Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Conditionally enable bounds checking for unsafe buffer pointers #71264

Open
wants to merge 11 commits into
base: main
Choose a base branch
from
Open

Conditionally enable bounds checking for unsafe buffer pointers #71264

wants to merge 11 commits into from

Conversation

DougGregor
Copy link
Member

@DougGregor DougGregor commented Jan 31, 2024 •
edited
Loading

Currently, bounds checking is only enabled for unsafe buffer pointers in debug builds. Introduce a new precondition check kind, _boundsCheckPrecondition, and use it for the bounds checking of unsafe buffer pointers. This check is enabled both in debug builds and in builds where the experimental feature UnsafePointerBoundsSafety is enabled.

Also add [unchecked:] and [uncheckedBounds:] subscripts to Collection and the various array and buffer types, which only perform bounds checking in debug builds. These can be used in places where the optimizer is not eliminating bounds checks that it should be.

@DougGregor
Copy link
Member Author

@swift-ci please smoke test

@DougGregor
Copy link
Member Author

@swift-ci please benchmark

@DougGregor DougGregor force-pushed the unsafe-ptr-bounds-checking branch from 99fd08b to 1a65a4e Compare January 31, 2024 02:11
@DougGregor
Copy link
Member Author

@swift-ci please smoke test

@DougGregor
Copy link
Member Author

@swift-ci please benchmark

@DougGregor
Copy link
Member Author

@swift-ci please smoke test

@DougGregor
Copy link
Member Author

Performance (x86_64): -O

Regression OLD NEW DELTA RATIO
FlattenListFlatMap 2853.0 4297.0 +50.6% 0.66x (?)
AngryPhonebook 275.414 359.116 +30.4% 0.77x (?)
AngryPhonebook.ASCII2.Small 113.833 148.0 +30.0% 0.77x (?)
ArrayAppendGenericStructs 1387.5 1790.0 +29.0% 0.78x (?)
LessSubstringSubstring 24.259 30.594 +26.1% 0.79x (?)
EqualSubstringString 24.364 30.692 +26.0% 0.79x (?)
EqualSubstringSubstring 24.613 30.92 +25.6% 0.80x (?)
UTF8Decode_InitFromData 142.667 177.7 +24.6% 0.80x (?)
EqualStringSubstring 24.815 30.8 +24.1% 0.81x (?)
EqualSubstringSubstringGenericEquatable 24.704 30.594 +23.8% 0.81x (?)
UTF8Decode_InitFromBytes 147.25 178.4 +21.2% 0.83x
String.replaceSubrange.String.Small 38.87 45.821 +17.9% 0.85x (?)
AngryPhonebook.Strasse 538.75 627.333 +16.4% 0.86x (?)
StringWordBuilderReservingCapacity 1140.714 1326.154 +16.3% 0.86x (?)
DistinctClassFieldAccesses 40.378 46.923 +16.2% 0.86x (?)
StringWordBuilder 1177.143 1363.636 +15.8% 0.86x (?)
SubstringEqualString 175.889 203.0 +15.4% 0.87x (?)
StringComparison_longSharedPrefix 210.2 241.222 +14.8% 0.87x (?)
PrefixWhileAnySequence 195.0 222.889 +14.3% 0.87x (?)
PrefixWhileSequence 195.1 222.0 +13.8% 0.88x (?)
String.replaceSubrange.String 9.478 10.783 +13.8% 0.88x (?)
AngryPhonebook.Cyrillic 596.5 678.0 +13.7% 0.88x (?)
String.initRepeating.1AsciiChar.Count100 476.5 541.333 +13.6% 0.88x (?)
AngryPhonebook.Armenian 569.75 645.667 +13.3% 0.88x
StringHasSuffixAscii 3462.0 3916.0 +13.1% 0.88x (?)
FlattenListLoop 1448.0 1624.0 +12.2% 0.89x (?)
Breadcrumbs.UTF16ToIdx.longASCII 39.313 43.582 +10.9% 0.90x (?)
StringBuilder 233.333 256.0 +9.7% 0.91x (?)
String.replaceSubrange.ArrChar.Small 39.333 42.739 +8.7% 0.92x (?)
StringEnumRawValueInitialization 394.815 424.762 +7.6% 0.93x (?)
 
