The extra control-flow seems to inhibit ARC and we end up with an extra
retain-release regressing LevenshteinDistance by 40%.
rdar://22479186
Swift SVN r31588
Revert "Add test cases to exercise the native String vs cocoa buffer String path."
Revert "stdlib: Add back a test I removed"
Revert "stdlib: Fix hasPrefix,hasSuffix tests"
Revert "stdlib: Add documentation for the cached ascii collation tables"
This reverts commit 31493, 31492, 31491, 31490, 31489.
There are linking errors in SwiftExternalProjects (we probably have to link
against libicucore somewhere).
Swift SVN r31543
directly construct a buffer of the right size
This safes a uniqueness check and unecessary code bloat when inlining
reserveCapacity.
rdar://22446738
Swift SVN r31540
Keeping this method in the swift dylib saves a few kilobytes for each
specialization of dictionary. This function is not on the critical
performance path and we don't gain much from pulling it and specializing
it in user code.
Swift SVN r31507
This commit annotates _assertionFailed with an attribute
that will keep in in the swift dylib. We don't care about
the performance of this function because it happens when
the program crashes. Keeping it in the swift dylib helps
in reducing the code size of the user app because it keeps
more of the printing and string.utf8 logic in the dylib.
Swift SVN r31506
Reapply of 31474 with a fix in _compareCocoaBuffer to use the bufferSizeRhs
variable instead of bufferSizeLhs for the right hand side buffer.
We no longer create intermediate NSString copies to compare and hash swift
Strings. Instead we call directly into the ICU library.
I measured a 1.2 to 2x improvement on dictionary benchmarks as a result of this.
The SuperChars benchmark is also about 1.2x faster because of this.
Pure ASCII comparison has gotten a little bit slower (20% on a pure comparison
micro-benchmark) because we no longer do a memcmp. Doing a memcmp on ASCII is
not the same as the default unicode collation. Instead we have to a string scan.
The default unicode collation does not order like ASCII does and ignores
characters (for example the \0 character).
rdar://18992510
Swift SVN r31489
Revert "stdlib: Add back a test I removed"
Revert "Add test cases to exercise the native String vs cocoa buffer String path."
Revert "stdlib: Move the darwin String implementation over to use the ICU library."
This reverts commit r31477, r31476, r31475, r31474.
Commit r31474 broke the ASAN build.
Swift SVN r31488
Allow untyped placeholder to take arbitrary type, but default to Void.
Add _undefined<T>() function, which is like fatalError() but has
arbitrary return type. In playground mode, merely warn about outstanding
placeholders instead of erroring out, and transform placeholders into
calls to _undefined(). This way, code with outstanding placeholders will
only crash when it attempts to evaluate such placeholders.
<rdar://problem/21167372> transform EditorPlaceholderExpr into fatalError()
Swift SVN r31481
We no longer create intermediate NSString copies to compare and hash swift
Strings. Instead we call directly into the ICU library.
I measured a 1.2 to 2x improvement on dictionary benchmarks as a result of this.
The SuperChars benchmark is also about 1.2x faster because of this.
Pure ASCII comparison has gotten a little bit slower (20% on a pure comparison
micro-benchmark) because we no longer do a memcmp. Doing a memcmp on ASCII is
not the same as the default unicode collation. Instead we have to a string scan.
The default unicode collation does not order like ASCII does and ignores
characters (for example the \0 character).
rdar://18992510
Swift SVN r31474
literal initializer to underlying failable initializer through
implicitly-failable initializer (proper solution is covered in
22276505).
<rdar://problem/20788137> Image literals should not be optional
Swift SVN r31229
This allows us to remove the workaround generic overloads for "print" without sacrificing good diagnostics for its unavailable overloads. (rdar://problem/21499048, rdar://problem/21582758, rdar://problem/22126141)
Swift SVN r31169
At -O I see about 10% improvement on Havlak, and 40% improvement on the recently
added array append benchmarks.
COWTree regresses by 15% because we hoist an append that is conditionally
executed. If we were to disallow hoisting of conditionally executed benchmarks
we would see big loses accross our benchmarks (for example, we would not hoist
out of nested loops anymore). So we will have to live with this regression.
rdar://17140639
Swift SVN r31120
