We cannot use spare bits or other overlapping storage layout tricks with fundamentally
address-only enums, and we can take advantage of this to do borrowing switches or other
in-place projections without copying the value. However, for resilient enums, the
implementation may use spare bit packing, but the type must be handled address-only
outside of its defining module, and we didn't have a way to express that with
borrowing switch. Optimization passes have also been running into problems with the
complexity that we were using `unchecked_take_enum_data_addr` sometimes as a pure
operation. This patch splits the instruction into three:
- `unchecked_inplace_enum_data_addr` represents a nondestructive in-place enum
projection. It is only allowed for enums whose projection operation is
nondestructive.
- `unchecked_take_enum_data_addr` represents a destructive enum projection,
invalidating the enum and leaving the payload to be further consumed.
This matches the current instruction's semantics.
- `unchecked_borrow_enum_data_addr` represents a borrowing enum projection.
The instruction takes a second operand for "scratch" space, which the
enum representation may be copied into in order to avoid invalidating the
enum value, so the result is dependent on the lifetime of both the
original enum and the scratch buffer. This allows for borrowing switches
over resilient enums.
`unchecked_borrow_enum_data_addr` is implemented by taking advantage of the
"address-only enums can't do spare bit optimization" property at runtime.
We inspect the operand type's bitwise-borrowability from its metadata. If
the type is bitwise-borrowable, then we are allowed to bitwise-copy the
enum to the scratch space and apply the projection to the scratch space,
preserving the original value. If the type is not bitwise-borrowable, then
we cannot use spare bit optimization in its layout, so we apply the
projection in-place.
Fixes rdar://174952822.
The `@export(interface)` and `@export(implementation)` attributes
SE-0497 are queried directly on AST nodes in several places within the
SIL pipeline. However, they don't persist when SIL functions are
serialized, meaning that clients of the original module might make
different assumptions about the availability of a given function's
definition.
Represent these attributes in a SIL function (as an optional
CodeGenerationModel), (de-)serialize them into the module, and add a
textual representation as SIL function attributes `[export_interface]`
and `[export_implementation]`.
Conformance entries are used for fast conformance lookup, which doesn't need to query the runtime's conformance lookup table.
A conformance entry specifies if the class conforms or does not conform to a protocol.
At runtime, a type cast instruction to an existential can directly load the witness table pointer from the VTable.
If null, the class does not conform to the protocol.
These two new invariants eliminate corner cases which caused bugs if optimization didn't handle them.
Also, it will significantly simplify lifetime completion.
The implementation basically consists of these changes:
* add a flag in SILFunction which tells optimization if they need to take care of infinite loops
* add a utility to break infinite loops
* let all optimizations remove unreachable blocks and break infinite loops if necessary
* add verification to check the new SIL invariants
The new `breakIfniniteLoops` utility breaks infinite loops in the control flow by inserting an "artificial" loop exit to a new dead-end block with an `unreachable`.
It inserts a `cond_br` with a `builtin "infinite_loop_true_condition"`:
```
bb0:
br bb1
bb1:
br bb1 // back-end branch
```
->
```
bb0:
br bb1
bb1:
%1 = builtin "infinite_loop_true_condition"() // always true, but the compiler doesn't know
cond_br %1, bb2, bb3
bb2: // new back-end block
br bb1
bb3: // new dead-end block
unreachable
```
The previous algorithm was doing an iterative forward data flow analysis
followed by a reverse data flow analysis. I suspect the history here is that
it was a reverse analysis, and that didn't really work for infinite loops,
and so complexity accumulated.
The new algorithm is quite straightforward and relies on the allocations
being properly jointly post-dominated, just not nested. We simply walk
forward through the blocks in consistent-with-dominance order, maintaining
the stack of active allocations and deferring deallocations that are
improperly nested until we deallocate the allocations above it. The only
real subtlety is that we have to delay walking into dead-end regions until
we've seen all of the edges into them, so that we can know whether we have
a coherent stack state in them. If the state is incoherent, we need to
remove any deallocations of previous allocations because we cannot talk
correctly about what's on top of the stack.