Improvement OLD NEW DELTA RATIO
KeyPathNestedStructs 6.853 3.324 -51.5% 2.06x
KeyPathsSmallStruct 21.438 10.537 -50.8% 2.03x
KeyPathReadPerformance 29.568 18.711 -36.7% 1.58x
Dictionary4 239.333 157.4 -34.2% 1.52x
Dictionary4OfObjects 263.833 189.429 -28.2% 1.39x
Array.removeAll.keepingCapacity.Object 7.376 5.75 -22.0% 1.28x (?)
UTF8Decode_InitDecoding 167.9 139.286 -17.0% 1.21x (?)
UTF8Decode_InitFromCustom_contiguous 167.154 139.385 -16.6% 1.20x (?)
Data.hash.Medium 28.074 24.29 -13.5% 1.16x (?)
UTF8Decode_InitFromCustom_noncontiguous 208.0 182.3 -12.4% 1.14x (?)
NormalizedIterator_emoji 365.231 323.034 -11.6% 1.13x (?)
NormalizedIterator_nonBMPSlowestPrenormal 474.5 430.233 -9.3% 1.10x (?)
ArrayAppendAscii 2043.4 1888.889 -7.6% 1.08x (?)
NormalizedIterator_slowerPrenormal 317.073 295.366 -6.8% 1.07x (?)

Code size: -O

Performance (x86_64): -Osize

Regression OLD NEW DELTA RATIO
Array.removeAll.keepingCapacity.Object 5.577 7.461 +33.8% 0.75x (?)
AngryPhonebook 273.237 360.267 +31.9% 0.76x (?)
AngryPhonebook.ASCII2.Small 113.706 147.786 +30.0% 0.77x (?)
EqualSubstringSubstring 23.844 30.279 +27.0% 0.79x (?)
EqualStringSubstring 23.833 30.259 +27.0% 0.79x (?)
EqualSubstringSubstringGenericEquatable 23.941 30.269 +26.4% 0.79x (?)
LessSubstringSubstringGenericComparable 24.946 31.345 +25.7% 0.80x (?)
EqualSubstringString 24.813 31.135 +25.5% 0.80x (?)
LessSubstringSubstring 24.917 31.258 +25.4% 0.80x (?)
UTF8Decode_InitFromData 142.091 177.727 +25.1% 0.80x (?)
UTF8Decode_InitFromBytes 144.0 178.222 +23.8% 0.81x (?)
StringWordBuilderReservingCapacity 1145.0 1355.385 +18.4% 0.84x (?)
SubstringEqualString 177.364 209.333 +18.0% 0.85x (?)
String.replaceSubrange.String.Small 39.103 46.0 +17.6% 0.85x (?)
StringComparison_longSharedPrefix 205.667 239.3 +16.4% 0.86x (?)
StringWordBuilder 1191.667 1385.0 +16.2% 0.86x (?)
AngryPhonebook.Strasse 539.5 627.0 +16.2% 0.86x (?)
String.replaceSubrange.String 9.588 10.986 +14.6% 0.87x (?)
Breadcrumbs.UTF16ToIdx.longASCII 39.322 44.673 +13.6% 0.88x (?)
AngryPhonebook.Cyrillic 597.333 678.0 +13.5% 0.88x (?)
String.initRepeating.1AsciiChar.Count100 477.25 541.25 +13.4% 0.88x (?)
AngryPhonebook.Armenian 570.25 645.667 +13.2% 0.88x (?)
NormalizedIterator_fastPrenormal 488.958 539.535 +10.3% 0.91x (?)
ArrayInClass 200.694 219.571 +9.4% 0.91x (?)
StringBuilder 235.25 257.0 +9.2% 0.92x (?)
SubstringFromLongString2 23.943 26.129 +9.1% 0.92x (?)
StringBuilderSmallReservingCapacity 242.857 263.5 +8.5% 0.92x (?)
Breadcrumbs.MutatedUTF16ToIdx.ASCII 2.476 2.666 +7.7% 0.93x (?)
 
Improvement OLD NEW DELTA RATIO
KeyPathsSmallStruct 21.479 10.589 -50.7% 2.03x
KeyPathNestedStructs 7.469 3.735 -50.0% 2.00x
KeyPathReadPerformance 29.571 18.738 -36.6% 1.58x
UTF8Decode_InitFromCustom_contiguous 169.0 140.083 -17.1% 1.21x
UTF8Decode_InitDecoding 168.4 141.909 -15.7% 1.19x (?)
Data.hash.Medium 28.074 24.265 -13.6% 1.16x (?)
UTF8Decode_InitFromCustom_noncontiguous 331.0 304.143 -8.1% 1.09x (?)

