Rewrite UnsafeRawBufferPointer.Iterator.next

[meg-gupta]

In the SIL generated from the current implementation, it isn't clear to the optimizer that _position == nil && _end != nil can never happen. This ends up creating a non-natural loop and none of the SIL and LLVM loop passes work for such loops. We have to find a way to fix this in SIL. Until then, rewrite so we get a natural loop.

Note: UnsafeBufferPointer.Iterator.next does not have this issue

[glessard]

@swift-ci please test

[meg-gupta]

@swift-ci test

[meg-gupta]

@swift-ci Please Apple Silicon benchmark

[meg-gupta]

------- Performance (arm64): -O -------

REGRESSION                                OLD       NEW        DELTA    RATIO    
ArrayAppendGenericStructs                 947.5     1053.636   +11.2%   **0.90x (?)**
SIMDRandomMask.Int8x16                    308.6     340.167    +10.2%   **0.91x (?)**
Diffing.Same                              2.577     2.84       +10.2%   **0.91x (?)**

IMPROVEMENT                               OLD       NEW        DELTA    RATIO    
UTF8Decode_InitDecoding                   131.25    96.238     -26.7%   **1.36x**
UTF8Decode_InitFromData                   132.067   96.95      -26.6%   **1.36x**
UTF8Decode_InitFromCustom_contiguous      131.706   97.957     -25.6%   **1.34x**
UTF8Decode_InitFromBytes                  137.8     103.526    -24.9%   **1.33x**
UTF8Decode_InitFromCustom_noncontiguous   151.0     116.368    -22.9%   **1.30x**
EqualSubstringString                      21.021    17.752     -15.6%   **1.18x**
EqualStringSubstring                      20.452    17.325     -15.3%   **1.18x**
EqualSubstringSubstring                   20.826    17.75      -14.8%   **1.17x**
LessSubstringSubstringGenericComparable   20.8      17.752     -14.7%   **1.17x**
LessSubstringSubstring                    20.78     17.752     -14.6%   **1.17x**
EqualSubstringSubstringGenericEquatable   20.281    17.752     -12.5%   **1.14x**
ObserverUnappliedMethod                   283.333   249.767    -11.8%   **1.13x (?)**

------- Code size: -O -------


####### Testing optimization level -Osize #######

------- Performance (arm64): -Osize -------

REGRESSION                                OLD       NEW       DELTA    RATIO    
Diffing.Same                              2.538     2.846     +12.1%   **0.89x**
FlattenListFlatMap                        2465.0    2678.0    +8.6%    **0.92x (?)**

IMPROVEMENT                               OLD       NEW       DELTA    RATIO    
UTF8Decode_InitDecoding                   132.938   96.286    -27.6%   **1.38x**
UTF8Decode_InitFromData                   131.941   96.619    -26.8%   **1.37x**
UTF8Decode_InitFromCustom_contiguous      131.706   96.545    -26.7%   **1.36x**
UTF8Decode_InitFromBytes                  137.313   101.9     -25.8%   **1.35x**
UTF8Decode_InitFromCustom_noncontiguous   227.6     190.083   -16.5%   **1.20x**
EqualStringSubstring                      20.391    17.41     -14.6%   **1.17x**
EqualSubstringString                      20.427    17.481    -14.4%   **1.17x**
EqualSubstringSubstringGenericEquatable   20.604    17.639    -14.4%   **1.17x**
EqualSubstringSubstring                   20.5      17.635    -14.0%   **1.16x**
LessSubstringSubstring                    20.396    17.619    -13.6%   **1.16x (?)**
LessSubstringSubstringGenericComparable   20.521    18.053    -12.0%   **1.14x**

------- Code size: -Osize -------

------- Performance (arm64): -Onone -------

REGRESSION                                OLD        NEW        DELTA    RATIO      
RawBufferCopyBytes                        10.015     12.455     +24.4%   **0.80x (?)**
Diffing.Same                              2.519      2.826      +12.2%   **0.89x (?)**
RawBuffer.1000.findLast                   44984.0    50271.0    +11.8%   **0.89x**  
RawBuffer.128.findLast                    6272.0     6943.0     +10.7%   **0.90x (?)**
RawBuffer.39.findLast                     2381.0     2571.0     +8.0%    **0.93x (?)**

IMPROVEMENT                               OLD        NEW        DELTA    RATIO      
ObjectiveCBridgeASCIIStringFromFile       0.12       0.0        -99.2%   **121.00x (?)**
UTF8Decode_InitFromData                   132.333    97.211     -26.5%   **1.36x**  
UTF8Decode_InitDecoding                   134.714    100.0      -25.8%   **1.35x**  
UTF8Decode_InitFromBytes                  138.063    102.526    -25.7%   **1.35x**  
UTF8Decode_InitFromCustom_contiguous      135.353    100.8      -25.5%   **1.34x**  
EqualSubstringSubstringGenericEquatable   22.892     19.54      -14.6%   **1.17x**  
LessSubstringSubstringGenericComparable   22.775     19.539     -14.2%   **1.17x**  
EqualSubstringString                      24.35      20.891     -14.2%   **1.17x**  
EqualStringSubstring                      24.048     20.723     -13.8%   **1.16x**  
LessSubstringSubstring                    23.592     20.413     -13.5%   **1.16x**  
EqualSubstringSubstring                   23.013     20.261     -12.0%   **1.14x**  
ArrayAppendGenericStructs                 1199.091   1080.0     -9.9%    **1.11x (?)**
DataAppendBytesMedium                     1714.615   1563.846   -8.8%    **1.10x (?)**
ArrayAppendToGeneric                      201.667    184.706    -8.4%    **1.09x (?)**

[glessard]

Nice!