<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><div class="">Andy Trick and I had this conversation Thursday, and I thought I'd capture it here.</div><div class=""><br class=""></div><div class=""><b class="">The Problem</b></div><div class=""><br class=""></div>It's a longstanding complaint that SIL uses drastically different code patterns for the same sequence of operations based on whether the types of the values involved are loadable or "address-only". For example, suppose I have Swift code like this:<div class=""> let x, y : T</div><div class=""> let z = x + y</div><div class=""><br class=""></div><div class="">If T is a loadable type, this will generate SIL somewhat like this:</div><div class=""> // %x and %y are values of type $T</div><div class="">%lhs = copy_value %x</div><div class="">%rhs = copy_value %y</div><div class=""> %operator = function_ref T.+</div><div class=""> %result = apply %operator(%lhs, %rhs)</div><div class=""> %z = %result</div><div class=""><br class=""></div><div class="">(copy_value doesn't currently exist, but it's easier to understand, and as it happens we're thinking of adding it back.)</div><div class=""><br class=""></div><div class="">If T is an address-only type, this will generate SIL somewhat like this:</div><div class=""> // %x and %y are values of type $*T</div><div class=""> %z = alloc_stack $T</div><div class=""> %lhs = alloc_stack $T</div><div class=""> copy_addr %x to [initialization] %lhs</div><div class=""><div class=""> %rhs = alloc_stack $T</div><div class=""> copy_addr %y to [initialization] %rhs</div></div><div class=""> %operator = function_ref T.+</div><div class=""> apply %operator(%z, %lhs, %rhs)</div><div class=""> dealloc_stack %rhs</div><div class=""> dealloc_stack %lhs</div><div class=""><br class=""></div><div class="">Notably, we're explicitly modeling the memory in which values are stored, which is both very verbose and — more importantly — loses any interesting SSA properties for tracking actual values around. And this has a bunch of secondary effects where various high-level operations like dynamic casts end up needing two different instructions based on whether the value is stored in memory. This is pretty dumb, and it's a major contributor to the reality that generic code is currently very poorly optimized.</div><div class=""><br class=""></div><div class="">It does, however, have some significant advantages: since the memory allocation is explicit, it's quite natural to express optimizations that e.g. hoist or sink those allocations, and the next level of lowering (IRGen) can be very simple.</div><div class=""><br class=""></div><div class=""><b class="">Addresses and address-only types</b></div><div class=""><b class=""><br class=""></b></div><div class="">Swift is an imperative language, and imperative languages make formal memory locations part of the high-level semantics. DI and SSA formation allow us to eliminate formal memory locations for most local variables, but class properties, global variables, escaped local variables, and pointers are all fundamentally un-SSA-able. Therefore we will always have some concept of an address.</div><div class=""><br class=""></div><div class="">But most values of address-only type aren't really being stored in that sort of formal memory location; we're just representing them that way in SIL. Why do we do this? What is it about a type that makes it address-only?</div><div class=""><br class=""></div><div class="">Well, some types are inherently "memory-pinned": something about their representation only makes sense, or is only implementable, if the value is always kept in memory:</div><div class=""> - The representation of the value may involve interior pointers, as with LLVM's SmallVector. This isn't currently a thing in Swift, but it's a possibility someday with the import of non-POD C++ types.</div><div class=""> - The address of the value may need to be registered elsewhere, as with weak references.</div><div class=""> - The value may allow internal mutation even from a shared reference, like a C++ class with a mutable field or a Rust atomic type; you can see weak references as analogous to this.</div><div class="">Such types are necessarily address-only at a low level.</div><div class=""><br class=""></div><div class="">Other types are address-only by abstraction: their representation isn't (fully) known, and so they have to be kept in memory (1) in case that representation includes another address-only value and (2) because there's no way to store a unbounded amount of data except in memory anyway.</div><div class=""><br class=""></div><div class="">But this sense of address-only types that we're describing is really a property of the final generated code. It's necessary for LLVM IR generation to know about it, but it's not obviously necessary for SIL to know about it beyond the implications of the inherently memory-pinned cases above:</div><div class=""> - it is not generally safe to assume that the stored properties of a non-POD C++ type remain invariant across moves</div><div class=""> - weak references *do* represent the same value across moves, but of course that value can change dynamically at any time anyway, per the rules of weak references</div><div class=""> - mutable fields can be modified even by a shared borrowed reference, and so (if they are modeled at all in SIL at all, rather than just leaving the type opaque) there must be some way to project a mutable address from a shared borrow and so on.</div><div class=""><br class=""></div><div class="">Our current representation of address-only types arises directly from the low-level operations that are required, which does admit some interesting optimizations on its own, but the disadvantages of having to support wildly divergent code paths and completely give up SSA use/def chains are crippling. What would SIL look like if we abandoned this entirely?