mean time, both forms print out the contents.62 63* References are used much more throughout the code base. In general, if a64 pointer is known to never be null, it is passed in as a reference instead of a65 pointer. For example, the instruction visitor class uses references instead66 of pointers, and that Pass subclasses now all receive references to Values67 instead of pointers, because they may never be null.68 69* The Function class now has helper functions for accessing the Arguments list.70 Instead of having to go through getArgumentList for simple things like71 iterator over the arguments, now the a*() methods can be used to access them.72 73