Code size: -Osize

Performance (x86_64): -Onone

Regression OLD NEW DELTA RATIO
ArrayAppendReserved 2547.143 3226.0 +26.7% 0.79x (?)
EqualStringSubstring 28.542 35.8 +25.4% 0.80x (?)
ObjectiveCBridgeStubDateMutation 348.6 435.75 +25.0% 0.80x (?)
EqualSubstringSubstring 28.72 35.828 +24.7% 0.80x (?)
LessSubstringSubstringGenericComparable 26.577 33.097 +24.5% 0.80x (?)
EqualSubstringSubstringGenericEquatable 27.656 34.4 +24.4% 0.80x (?)
UTF8Decode_InitFromData 143.833 178.6 +24.2% 0.81x
UTF8Decode_InitFromBytes 146.75 181.1 +23.4% 0.81x
EqualSubstringString 29.063 35.815 +23.2% 0.81x (?)
LessSubstringSubstring 28.656 35.207 +22.9% 0.81x (?)
StringWordBuilder 1386.667 1647.273 +18.8% 0.84x
StringWordBuilderReservingCapacity 1350.833 1592.5 +17.9% 0.85x
BinaryFloatingPointPropertiesBinade 53.174 61.857 +16.3% 0.86x (?)
AngryPhonebook.Strasse 540.0 627.333 +16.2% 0.86x (?)
String.replaceSubrange.String.Small 41.148 47.32 +15.0% 0.87x (?)
ObjectiveCBridgeStubDateAccess 3258.0 3726.0 +14.4% 0.87x (?)
PrefixCountableRange 261.714 297.714 +13.8% 0.88x (?)
SuffixCountableRange 90.941 103.0 +13.3% 0.88x (?)
AngryPhonebook.Cyrillic 599.25 678.333 +13.2% 0.88x (?)
AngryPhonebook.Armenian 570.5 645.667 +13.2% 0.88x (?)
DropLastCountableRange 90.895 102.75 +13.0% 0.88x (?)
DropFirstCountableRange 261.714 295.571 +12.9% 0.89x (?)
ArraySetElement 809.5 911.0 +12.5% 0.89x
XorLoop 4489.0 5018.0 +11.8% 0.89x (?)
BinaryFloatingPointPropertiesUlp 70.778 79.0 +11.6% 0.90x (?)
DataSubscriptMedium 56.625 63.158 +11.5% 0.90x (?)
Data.hash.Empty 99.133 109.643 +10.6% 0.90x (?)
ObjectiveCBridgeStubNSDateRefAccess 3369.0 3723.0 +10.5% 0.90x (?)
String.replaceSubrange.String 11.659 12.851 +10.2% 0.91x (?)
BinaryFloatingPointPropertiesNextUp 67.458 74.3 +10.1% 0.91x (?)
DataSubscriptSmall 76.118 82.75 +8.7% 0.92x (?)
ObjectiveCBridgeStubToNSStringRef 111.5 120.923 +8.5% 0.92x (?)
 
Improvement OLD NEW DELTA RATIO
KeyPathsSmallStruct 21.939 10.1 -54.0% 2.17x
SubstringRemoveFirst1 0.222 0.143 -35.4% 1.55x (?)
KeyPathNestedStructs 16.769 13.185 -21.4% 1.27x (?)
ArrayAppendGenericStructs 1390.0 1130.0 -18.7% 1.23x (?)
UTF8Decode_InitDecoding 175.333 148.0 -15.6% 1.18x (?)
UTF8Decode_InitFromCustom_contiguous 177.7 150.417 -15.4% 1.18x (?)
KeyPathReadPerformance 47.167 40.462 -14.2% 1.17x (?)
NSStringConversion.InlineBuffer.ASCII 5884.0 5193.0 -11.7% 1.13x (?)
NSStringConversion.InlineBuffer.UTF8 3574.0 3202.0 -10.4% 1.12x (?)
Data.hash.Medium 32.44 29.091 -10.3% 1.12x (?)

Code size: -swiftlibs

How to read the data The tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.

If you see any unexpected regressions, you should consider fixing the
regressions before you merge the PR.

Noise: Sometimes the performance results (not code size!) contain false
alarms. Unexpected regressions which are marked with '(?)' are probably noise.
If you see regressions which you cannot explain you can try to run the
benchmarks again. If regressions still show up, please consult with the
performance team (@eeckstein).

Hardware Overview 
  Model Name: Mac mini
  Model Identifier: Macmini8,1
  Processor Name: 6-Core Intel Core i7
  Processor Speed: 3.2 GHz
  Number of Processors: 1
  Total Number of Cores: 6
  L2 Cache (per Core): 256 KB
  L3 Cache: 12 MB
  Memory: 32 GB

@glessard
Copy link
Contributor

Nice!

@benrimmington
Copy link
Contributor

Collection has three _failEarlyRangeCheck requirements and default implementations.

  • Should the default implementations be changed, to match the documentation comments?

  • UnsafeRawBufferPointer doesn't implement any of the methods, but it has a TODO comment.

  • UnsafeBufferPointer doesn't implement the method with a ClosedRange parameter, and neither do many other types.

