Add AArch64 PMULL binary-polynomial multiply/square kernels

Port the GF(2)[x] BinPoly kernel family to ARMv8 PMULL. The fixed-size
SMALL cases (1..10 limbs)

Author: Xor-el

CryptoLib.Benchmark/Delphi/CryptoLib.BenchmarkConsole.dpr

@@ -361,6 +361,7 @@ uses
   ClpBinPolySimdMedium in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdMedium.pas',
   ClpBinPolySimdLarge in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdLarge.pas',
   ClpBinPolyX86Backend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyX86Backend.pas',
+  ClpBinPolyArmBackend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyArmBackend.pas',
   ClpBinPolyMulBase in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolyMulBase.pas',
   ClpItohTsujiiInv in '..\..\CryptoLib\src\Math\BinPoly\ClpItohTsujiiInv.pas',
   ClpBinPolys in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolys.pas',

CryptoLib.Examples/Delphi.Examples/CryptoLib.Examples.dpr

@@ -375,6 +375,7 @@ uses
   ClpBinPolySimdMedium in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdMedium.pas',
   ClpBinPolySimdLarge in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdLarge.pas',
   ClpBinPolyX86Backend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyX86Backend.pas',
+  ClpBinPolyArmBackend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyArmBackend.pas',
   ClpBinPolyMulBase in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolyMulBase.pas',
   ClpItohTsujiiInv in '..\..\CryptoLib\src\Math\BinPoly\ClpItohTsujiiInv.pas',
   ClpBinPolys in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolys.pas',

CryptoLib.Tests/Delphi.Tests/CryptoLib.Tests.Mobile.dpr

@@ -356,6 +356,7 @@ uses
   ClpBinPolySimdMedium in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdMedium.pas',
   ClpBinPolySimdLarge in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdLarge.pas',
   ClpBinPolyX86Backend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyX86Backend.pas',
+  ClpBinPolyArmBackend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyArmBackend.pas',
   ClpBinPolyMulBase in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolyMulBase.pas',
   ClpItohTsujiiInv in '..\..\CryptoLib\src\Math\BinPoly\ClpItohTsujiiInv.pas',
   ClpBinPolys in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolys.pas',

CryptoLib.Tests/Delphi.Tests/CryptoLib.Tests.Mobile.dproj

@@ -699,6 +699,7 @@
         <DCCReference Include="..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdMedium.pas"/>
         <DCCReference Include="..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdLarge.pas"/>
         <DCCReference Include="..\..\CryptoLib\src\Simd\Backend\ClpBinPolyX86Backend.pas"/>
+        <DCCReference Include="..\..\CryptoLib\src\Simd\Backend\ClpBinPolyArmBackend.pas"/>
         <DCCReference Include="..\..\CryptoLib\src\Math\BinPoly\ClpBinPolyMulBase.pas"/>
         <DCCReference Include="..\..\CryptoLib\src\Math\BinPoly\ClpItohTsujiiInv.pas"/>
         <DCCReference Include="..\..\CryptoLib\src\Math\BinPoly\ClpBinPolys.pas"/>

CryptoLib.Tests/Delphi.Tests/CryptoLib.Tests.dpr

@@ -375,6 +375,7 @@ uses
   ClpBinPolySimdMedium in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdMedium.pas',
   ClpBinPolySimdLarge in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolySimdLarge.pas',
   ClpBinPolyX86Backend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyX86Backend.pas',
+  ClpBinPolyArmBackend in '..\..\CryptoLib\src\Simd\Backend\ClpBinPolyArmBackend.pas',
   ClpBinPolyMulBase in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolyMulBase.pas',
   ClpItohTsujiiInv in '..\..\CryptoLib\src\Math\BinPoly\ClpItohTsujiiInv.pas',
   ClpBinPolys in '..\..\CryptoLib\src\Math\BinPoly\ClpBinPolys.pas',

CryptoLib.Tests/src/Math/BinPoly/BinPolyTests.pas

@@ -40,6 +40,10 @@ interface
   ClpBinPolyX86Backend,
   ClpBinPolySimd,
 {$ENDIF CRYPTOLIB_X86_SIMD}
+{$IFDEF CRYPTOLIB_AARCH64_ASM}
+  ClpBinPolyArmBackend,
+  ClpBinPolySimd,
+{$ENDIF CRYPTOLIB_AARCH64_ASM}
   ClpBinPolyMulBaseBinomialReduce,
   CryptoLibTestBase;
 
@@ -191,6 +195,20 @@   TTestBinPolyX86 = class(TBinPolyBackendTestBase)
   end;
 {$ENDIF CRYPTOLIB_X86_SIMD}
 
+{$IFDEF CRYPTOLIB_AARCH64_ASM}
+  /// <summary>
+  /// ARM (PMULL) instantiation of the BinPoly backend suite. Registered
+  /// only when CRYPTOLIB_AARCH64_ASM is defined.
+  /// </summary>
+  TTestBinPolyArm = class(TBinPolyBackendTestBase)
+  strict protected
+    function BackendSupported: Boolean; override;
+    function CreateBackendMul(AN: Int32; const AReduce: IBinPolyReduce)
+      : IBinPolyMul; override;
+    function BackendLabel: String; override;
+  end;
+{$ENDIF CRYPTOLIB_AARCH64_ASM}
+
 implementation
 
 const
@@ -1783,6 +1801,29 @@ function TTestBinPolyX86.BackendLabel: String;
 
 {$ENDIF CRYPTOLIB_X86_SIMD}
 
+{$IFDEF CRYPTOLIB_AARCH64_ASM}
+
+{ TTestBinPolyArm }
+
+function TTestBinPolyArm.BackendSupported: Boolean;
+begin
+  Result := TBinPolyArmBackend.IsSupported;
+end;
+
+function TTestBinPolyArm.CreateBackendMul(AN: Int32;
+  const AReduce: IBinPolyReduce): IBinPolyMul;
+begin
+  if not TBinPolySimd.TryCreateBinPolyMul(AN, AReduce, Result) then
+    Result := nil;
+end;
+
+function TTestBinPolyArm.BackendLabel: String;
+begin
+  Result := 'ARM';
+end;
+
+{$ENDIF CRYPTOLIB_AARCH64_ASM}
+
 initialization
 
 {$IFDEF FPC}
@@ -1799,4 +1840,12 @@ initialization
 {$ENDIF FPC}
 {$ENDIF CRYPTOLIB_X86_SIMD}
 
+{$IFDEF CRYPTOLIB_AARCH64_ASM}
+{$IFDEF FPC}
+  RegisterTest(TTestBinPolyArm);
+{$ELSE}
+  RegisterTest(TTestBinPolyArm.Suite);
+{$ENDIF FPC}
+{$ENDIF CRYPTOLIB_AARCH64_ASM}
+
 end.