more cleaning

Author: Daniel Lemire

README.md

@@ -27,12 +27,14 @@ as a reference (`System.Buffers.Text.Base64.DecodeFromUtf8`).
 
 | processor       | SimdBase64(GB/s) | .NET speed (GB/s) | speed up |
 |:----------------|:------------------------|:-------------------|:-------------------|
-| Apple M2 processor (ARM)   | 6.3                      | 3.8               | 1.7 x |
-| Intel Ice Lake (AVX2)   | 5.8                      | 3.4              | 1.7 x |
-| Intel Ice Lake (SSSE3)   | 4.7                      | 3.4              | 1.4 x |
+| Apple M2 processor (ARM)   | 6.5                      | 3.8               | 1.7 x |
+| Intel Ice Lake (AVX2)   | 6.6                      | 3.4              | 1.9 x |
+| Intel Ice Lake (SSSE3)   | 4.9                     | 3.4              | 1.4 x |
 
 Our results are more impressive when comparing against the standard base64 string decoding
-function (`Convert.FromBase64String(mystring)`), but we omit these results for simplicity.
+function (`Convert.FromBase64String(mystring)`), but it is explained in part by the fact
+that the .NET team did not accelerated them using SIMD instructions. Thus we omit them, only
+comparing with the SIMD-accelerated .NET functions.
 
 ## Requirements
 

benchmark/Benchmark.cs

@@ -152,9 +152,9 @@ public Config()
             }
         }
         // Parameters and variables for real data
-        [Params(
-                               //@"data/email/" //,
-                               @"data/dns/swedenzonebase.txt"
+        [Params(
+                               @"data/email/" //,
+                               //@"data/dns/swedenzonebase.txt"
                 )]
 #pragma warning disable CA1051
         public string? FileName;

src/Base64Scalar.cs

@@ -151,7 +151,7 @@ public unsafe static OperationStatus DecodeFromBase64Scalar(ReadOnlySpan<byte> s
                                 if (MatchSystem(Endianness.BIG))
                                 {
                                     triple <<= 8;
-                                    Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 2);
+                                    Buffer.MemoryCopy(&triple, dst, 2, 2);
                                 }
                                 else
                                 {
@@ -180,7 +180,7 @@ public unsafe static OperationStatus DecodeFromBase64Scalar(ReadOnlySpan<byte> s
                         if (MatchSystem(Endianness.BIG))
                         {
                             triple <<= 8;
-                            Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 3);
+                            Buffer.MemoryCopy(&triple, dst, 3, 3);
                         }
                         else
                         {
@@ -300,7 +300,7 @@ public unsafe static OperationStatus DecodeFromBase64Scalar(ReadOnlySpan<char> s
                                 if (MatchSystem(Endianness.BIG))
                                 {
                                     triple <<= 8;
-                                    Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 2);
+                                    Buffer.MemoryCopy(&triple, dst, 2, 2);
                                 }
                                 else
                                 {
@@ -329,7 +329,7 @@ public unsafe static OperationStatus DecodeFromBase64Scalar(ReadOnlySpan<char> s
                         if (MatchSystem(Endianness.BIG))
                         {
                             triple <<= 8;
-                            Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 3);
+                            Buffer.MemoryCopy(&triple, dst, 3, 3);
                         }
                         else
                         {
@@ -457,7 +457,7 @@ public unsafe static OperationStatus SafeDecodeFromBase64Scalar(ReadOnlySpan<byt
                                 if (MatchSystem(Endianness.BIG))
                                 {
                                     triple <<= 8;
-                                    Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 2);
+                                    Buffer.MemoryCopy(&triple, dst, 2, 2);
                                 }
                                 else
                                 {
@@ -492,7 +492,7 @@ public unsafe static OperationStatus SafeDecodeFromBase64Scalar(ReadOnlySpan<byt
                         if (MatchSystem(Endianness.BIG))
                         {
                             triple <<= 8;
-                            Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 3);
+                            Buffer.MemoryCopy(&triple, dst, 3, 3);
                         }
                         else
                         {
@@ -548,7 +548,6 @@ public unsafe static OperationStatus SafeDecodeFromBase64Scalar(ReadOnlySpan<cha
                                 return OperationStatus.DestinationTooSmall;
                             }
                             Buffer.MemoryCopy(bufferPtr, dst, 3, 3);
-                            //Marshal.Copy(buffer, 0, (IntPtr)dst, 3); // optimization opportunity: copy 4 bytes
                             dst += 3;
                             src += 4;
                         }
@@ -621,7 +620,7 @@ public unsafe static OperationStatus SafeDecodeFromBase64Scalar(ReadOnlySpan<cha
                                 if (MatchSystem(Endianness.BIG))
                                 {
                                     triple <<= 8;
-                                    Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 2);
+                                    Buffer.MemoryCopy(&triple, dst, 2, 2);
                                 }
                                 else
                                 {
@@ -656,7 +655,7 @@ public unsafe static OperationStatus SafeDecodeFromBase64Scalar(ReadOnlySpan<cha
                         if (MatchSystem(Endianness.BIG))
                         {
                             triple <<= 8;
-                            Marshal.Copy(BitConverter.GetBytes(triple), 0, (IntPtr)dst, 3);
+                            Buffer.MemoryCopy(&triple, dst, 3, 3);
                         }
                         else
                         {