@DougGregor
Copy link
Member Author

DougGregor commented Jan 31, 2024 •
edited
Loading

@benrimmington (EDITED) the default implementations use _precondition, which is the right thing: we're already unconditionally doing bounds checking, so there's nothing to update.

What do you think about providing specific implementations of these for UnsafeRawBufferPointer, @glessard?

As for the ClosedRange version... I don't know why we're missing them, but I'd like to separate out that task from this bounds-safety one.

DougGregor and others added 6 commits January 31, 2024 10:33
@DougGregor
Introduce an assertion configuration for "release with bounds safety"
b9e1b17 
To support the introduction of bounds checks for the unsafe buffer pointer
types, introduce a new assertion confirmation that is for release builds
that also have this buffer safety. It can be enabled with the experimental
feature `UnsafePointerBoundsSafety` in normal builds, or (for testing)
the `ReleaseWithBoundsSafety` assert configuration.

The end goal of this approach to change the experimental feature to
an upcoming feature, which will also enabled with Swift 6.
@glessard @DougGregor
Conditionally enable bounds checking for unsafe buffer pointers
b589273 
Currently, bounds checking is only enabled for unsafe buffer pointers
in debug builds. Introduce a new precondition check kind,
`_boundsCheckPrecondition`, and use it for the bounds checking of
unsafe buffer pointers. This check is enabled both in debug builds
and in builds where the experimental feature `UnsafePointerBoundsSafety`
is enabled.
@DougGregor
Switch UnsafeRawBufferPointer over to _boundsCheckPrecondition
faa3b1a
@DougGregor
Upgrade a few more _debugPreconditions to _boundsCheckPrecondition
6bd528d
@DougGregor
Use _boundsCheckPrecondition for unsafe buffer pointer index manipula…
5c0b803 
…tion
@DougGregor
Add "unchecked" subscripts for buffers and arrays
bf9d50c 
Introduce `subscript(unchecked:)` and `subscript(uncheckedBounds:)`
operations for the various buffer and array types, allowing one to
explicitly opt out of bounds checking for release builds.
@DougGregor DougGregor force-pushed the unsafe-ptr-bounds-checking branch from 1a65a4e to bf9d50c Compare January 31, 2024 18:33
@DougGregor
Copy link
Member Author

@swift-ci please test

@DougGregor
Copy link
Member Author

@swift-ci please benchmark

@DougGregor
Copy link
Member Author

Now updated with comprehensive checking, as well as the subscript(unchecked:) and subscript(uncheckedBounds:) operations.

@ahoppen ahoppen removed their request for review January 31, 2024 23:08
@DougGregor
Copy link
Member Author

From the last benchmark run...

Performance (x86_64): -O

Regression OLD NEW DELTA RATIO
AngryPhonebook 278.091 433.054 +55.7% 0.64x
AngryPhonebook.ASCII2.Small 113.938 175.417 +54.0% 0.65x
SubstringEqualString 165.273 222.667 +34.7% 0.74x
SequenceAlgosAnySequence 19570.0 24822.222 +26.8% 0.79x (?)
DropLastAnySeqCntRangeLazy 31955.0 40516.0 +26.8% 0.79x
DropLastAnySeqCRangeIterLazy 32507.0 41074.0 +26.4% 0.79x
SuffixAnySeqCntRangeLazy 31915.0 40029.0 +25.4% 0.80x
SuffixAnySeqCRangeIterLazy 32191.0 40117.0 +24.6% 0.80x
UTF8Decode_InitDecoding 139.727 172.818 +23.7% 0.81x
UTF8Decode_InitFromCustom_contiguous 140.154 173.077 +23.5% 0.81x (?)
UTF8Decode_InitFromCustom_noncontiguous 180.143 212.0 +17.7% 0.85x
String.replaceSubrange.String.Small 40.741 47.533 +16.7% 0.86x (?)
SubstringEquatable 328.833 382.8 +16.4% 0.86x
Data.hash.Medium 24.3 28.094 +15.6% 0.86x (?)
NormalizedIterator_fastPrenormal 464.231 534.419 +15.1% 0.87x (?)
String.replaceSubrange.Substring 11.0 12.643 +14.9% 0.87x (?)
SuffixAnyCollectionLazy 15769.0 17894.0 +13.5% 0.88x
NormalizedIterator_latin1 168.545 190.783 +13.2% 0.88x (?)
StringWordBuilderReservingCapacity 1205.714 1364.615 +13.2% 0.88x (?)
String.initRepeating.1AsciiChar.Count100 483.0 543.5 +12.5% 0.89x (?)
StringHasPrefixAscii 3703.333 4166.0 +12.5% 0.89x (?)
StringDistance.utf16.mixed 71.333 80.0 +12.1% 0.89x (?)
CStringLongNonAscii 139.429 156.333 +12.1% 0.89x (?)
PrefixAnyCollectionLazy 47199.0 52908.0 +12.1% 0.89x
StringHasSuffixUnicode 72176.471 80388.889 +11.4% 0.90x (?)
StringWordBuilder 1264.286 1406.667 +11.3% 0.90x (?)
DropLastAnyCollectionLazy 15884.0 17642.0 +11.1% 0.90x (?)
DropFirstAnyCollectionLazy 47420.0 52582.0 +10.9% 0.90x (?)
Breadcrumbs.IdxToUTF16Range.longMixed 273.0 302.0 +10.6% 0.90x (?)
StringHasSuffixAscii 3800.0 4200.0 +10.5% 0.90x (?)
AngryPhonebook.Strasse 561.75 620.667 +10.5% 0.91x (?)
AngryPhonebook.Cyrillic 607.75 669.0 +10.1% 0.91x (?)
NormalizedIterator_ascii 93.154 102.286 +9.8% 0.91x (?)
CxxStringConversion.cxx.to.swift 113.5 124.5 +9.7% 0.91x (?)
AngryPhonebook.Armenian 583.5 638.0 +9.3% 0.91x (?)
NormalizedIterator_emoji 311.484 340.296 +9.2% 0.92x (?)
Breadcrumbs.MutatedIdxToUTF16.ASCII 2.5 2.728 +9.1% 0.92x (?)
String.replaceSubrange.Substring.Small 44.885 48.952 +9.1% 0.92x (?)
Breadcrumbs.IdxToUTF16.longMixed 734.5 795.0 +8.2% 0.92x (?)
KeyPathWritePerformance 27.0 29.081 +7.7% 0.93x (?)
DropLastSequenceLazy 296.286 318.833 +7.6% 0.93x (?)
 
