In the previous post, we discussed escape analysis as implemented in the Java Swift compiler and discussed in the corresponding WRL research report. Specifically we looked at references that can escape from the current method or thread. This helps us do optimizations such as allocating the objects on the stack or on the heap and the removal of unnecessary synchronization. We found that the Swift compiler uses the summary of the method together with the inter procedural analysis. It finds the usages of the references to determine if the references are being held globally or used in an array that can let it escape. With this analysis, Swift looks for values which are either objects that are directly allocated in the method or are newly allocated unaliased objects returned by method calls. It makes a conservative selection and comes up with a pruned list. For methods that don't require synchronization, it adds the corresponding operations to the SSA graph.
We review Alias analysis next. Here Swift attempts to relax data dependencies in the SSA graph. By alias, we mean that two memory locations are not the same  when they are accessed. For example, if a loop contains two array store operations then there is a phi node in the Swift IR at the top of the loop and this merges the initial global store entering the loop with that at the end of each iteration. The values of the variables loaded inside of the loop cannot move outside the loop. Therefore they are reloaded in each iteration.