</div><div class=""><br class=""></div><div class=""><b class="">All types as SIL scalars</b></div><div class=""><b class=""><br class=""></b></div><div class="">The fundamental issue here is that IR-level lowering does need to place memory-pinned values into memory. Suppose we take a value, use it as an enum case payload, inject that enum into an Any, and pass that to a function. In a future world where we consistently SSA all local values, this ideally looks something like this:</div><div class=""><br class=""></div><div class=""> // %value is a $T</div><div class=""> %enum = enum #MyEnum.foo, %value : $T</div><div class=""> %any = existential $Any, %enum</div><div class=""> %fn = function_ref @bar</div><div class=""> apply %fn(%any)</div><div class=""><br class=""></div><div class="">If all of these types were loadable, lowering this to IR doesn't impose any abstraction costs, because we can just manipulate it as a bit-pattern in memory, which LLVM (really, any reasonable compiler infrastructure) should be quite good at analyzing and optimizing. But if all of these types are address-only, that's not obviously true. Let's look at the details of lowering to memory.</div><div class=""><br class=""></div><div class="">Because all types are movable — even in a world with Rust-style ownership — there's a fairly straightforward lowering that works for all possible functions: every address-only SIL value gets its own allocation which is deallocated along the dominance frontier of the value. Indirect arguments use the passed-in buffer. Basic block arguments are copied from the allocation for the branch argument value to the allocation for the corresponding destination block parameter. Indirect return values are copied from the allocation for the return value to the pointer passed in.</div><div class=""><br class=""></div><div class="">This admits some obvious optimizations. If "owned" SIL values are pseudo-linear — i.e. along every path from their definition point, it is statically evident that they are (optionally) borrowed multiple times, consumed exactly once, and then never used again — then the copies can instead be moves. Standard register-allocation algorithms can recognize when basic block parameters can use the same location as their arguments. SIL only permits a single return instruction, so there's no reason not to allocate returned values directly into the return slot. These optimizations will tend to eliminate a lot of moves.</div><div class=""><br class=""></div><div class="">However, there's another problem, which is injections. Consider the example above. %any is an opaque existential, and initializing it formally involves allocating a buffer within the existential and moving the argument into that buffer. Ideally, we would allocate that buffer first and then simply allocate %enum in-place into that buffer. In this simple example, that's easy. But if the control flow were more complex, detecting that this is possible becomes significantly more challenging, as does ensuring that the buffer is properly cleaned up along all paths. For example, suppose that %value were computed by calling a throwing function:</div><div class=""><br class=""></div><div class=""> try_apply %someFunction() normal %cont, unwind %handler</div><div class="">cont(%value: $T):</div><div class=""> %enum = enum #MyEnum.foo, %value : $T</div><div class=""> %any = existential $Any, %enum</div><div class=""> %fn = function_ref @bar</div><div class=""> apply %fn(%any)</div><div class="">handler(%error: $Error):</div><div class=""> throw $error</div><div class=""><br class=""></div><div class="">Naive allocation here is going to introduce a lot of moves. Optimally, we would receive the return value from %someFunction directly in the payload of %enum, which we want to build directly into the allocated existential buffer of %any. But to do this, we actually need to allocate that existential buffer before executing the try_apply; and if the try_apply throws, we need to deallocate that existential buffer in the handler block. The need to retroactively insert this kind of clean-up code adds a lot of complexity to this allocation approach. Moreover, it's quite possible that complex intermediate control — for example, if there's a loop somewhere between the definition of a value and its consuming use — will tend to block this kind of analysis and cause more unnecessary moves.</div><div class=""><br class=""></div><div class="">In contrast, the current SIL-generation scheme tends to be aware of at least the immediate local uses of values and therefore emits a lot of these kinds of initialization "in-place" already. But that said, it does proceed from a simple recursive examination of the AST, which means it will miss examples that are just slightly more opaque, like binding a return value as a let and only then returning it. That kind of thing is currently left to the optimizer for no real good reason.</div><div class=""><br class=""></div><div class=""><b class="">Summary</b></div><div class=""><b class=""><br class=""></b></div><div class="">Overall, I think representing local address-only values as SIL scalars with full SSA use/def chains is really promising, and I do think we can write an allocation pass that does an acceptable job eliminating unnecessary moves. In order to actually do this, though, I think we need two things:</div><div class=""> - We need SIL to be "pseudo-linear" as discussed above. We really don't want the allocation pass to have to worry about keeping values alive past consumptions, and thus potentially having to insert copies instead of moves.</div><div class=""> - We need the allocation pass to handle exits before initialization (like with try_apply) and other sorts of interfering control flow. It will not be acceptable for this to be a block-local analysis with only a trivial amount of cross-block argument merging.</div><div class=""><br class=""></div><div class="">John.</div></body></html>