Since a4e9109 (#17396), both the hashes and the equality of numeric
types inside of AnyHashable do not follow the rules that this part of
the comment was talking about.
I couldn't find an easy example that shows the same behaviour, so I
decided to remove the comment completely.
LLVM r355981 changed various intrinsic functions, including expect,
to require immediate arguments. Swift's _branchHint function has an
expected value that is passed in as an argument, so that it cannot
use LLVM's expect intrinsic. The good news is that _branchHint is only
ever used with immediate arguments, so we can just move the intrinsic
into _fastPath and _slowPath and use those instead of _branchHint.
As was noted in the documentation, the _fastPath and _slowPath names are
confusing but we have passed the point where we can simply rename them.
We could add new names but would still need to keep the old ones around
for binary compatibility, and it is not clear that it is worth the
trouble. I have removed that note from the documentation.
This avoids us having to pattern match every source file which should
help speed up the CMake generation. A secondary optimization is
possible with CMake 3.14 which has the ability to remove the last
extension component without having to resort to regular expressions. It
also helps easily identify the GYB'ed sources.
1. Move discussion of `DBL_EPSILON` etc. onto `ulpOfOne` instead of `ulp`.
2. Add text explaining that `ulpOfOne` is a poor value to use for approximate comparison.
This adds a mostly flow-insensitive analysis that runs before the
dominator-based transformations. The analysis is simple and efficient
because it only needs to track data flow of currently in-scope
accesses. The original dominator tree walk remains simple, but it now
checks the flow insensitive analysis information to determine general
correctness. This is now correct in the presence of all kinds of nested
static and dynamic nested accesses, call sites, coroutines, etc.
This is a better compromise than:
(a) disabling the pass and taking a major performance loss.
(b) converting the pass itself to full-fledged data flow driven
optimization, which would be more optimal because it could remove
accesses when nesting is involved, but would be much more expensive
and complicated, and there's no indication that it's useful.
The new approach is also simpler than adding more complexity to
independently handle to each of many issues:
- Nested reads followed by a modify without a false conflict.
- Reads nested within a function call without a false conflict.
- Conflicts nested within a function call without dropping enforcement.
- Accesses within a generalized accessor.
- Conservative treatment of invalid storage locations.
- Conservative treatment of unknown apply callee.
- General analysis invalidation.
Some of these issues also needed to be considered in the
LoopDominatingAccess sub-pass. Rather than fix that sub-pass, I just
integrated it into the main pass. This is a simplification, is more
efficient, and also handles nested loops without creating more
redundant accesses. It is also generalized to:
- hoist non-uniquely identified accesses.
- Avoid unnecessarily promoting accesses inside the loop.
With this approach we can remove the scary warnings and caveats in the
comments.
While doing this I also took the opportunity to eliminate quadratic
behavior, make the domtree walk non-recursive, and eliminate cutoff
thresholds.
Note that simple nested dynamic reads to identical storage could very
easily be removed via separate logic, but it does not fit with the
dominator-based algorithm. For example, during the analysis phase, we
could simply mark the "fully nested" read scopes, then convert them to
[static] right after the analysis, removing them from the result
map. I didn't do this because I don't know if it happens in practice.
Windows uses `char *` for the `va_list` type even on x86_64. Restore
the correct selection of `__VaListBuilder` for Windows x86_64. This
partially fixes variadic functions.
We've been collecting the location info for some time now, but
apparently never printed it in no-assert builds of the stdlib, which
means this functionality was never available to the users.
With this change, the location will be printed depending on the
debug/release build configuration of the program, not stdlib.
Somewhat addresses: <rdar://problem/42980523>
- Add an underscore to private/internal symbols
- Make access levels explicit, even when they’re implied from context
- Conform to max line length of 80 characters
- Conformances declared on separate extensions that implement them
- Arrange member declarations so that stored properties appear first
- Expand single-letter function and type parameter names where there is an obvious name that’s more descriptive
- RangeReplaceableCollection.fastApplicationEnumeration → CollectionDifference._fastEnumeratedApply
String.Index has an encodedOffset-based initializer and computed
property that exists for serialization purposes. It was documented as
UTF-16 in the SE proposal introducing it, which was String's
underlying encoding at the time, but the dream of String even then was
to abstract away whatever encoding happend to be used.
Serialization needs an explicit encoding for serialized indices to
make sense: the offsets need to align with the view. With String
utilizing UTF-8 encoding for native contents in Swift 5, serialization
isn't necessarily the most efficient in UTF-16.
Furthermore, the majority of usage of encodedOffset in the wild is
buggy and operates under the assumption that a UTF-16 code unit was a
Swift Character, which isn't even valid if the String is known to be
all-ASCII (because CR-LF).
This change introduces a pair of semantics-preserving alternatives to
encodedOffset that explicitly call out the UTF-16 assumption. These
serve as a gentle off-ramp for current mis-uses of encodedOffset.
Compilation of code sample snippets was broken due to using the same
identifier as the `type` function itself to store its result resulting
in:
```swift
func printGenericInfo<T>(_ value: T) {
let type = type(of: value)
print("'\(value)' of type '\(type)'")
}
// error: repl.swift:2:16: error: variable used within its own initial
// value
// let type = type(of: value)
// ^
```
Result:
- Snippets are more copy&paste friendly.
- Resolves https://bugs.swift.org/browse/SR-9915
This fixes a major perform bug involving array initialization from any
contiguously stored collection. This is not a recent regression. This fix
results in a 10,000X speedup (that's 4 zeros) for this code path:
func initializeFromSlice(_ a: [Int]) -> [Int] {
return Array<Int>(a[...])
}
A benchmark is included.
This fix updates various initializers to handle incoming empty buffers
that happen to have a nil base. They should simply create another
buffer with nil base rather than crashing!
It is valid for an Unsafe[Raw]BufferPointer can have a nil base
address. This allows round-tripping with C code that takes a
pointer/length pair and uses `0` as the pointer value.
The original design wrongly assumed that we would use a sentinel value
for empty buffers and was never updated for or tested with the current
design.
Fixes <rdar://problem/47946984> Regression in Foundation.Data's
UnsafeBufferPointer constructor.