Improvement OLD NEW DELTA RATIO
KeyPathNestedStructs 6.844 3.286 -52.0% 2.08x
KeyPathsSmallStruct 21.477 10.543 -50.9% 2.04x
KeyPathReadPerformance 29.568 18.759 -36.6% 1.58x (?)
Dictionary4 191.833 156.4 -18.5% 1.23x
Dictionary4OfObjects 219.333 183.429 -16.4% 1.20x
FlattenListLoop 1624.0 1387.0 -14.6% 1.17x (?)
Breadcrumbs.UTF16ToIdx.longASCII 45.759 39.328 -14.1% 1.16x (?)
StringBuilderSmallReservingCapacity 265.714 239.125 -10.0% 1.11x (?)

Code size: -O

Performance (x86_64): -Osize

Regression OLD NEW DELTA RATIO
AngryPhonebook 273.125 432.297 +58.3% 0.63x
AngryPhonebook.ASCII2.Small 113.625 174.917 +53.9% 0.65x
SubstringEqualString 168.846 213.889 +26.7% 0.79x (?)
SequenceAlgosAnySequence 19580.0 24750.0 +26.4% 0.79x (?)
SuffixAnySeqCntRangeLazy 31745.0 39952.0 +25.9% 0.79x
SuffixAnySeqCRangeIterLazy 31879.0 40035.0 +25.6% 0.80x
DropLastAnySeqCntRangeLazy 32245.0 40122.0 +24.4% 0.80x
DropLastAnySeqCRangeIterLazy 32623.0 40183.0 +23.2% 0.81x
UTF8Decode_InitFromCustom_contiguous 140.615 172.091 +22.4% 0.82x (?)
UTF8Decode_InitDecoding 139.636 170.0 +21.7% 0.82x (?)
Data.hash.Medium 24.292 28.517 +17.4% 0.85x (?)
String.replaceSubrange.Substring 11.111 12.882 +15.9% 0.86x (?)
SubstringEquatable 335.0 385.8 +15.2% 0.87x (?)
String.replaceSubrange.String.Small 40.621 46.769 +15.1% 0.87x (?)
StringHasPrefixAscii 3652.0 4162.0 +14.0% 0.88x (?)
OpenClose 52.261 58.818 +12.5% 0.89x (?)
CStringLongNonAscii 139.615 156.545 +12.1% 0.89x (?)
StringDistance.utf16.mixed 70.5 78.933 +12.0% 0.89x (?)
PrefixAnyCollectionLazy 47927.0 53425.0 +11.5% 0.90x (?)
StringWordBuilderReservingCapacity 1216.667 1345.0 +10.5% 0.90x (?)
SuffixAnyCollectionLazy 15891.0 17550.0 +10.4% 0.91x (?)
StringHasSuffixUnicode 72200.0 79736.842 +10.4% 0.91x (?)
AngryPhonebook.Strasse 560.75 619.25 +10.4% 0.91x (?)
NormalizedIterator_ascii 91.923 101.5 +10.4% 0.91x (?)
StringHasSuffixAscii 3812.0 4204.0 +10.3% 0.91x (?)
RemoveWhereFilterString 220.8 243.4 +10.2% 0.91x (?)
NSStringConversion.Rebridge 50.8 55.974 +10.2% 0.91x (?)
Breadcrumbs.IdxToUTF16Range.longMixed 273.667 301.333 +10.1% 0.91x (?)
String.initRepeating.1AsciiChar.Count100 493.25 540.5 +9.6% 0.91x (?)
StringWordBuilder 1259.286 1379.231 +9.5% 0.91x (?)
StringFromLongWholeSubstring 2.827 3.075 +8.8% 0.92x (?)
StringDistance.utf16.ascii 138.882 151.0 +8.7% 0.92x (?)
RomanNumbers2 455.25 492.75 +8.2% 0.92x (?)
Chars2 3227.273 3493.103 +8.2% 0.92x (?)
ArrayAppendAscii 1876.375 2028.667 +8.1% 0.92x (?)
Breadcrumbs.MutatedIdxToUTF16.ASCII 2.534 2.737 +8.0% 0.93x (?)
DropLastAnyCollectionLazy 16502.0 17763.0 +7.6% 0.93x (?)
 
