Jpeg encoding code optimization

src/ImageSharp/Common/Helpers/Numerics.cs

@@ -23,6 +23,28 @@ internal static class Numerics
  private const int ShuffleAlphaControl = 0b_11_11_11_11;
 #endif
 
+#if !SUPPORTS_BITOPERATIONS
+ /// <summary>
+ /// Gets the counts the number of bits needed to hold an integer.
+ /// </summary>
+ private static ReadOnlySpan<byte> BitCountLut => new byte[]
+ {
+ 0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5,
+ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+ 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8,
+ };
+#endif
+
  /// <summary>
  /// Determine the Greatest CommonDivisor (GCD) of two numbers.
  /// </summary>
@@ -756,7 +778,7 @@ public static float Lerp(float value1, float value2, float amount)
  /// widening them to 32-bit integers and performing four additions.
  /// </summary>
  /// <remarks>
- /// <code>byte(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)</code>
+ /// <c>byte(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)</c>
  /// is widened and added onto <paramref name="accumulator"/> as such:
  /// <code>
  /// accumulator += i32(1, 2, 3, 4);
@@ -825,5 +847,26 @@ public static int EvenReduceSum(Vector256<int> accumulator)
  return Sse2.ConvertToInt32(vsum);
  }
 #endif
+
+ /// <summary>
+ /// Calculates how many minimum bits needed to store given value.
+ /// </summary>
+ /// <param name="number">Unsigned integer to store</param>
+ /// <returns>Minimum number of bits needed to store given value</returns>
+ [MethodImpl(MethodImplOptions.AggressiveInlining)]
+ public static int MinimumBitsToStore16(uint number)
+ {
+#if !SUPPORTS_BITOPERATIONS
+ if (number < 0x100)
+ {
+ return BitCountLut[(int)number];
+ }
+
+ return 8 + BitCountLut[(int)number >> 8];
+#else
+ const int bitInUnsignedInteger = sizeof(uint) * 8;
+ return bitInUnsignedInteger - BitOperations.LeadingZeroCount(number);
+#endif
+ }
  }
 }

src/ImageSharp/Common/Helpers/SimdUtils.HwIntrinsics.cs

@@ -532,6 +532,7 @@ private static void Shuffle4Slice3(
  /// <summary>
  /// Performs a multiplication and an addition of the <see cref="Vector256{T}"/>.
  /// </summary>
+ /// <remarks>ret = (vm0 * vm1) + va</remarks>
  /// <param name="va">The vector to add to the intermediate result.</param>
  /// <param name="vm0">The first vector to multiply.</param>
  /// <param name="vm1">The second vector to multiply.</param>
@@ -552,6 +553,30 @@ public static Vector256<float> MultiplyAdd(
  }
  }
 
+ /// <summary>
+ /// Performs a multiplication and a substraction of the <see cref="Vector256{T}"/>.
+ /// </summary>
+ /// <remarks>ret = (vm0 * vm1) - vs</remarks>
+ /// <param name="vs">The vector to substract from the intermediate result.</param>
+ /// <param name="vm0">The first vector to multiply.</param>
+ /// <param name="vm1">The second vector to multiply.</param>
+ /// <returns>The <see cref="Vector256{T}"/>.</returns>
+ [MethodImpl(InliningOptions.ShortMethod)]
+ public static Vector256<float> MultiplySubstract(
+ in Vector256<float> vs,
+ in Vector256<float> vm0,
+ in Vector256<float> vm1)
+ {
+ if (Fma.IsSupported)
+ {
+ return Fma.MultiplySubtract(vm1, vm0, vs);
+ }
+ else
+ {
+ return Avx.Subtract(Avx.Multiply(vm0, vm1), vs);
+ }
+ }
+
  /// <summary>
  /// <see cref="ByteToNormalizedFloat"/> as many elements as possible, slicing them down (keeping the remainder).
  /// </summary>