Out of SILGen, we'll get the non-indirect SSA for throwing
the error. AddressLowering then converts a `throw` into
`throw_addr` to match the function convention. Similarly, a
try_apply gets rewritten to pass the error address to the
error successor block.
resolves rdar://158171053
This is done by splitting the `begin_borrow` of the whole struct into individual borrows of the fields (for trivial fields no borrow is needed).
And then sinking the `struct` to it's consuming use(s).
```
%3 = struct $S(%nonTrivialField, %trivialField) // owned
...
%4 = begin_borrow %3
%5 = struct_extract %4, #S.nonTrivialField
%6 = struct_extract %4, #S.trivialField
use %5, %6
end_borrow %4
...
end_of_lifetime %3
```
->
```
...
%5 = begin_borrow %nonTrivialField
use %5, %trivialField
end_borrow %5
...
%3 = struct $S(%nonTrivialField, %trivialField)
end_of_lifetime %3
```
This optimization is important for Array code where the Array buffer is constantly wrapped into structs and then extracted again to access the buffer.
A dead `destructure_struct` with an owned argument can appear for a non-copyable or non-escapable struct which has only trivial elements.
The instruction is not trivially dead because it ends the lifetime of its operand.
Fixes an ownership verification error.
In 025902fa99,
`SILFunction::getSourceFile()`
is introduced. This patch makes use of that to avoid duplicating the
logic in `static SourceFile &getSourceFile(SILFunction *f)` in
lib/SILOptimizer/Differentiation/ADContext.cpp.
Beside supporting OSSA, this change significantly simplifies the pass.
The main change is that instead of starting at a closure (e.g. `partial_apply`) and finding all call sites, we now start at a call site and look for closures for all arguments. This makes a lot of things much simpler, e.g. not so many intermediate data structures are required to track all the states.
I needed to remove the 3 unit tests because the things those tests were testing are not there anymore. However, the pass is tested with a lot of sil tests (and I added quite a few), which should give good test coverage.
The old ClosureSpecializer pass is still kept in place, because at that point in the pipeline we don't have OSSA, yet. Once we have that, we can replace the old pass withe the new one.
However, the autodiff closure specializer already runs in the OSSA pipeline and there the new changes take effect.
https://github.com/swiftlang/swift/pull/67920 introduced a mandatory SIL
optimization that eliminates switch statement dispatch for enum elements that
are provably unavailable at runtime. This optimization helps eliminate
unreachable code, but it's also load bearing because it removes references to
enum element symbols that may not preset at link time. The optimization was
checking the wrong condition to determine whether an enum element is
unavailable at runtime and so the optimization wasn't working for code using
custom availability domains.
Resolves rdar://161846311.
Unowned result conventions do not work well with OSSA. Retain the result
right after the call when we come out of OSSA so we can treat the
returned value as if it was owned when we do optimizations.
This fix a miscompilation due to the DestroyAddrHoisting pass hoisting
destroys above copies with unowned sources. When the destroyed object
was the last reference to the pointed memory the copy is happening too
late resulting in a use after free.
rdar://160462854
This was a wrong use of the InstructionDeleter.
When replacing a mark_dependence with a new one the old one has to be deleted _without_ fixing its lifetime.
Otherwise destroy_value instructions are inserted, which is obviously wrong.
Fixes a SIL ownership verification error.
rdar://161831308
The intent for `@inline(always)` is to act as an optimization control.
The user can rely on inlining to happen or the compiler will emit an error
message.
Because function values can be dynamic (closures, protocol/class lookup)
this guarantee can only be upheld for direct function references.
In cases where the optimizer can resolve dynamic function values the
attribute shall be respected.
rdar://148608854
No update is needed for the values they produce. This pass should
really be refactored not to crash on instructions that aren't explicitly
listed or at least not to compile if not every instruction is listed.
rdar://161371112
Was mistakenly counting a 'store' as
a copying instruction, when it's only a
consuming one.
SILGen was not handling lvalues that are
addresses for loadable types correctly,
when emitting a CopyExpr in ManualOwnership.