Improvement OLD NEW DELTA RATIO
KeyPathsSmallStruct 21.511 10.571 -50.9% 2.03x
KeyPathNestedStructs 7.471 3.806 -49.0% 1.96x
KeyPathReadPerformance 29.629 18.762 -36.7% 1.58x
FlattenListFlatMap 3800.0 2620.0 -31.1% 1.45x (?)
FlattenListLoop 1386.0 986.0 -28.9% 1.41x (?)
Breadcrumbs.UTF16ToIdx.longASCII 43.586 39.313 -9.8% 1.11x
StringAdder 309.333 280.0 -9.5% 1.10x (?)
CharacterRecognizer.ascii 102.722 93.647 -8.8% 1.10x (?)
String.replaceSubrange.String 10.782 9.833 -8.8% 1.10x (?)
UTF8Decode_InitFromData_ascii_as_ascii 181.0 166.273 -8.1% 1.09x (?)
Data.hash.Empty 56.63 52.25 -7.7% 1.08x (?)
StringInterpolationManySmallSegments 6652.381 6214.286 -6.6% 1.07x (?)

Code size: -Osize

Performance (x86_64): -Onone

Regression OLD NEW DELTA RATIO
StringWithCString2 0.006 0.01 +57.1% 0.64x
PrefixWhileCountableRange 22762.0 29687.0 +30.4% 0.77x
PrefixWhileAnyCollection 46939.0 59927.0 +27.7% 0.78x
DropFirstAnySeqCRangeIter 32560.0 41425.0 +27.2% 0.79x
SuffixAnyCollection 8090.0 10275.0 +27.0% 0.79x
PrefixAnyCollection 23978.0 30259.0 +26.2% 0.79x
DropLastAnyCollection 7975.0 10064.0 +26.2% 0.79x
SequenceAlgosAnySequence 19922.222 25075.0 +25.9% 0.79x (?)
DropWhileAnyCollection 31794.0 39928.0 +25.6% 0.80x
DropFirstAnyCollection 24083.0 30244.0 +25.6% 0.80x
DropFirstAnySeqCntRange 32697.0 41051.0 +25.5% 0.80x
DropFirstAnySeqCntRangeLazy 32776.0 41012.0 +25.1% 0.80x
DropWhileCountableRange 7931.0 9919.0 +25.1% 0.80x
ObjectiveCBridgeStubDateMutation 348.6 435.75 +25.0% 0.80x
SequenceAlgosRange 1955680.0 2436690.0 +24.6% 0.80x
DropWhileAnySeqCRangeIter 37170.0 46162.0 +24.2% 0.81x
LazilyFilteredRange 849290.0 1050740.0 +23.7% 0.81x
PrefixAnySeqCRangeIter 26370.0 32601.0 +23.6% 0.81x
SuffixAnySeqCntRangeLazy 34838.0 43043.0 +23.6% 0.81x (?)
DropWhileAnySeqCntRange 37157.0 45860.0 +23.4% 0.81x
PrefixAnySeqCntRangeLazy 25931.0 31999.0 +23.4% 0.81x
SuffixAnySeqCntRange 34162.0 42141.0 +23.4% 0.81x
PrefixAnySeqCntRange 26529.0 32689.0 +23.2% 0.81x
SuffixAnySeqCRangeIterLazy 34983.0 43062.0 +23.1% 0.81x
PrefixAnySeqCRangeIterLazy 25926.0 31905.0 +23.1% 0.81x
DropFirstAnySeqCRangeIterLazy 33295.0 40957.0 +23.0% 0.81x
SuffixAnySeqCRangeIter 34860.0 42826.0 +22.9% 0.81x
DropLastAnySeqCntRange 35998.0 44211.0 +22.8% 0.81x
DropWhileCountableRangeLazy 35941.0 44097.0 +22.7% 0.82x
DropWhileAnyCollectionLazy 36321.0 44544.0 +22.6% 0.82x
PrefixWhileCountableRangeLazy 27643.0 33896.0 +22.6% 0.82x
DropWhileAnySeqCRangeIterLazy 36296.0 44420.0 +22.4% 0.82x
DropWhileAnySeqCntRangeLazy 36202.0 44265.0 +22.3% 0.82x
DropLastAnySeqCRangeIterLazy 36758.0 44861.0 +22.0% 0.82x
UTF8Decode_InitDecoding 146.0 178.0 +21.9% 0.82x (?)
DropLastAnySeqCRangeIter 36566.0 44534.0 +21.8% 0.82x
PrefixWhileAnySeqCntRangeLazy 28045.0 34101.0 +21.6% 0.82x
DropLastAnySeqCntRangeLazy 36829.0 44775.0 +21.6% 0.82x
