As there are no instructions left which produce multiple result values, this is a NFC regarding the generated SIL and generated code.
Although this commit is large, most changes are straightforward adoptions to the changes in the ValueBase and SILValue classes.
And use the new project_existential_box to get to the address value.
SILGen now generates a project_existential_box for each alloc_existential_box.
And IRGen re-uses the address value from the alloc_existential_box if the operand of project_existential_box is an alloc_existential_box.
This lets the generated code be the same as before.
This change is needed for the next update to ToT LLVM. It can be put
into place now without breaking anything so I am committing it now.
The churn upstream on ilist_node is neccessary to remove undefined
behavior. Rather than updating the different ilist_node patches for the
hacky change required to not use iterators, just use iterators and keep
everything as ilist_nodes. Upstream they want to eventually do this, so
it makes sense for us to just do it now.
Please do not introduce new invocations of
ilist_node::get{Next,Prev}Node() into the tree.
And use project_box to get to the address value.
SILGen now generates a project_box for each alloc_box.
And IRGen re-uses the address value from the alloc_box if the operand of project_box is an alloc_box.
This lets the generated code be the same as before.
Other than that most changes of this (quite large) commit are straightforward.
its fields are initialized. Before:
t.swift:18:24: error: variable 'self.B' captured by a closure before being initialized
after:
t.swift:19:24: error: 'self' captured by a closure before all members were initialized
self.A.withCString { cString -> () in
^
t.swift:14:7: note: 'self.B' not initialized
var B: Int
^
This drives home the fact that 'self' is being captured here, not the individual
properties.
Partial applications of a root self value are an escape point, not a load. This
improves the diagnostic in this case from:
t.swift:18:24: error: variable 'self.B' used before being initialized
self.A.withCString { cString -> () in
^
to:
t.swift:18:24: error: variable 'self.B' captured by a closure before being initialized
self.A.withCString { cString -> () in
^
When a root class delegates to a non-class-bound protocol method, the self value
gets wrapped up in a SIL alloc_stack so it can be passed by address. Recognize
that the store involved is doing this, so we can provide a more specific diagnostic.
Before this, we produced:
variable 'self.x' used before being initialized
Now we produce:
error: use of 'self' in method call 'getg' before all stored properties are initialized
note: 'self.x' not initialized
This came up with other changes I have to modify the optimizer
pipeline. We shouldn't assert if we have an alloc_stack/dealloc_stack
where the only other use of the alloc_stack is a debug_value_addr.
It's easy to avoid this by removing allocations that don't have real
uses prior to attempting to handle the ones that do have real uses (as
opposed to the other way around).
In a bunch of use-cases we use stripSinglePredecessorArgs to eliminate this
case. There is no reason to assume that this is being done in the caller of
RCIdentity. Lets make sure that we handle this case here.
rdar://24156136
In all of the cases where this is being used, we already immediately perform an
unreachable if we find a TermKind::Invalid. So simplify the code and move it
into the conversion switch itself.
While I'm in this code, generalize it to propagate the original
type information in more cases, including when the original type
is still dependent, and teach it to handle existential metatypes.
rdar://24114020
The code in question was the following:
auto *RetainArray = dyn_cast_or_null<StrongRetainInst>(getInstBefore(Call));
if (!RetainArray && MayHaveBridgedObjectElementType)
return false;
auto *ReleaseArray = dyn_cast_or_null<StrongReleaseInst>(getInstAfter(Call));
if (!ReleaseArray && MayHaveBridgedObjectElementType)
return false;
if (ReleaseArray &&
ReleaseArray->getOperand() != RetainArray->getOperand())
return false;
There is no check in the last if if RetainArray is not nullptr even though it is
possible for it to be so.
Found by clang static analyzer.
This eliminates some minor overheads, but mostly it eliminates
a lot of conceptual complexity due to the overhead basically
appearing outside of its context.
On the whole it looks like this currently benefits performance.
As with the devirtualization pass, once the updated inliner is
committed, the position of this pass in the pipeline will change.
It looks like this has minimal performance impact either way. Once the
changes to make the inliner a function pass are committed, the position
of this in the pipeline will change.
