In order for the runtime demangler to be able to find ObjC classes and protocols, it needs to
have the runtime name of the declaration be in the mangled name. Only do this for runtime manglings,
to minimize the potential ABI impact for symbol names that already have the source-level names of
ObjC entities baked in. Fixes SR-12169 | rdar://59306590.
* Don't allocate breadrumbs pointer if under threshold
* Increase breadrumbs threshold
* Linear 16-byte bucketing until 128 bytes, malloc_size after
* Allow cap less than _SmallString.capacity (bridging non-ASCII)
This change decreases the amount of heap usage for moderate-length
strings (< 64 UTF-8 code units in length) and increases the amount of
spare code unit capacity available (less growth needed).
Average improvements for moderate-length strings:
* 64-bit: on average, 8 bytes saved and 4 bytes of extra capacity
* 32-bit: on average, 4 bytes saved and 6 bytes of extra capacity
Additionally, on 32-bit, large-length strings also gain an average of
6 bytes of extra spare capacity.
Details:
On 64-bit, half of moderate-length allocations will save 16 bytes
while the other half get an extra 8 bytes of spare capacity.
On 32-bit, a quarter of moderate-length allocations will save 16
bytes, and the rest get an extra 4 bytes of spare
capacity. Additionally, 32-bit string's storage class now claims its
full allocation, which is its birthright. Prior to this change, we'd
have on average 1.5 bytes of spare capacity, and now we have 7.5 bytes
of spare capacity.
Breadcrumbs threshold is increased from the super-conservative 32 to
the pretty-conservative 64. Some speed improvements are incorporated
in this change, but more are in flight. Even without those eventual
improvements, this is a worthwhile change (ASCII is still fast-pathed
and irrelevant to breadcrumbing).
For a complex real-world workload, this amounts to around a 5%
improvement to transient heap usage due to all strings and a 4%
improvement to peak heap usage due to all strings. For moderate-length
strings specifically, this gives around 11% improvement to both.
Consider following example:
```swift
enum E {
case foo((x: Int, y: Int))
case bar(x: Int, y: Int)
}
func test(e: E) {
if case .foo(let x, let y) = e {}
if case .bar(let tuple) = e {}
}
```
Both of `if case` expressions have to be supported:
1. `case .foo(let x, let y) = e` allows a single tuple
parameter to be "destructured" into multiple arguments.
2. `case .bar(let tuple) = e` allows to match multiple
parameters with a single tuple argument.
Resolves: rdar://problem/60061646
parseClosureSignatureIfPresent.
Otherwise, closure parameters that were parsed as "potentially destructured"
will fail constraint generation, even after the parser has decided
they are not destructured.
This is a follow up to changes related to contextual availability
(https://github.com/apple/swift/pull/29921) which increased score
for unavailable declarations only if they were overloaded but
overlooked a case of a single unavailable choice.
Resolve: rdar://problem/60047439
Since constraint system now handles pattern matching it has
to preverse label matching semantics which existed in original
code: if pattern element has a label it has to match the one
in the tuple type it's matched against.
Resolves: rdar://problem/60048356
Specifically if we have a begin_access [read] or a begin_access [modify] that is
never actually written to. In that case if we can prove that the load [copy] is
completely within the exclusive access region, we will load_borrow. Example:
```
%0 = begin_access [read] ...
%1 = load [copy] %0
...
destroy_value %1
end_access %0
```
=>
```
%0 = begin_access [read] ...
%1 = load_borrow %0
...
end_borrow %1
end_access %0
```
rdar://60064692
Without %target-run, DYLD_LIBRARY_PATH won’t apply and the target will try to load OS-supplied libraries rather than the ones we just built even if use_os_stdlib isn’t enabled.
This was set before config.target_run has fully set up, leading to resilience tests using OS libraries instead of the just-built libs in some cases at least.
Specifically, we do not handle cases where we use projections,
open_existential_addr, or load_borrow. We should be able to handle those in the
future.
The best part is that ossa makes this really easy to do correctly.
rdar://60064817
I enabled this recently by piggybacking it in the mega-commit:
commit badc5658bb
Author: Andrew Trick <atrick@apple.com>
Date: Mon Mar 2 16:23:53 2020
Fix SIL MemBehavior queries with access markers.
It miscompiles SwiftNIO's NIOPerformanceTester benchmark causing it to
segfault immediately.
We may consider debugging this to reenable it in the future...
<rdar://60068448> Teach IRGen to emit !invariant metadata for 'let' variables
Fixes <rdar://60038603> Swift CI: SwiftNIO_macOS NIOHTTP2_macOS fails
The "opened type" of a subscript reference has all references to Self
immediately substituted out, which destroys the link between Self and
the opened existential type we form for the "fully opened type". This
means, in general, we cannot use this type when rewriting subscript
applies, or we'll miss closing opened existentials.
Use the "fully opened type" everywhere except the AnyObject subscript
path. Then, add an assertion that AnyObject subscripts never involve
opened archetypes that we would have to close.
Resolves SR-11748, rdar://57092093