The reason I'm doing this, besides it just being a simpler and hopefully
faster algorithm, is that modeling some of the uses of the async stack
allocator properly requires builtins that cannot just be semantically
reordered. That should be somewhat easier to handle with the new approach,
although really (1) we should not have runtime functions that need this and
(2) we're going to need a conservatively-correct solution that's different
from this anyway because hoisting allocations is *also* limited in its own
way.
I've attached a rather pedantic proof of the correctness of the algorithm.
The thing that concerns me most about the rewritten pass is that it isn't
actually validating joint post-dominance on input, so if you give it bad
input, it might be a little mystifying to debug the verifier failures.
This instruction converts Builtin.ImplicitActor to Optional<any Actor>. In the
process of doing so, it masks out the bits we may have stolen from the witness
table pointer of Builtin.ImplicitActor. The bits that we mask out are the bottom
two bits of the top nibble of the TBI space on platforms that support TBI (that
is bit 60,61 on arm64). On platforms that do not support TBI, we just use the
bottom two tagged pointer bits (0,1).
By using an instruction, we avoid having to represent the bitmasking that we are
performing at the SIL level and can instead just make the emission of the
bitmasking an IRGen detail. It also allows us to move detection if we are
compiling for AArch64 to be an IRGen flag instead of a LangOpts flag.
The instruction is a guaranteed forwarding instruction since we want to treat
its result as a borrowed projection from the Builtin.ImplicitActor.
The asmname attribute allows one to specify the name that will be used
when lowering a given SIL declaration to LLVM IR. It is not currently
exposed in the surface language.
Make sure this attribute round-trips through the parser and
serialization.
Part of rdar://137014448O.
Based on what I see from OwnershipModelEliminator,
`%x = load [copy] %y` can turn into a plain load
followed by a retain of the loaded value.
See `NonTrivialLoadableTypeLowering::emitLoad`
When extending a coroutine, handle the end_borrow instruction used to end a
coroutine lifetime at a dead-end block.
Fixes rdar://153479358 (Compiler crash when force-unwrapping optional ~Copyable type)
It derives the address of the first element of a vector, i.e. a `Builtin.FixedArray`, from the address of the vector itself.
Addresses of other vector elements can then be derived with `index_addr`.
When performing a dynamic cast to an existential type that satisfies
(Metatype)Sendable, it is unsafe to allow isolated conformances of any
kind to satisfy protocol requirements for the existential. Identify
these cases and mark the corresponding cast instructions with a new flag,
`[prohibit_isolated_conformances]` that will be used to indicate to the
runtime that isolated conformances need to be rejected.
The problem with `is_escaping_closure` was that it didn't consume its operand and therefore reference count checks were unreliable.
For example, copy-propagation could break it.
As this instruction was always used together with an immediately following `destroy_value` of the closure, it makes sense to combine both into a `destroy_not_escaped_closure`.
It
1. checks the reference count and returns true if it is 1
2. consumes and destroys the operand
This is used for synthetic uses like _ = x that do not act as a true use but
instead only suppress unused variable warnings. This patch just adds the
instruction.
Eventually, we can use it to move the unused variable warning from Sema to SIL
slimmming the type checker down a little bit... but for now I am using it so
that other diagnostic passes can have a SIL instruction (with SIL location) so
that we can emit diagnostics on code like _ = x. Today we just do not emit
anything at all for that case so a diagnostic SIL pass would not see any
instruction that it could emit a diagnostic upon. In the next patch of this
series, I am going to add SILGen support to do that.
With a focus on updating the documentation of Ownership SSA.
The main changes are:
* Created a new directory `docs/SIL` and moved all SIL-related files into this directory.
* Converted `rst` files to markdown.
* Extracted sections from `SIL.md` which go into very much detail - including the instruction reference - into separate files: `Types.md`, `Ownership.md`, `FunctionAttributes.md`, `Instructions.md`. Those files are referenced from `SIL.md` at the relevant places.
* Rewrote and updated the OSSA part in `SIL.md`
* Removed a few sections, which are not relevant anymore, like "Value dependency" (which is replaced by ownership concepts).
* Fixed and improved a lot of small things.
Emil Pedersen
Joe Groff
Doug Gregor
Erik Eckstein
Anthony Latsis
Jamie
John McCall
Meghana Gupta
Michael Gottesman
Kavon Farvardin
Janat Baig
Nate Chandler
Andrew Trick
Alex Martini