PrefixWhileAnyCollectionLazy 28343.0 34454.0 +21.6% 0.82x
PrefixWhileAnySeqCRangeIterLazy 28241.0 34297.0 +21.4% 0.82x
UTF8Decode_InitFromCustom_contiguous 148.583 180.333 +21.4% 0.82x (?)
DictionaryGroup 4783.0 5740.0 +20.0% 0.83x (?)
BinaryFloatingPointPropertiesUlp 66.545 78.944 +18.6% 0.84x (?)
PrefixWhileAnySeqCntRange 32611.0 38449.0 +17.9% 0.85x
DropLastCountableRangeLazy 14793.0 17378.0 +17.5% 0.85x
ReversedBidirectional 55925.0 65571.0 +17.2% 0.85x
PrefixWhileAnySeqCRangeIter 33532.0 39046.0 +16.4% 0.86x
String.replaceSubrange.String.Small 42.364 49.125 +16.0% 0.86x (?)
String.replaceSubrange.Substring 12.528 14.5 +15.7% 0.86x (?)
BinaryFloatingPointPropertiesBinade 53.136 61.478 +15.7% 0.86x (?)
DropFirstCountableRangeLazy 44953.0 51751.0 +15.1% 0.87x
PrefixCountableRangeLazy 45160.0 51926.0 +15.0% 0.87x
RemoveWhereQuadraticInts 7508.0 8564.0 +14.1% 0.88x (?)
SuffixCountableRangeLazy 15152.0 17264.0 +13.9% 0.88x
StringWordBuilderReservingCapacity 1342.308 1528.333 +13.9% 0.88x (?)
Sim2DArray 5518.0 6280.0 +13.8% 0.88x (?)
CharacterPropertiesPrecomputed 1786.154 2023.636 +13.3% 0.88x (?)
StringHasPrefixAscii 5295.0 5993.333 +13.2% 0.88x
StringWordBuilder 1406.667 1590.909 +13.1% 0.88x (?)
DropLastCountableRange 90.941 102.75 +13.0% 0.89x (?)
SuffixCountableRange 90.737 102.125 +12.6% 0.89x (?)
DropFirstCountableRange 262.143 294.857 +12.5% 0.89x (?)
PrefixCountableRange 261.875 294.429 +12.4% 0.89x (?)
RemoveWhereFilterString 318.714 357.0 +12.0% 0.89x (?)
RemoveWhereQuadraticStrings 8221.0 9207.0 +12.0% 0.89x (?)
SIMDReduce.Int8 6422.0 7180.0 +11.8% 0.89x (?)
StringHasSuffixAscii 5352.5 5976.667 +11.7% 0.90x (?)
DataSubscriptMedium 56.619 63.158 +11.5% 0.90x (?)
StringInterpolationManySmallSegments 7453.333 8309.091 +11.5% 0.90x (?)
BinaryFloatingPointPropertiesNextUp 67.476 75.05 +11.2% 0.90x (?)
DropFirstAnyCollectionLazy 85641.0 94964.0 +10.9% 0.90x (?)
StringHasSuffixUnicode 74400.0 82473.684 +10.9% 0.90x (?)
DataCreateEmpty 807.895 893.0 +10.5% 0.90x (?)
AngryPhonebook.Strasse 561.75 620.0 +10.4% 0.91x (?)
CharacterPropertiesStashedMemo 2064.286 2277.143 +10.3% 0.91x (?)
String.initRepeating.1AsciiChar.Count100 492.5 542.667 +10.2% 0.91x (?)
CStringLongNonAscii 156.615 172.545 +10.2% 0.91x (?)
AngryPhonebook.Cyrillic 607.0 668.333 +10.1% 0.91x (?)
DataCreateMedium 126600.0 139000.0 +9.8% 0.91x (?)
SuffixSequence 10841.0 11886.0 +9.6% 0.91x (?)
DropLastAnyCollectionLazy 28813.0 31565.0 +9.6% 0.91x (?)
AngryPhonebook.Armenian 581.5 637.0 +9.5% 0.91x (?)
PrefixAnyCollectionLazy 86548.0 94567.0 +9.3% 0.92x (?)
SuffixAnyCollectionLazy 28785.0 31443.0 +9.2% 0.92x (?)
SuffixSequenceLazy 10886.0 11841.0 +8.8% 0.92x (?)
String.replaceSubrange.Substring.Small 46.32 50.269 +8.5% 0.92x (?)
DataSubscriptSmall 76.4 82.75 +8.3% 0.92x (?)
Data.hash.Empty 100.2 108.182 +8.0% 0.93x (?)
 
Improvement OLD NEW DELTA RATIO
KeyPathsSmallStruct 21.909 10.09 -53.9% 2.17x
KeyPathNestedStructs 16.897 13.192 -21.9% 1.28x
KeyPathReadPerformance 46.684 40.786 -12.6% 1.14x (?)
CharIndexing_punctuated_unicodeScalars_Backwards 66280.0 59920.0 -9.6% 1.11x (?)
StringEnumRawValueInitialization 7232.0 6743.333 -6.8% 1.07x (?)
Set.isSuperset.Seq.Int.Empty 1289.0 1202.0 -6.7% 1.07x (?)

Code size: -swiftlibs

How to read the data The tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.

If you see any unexpected regressions, you should consider fixing the
regressions before you merge the PR.

Noise: Sometimes the performance results (not code size!) contain false
alarms. Unexpected regressions which are marked with '(?)' are probably noise.
If you see regressions which you cannot explain you can try to run the
benchmarks again. If regressions still show up, please consult with the
performance team (@eeckstein).

Hardware Overview 
  Model Name: Mac mini
  Model Identifier: Macmini8,1
  Processor Name: 6-Core Intel Core i7
  Processor Speed: 3.2 GHz
  Number of Processors: 1
  Total Number of Cores: 6
  L2 Cache (per Core): 256 KB
  L3 Cache: 12 MB
  Memory: 32 GB

@DougGregor
Copy link
Member Author

@swift-ci please smoke test

@eeckstein
Copy link
Contributor

Interestingly, benchmarks like AngryPhonebook don't use any unsafe stuff. It would be good to look at least into one of those benchmark regressions and try to understand the underlying issue

@eeckstein
Copy link
Contributor

Maybe the regression stem from the internals for the String implementation, where it would be okay to opt out of bounds checking (because it's stdlib internal and behind the "safety border")

@stephentyrone
Copy link
Contributor

Yes, that's correct.

@glessard
Copy link
Contributor

glessard commented Feb 1, 2024

What do you think about providing specific implementations of these for UnsafeRawBufferPointer?

I think the reason we hadn't done so thus far is that we got away with it. If we no longer get away with it, then we'll add them.

@DougGregor
Copy link
Member Author

@swift-ci please smoke test

@DougGregor
Add subscript(unchecked:) and subscript(uncheckedBounds:) to Collection
25379e6 
Introduce new requirements for the unchecked subscripting operations into
Collection itself. This makes it possible for generic algorithms to
use the unchecked subscript operations, too, when it is determined that
the optimizer is unable to eliminate specific unnecessary bounds checks.
@DougGregor
Copy link
Member Author

@swift-ci please test

@DougGregor
Copy link
Member Author

@swift-ci please test

@atrick atrick self-requested a review May 2, 2024 19:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ktoso ktoso Awaiting requested review from ktoso ktoso is a code owner

@hborla hborla Awaiting requested review from hborla hborla is a code owner

@xedin xedin Awaiting requested review from xedin xedin is a code owner

@bnbarham bnbarham Awaiting requested review from bnbarham bnbarham is a code owner

@CodaFi CodaFi Awaiting requested review from CodaFi CodaFi is a code owner

@hamishknight hamishknight Awaiting requested review from hamishknight hamishknight is a code owner

@rintaro rintaro Awaiting requested review from rintaro rintaro is a code owner

@slavapestov slavapestov Awaiting requested review from slavapestov slavapestov is a code owner

@tshortli tshortli Awaiting requested review from tshortli tshortli is a code owner

@atrick atrick Awaiting requested review from atrick

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@DougGregor @glessard @benrimmington @eeckstein @stephentyrone