BizHawk/EMU7800/Core/BufferElement.cs

72 lines
2.5 KiB
C#

namespace EMU7800.Core
{
/*
* why this sucks:
* A read costs 3 shifts and an or. A write costs 2 shifts. Additional shifts are
* needed elsewhere to figure out which item in a BufferElement[] to access. Because
* the encapsulation is for a BufferElement and not a whole array of them, code elsewhere
* is gunked up with 'BufferElement.SIZE' shifts. If the 32 bit "alias" was actually used,
* there might be some purpose to this code: but it's only used for a ZeroMemory()
* replacement. Every use of BufferElement in the code is a BufferElement[] used as a gunked
* up replacement for a byte[].
*
* A small speed increase was observed hacking this out; but my motivation was more about cleaness
* and stomping out bad ideas.
*/
/*
/// <summary>
/// Frames are composed of <see cref="BufferElement"/>s,
/// that group bytes into machine words for efficient array processing.
/// Bytes are packed in logical little endian order.
/// </summary>
public struct BufferElement
{
/// <summary>
/// The number of bytes contained within a <see cref="BufferElement"/>.
/// </summary>
public const int SIZE = 4; // 2^SHIFT
/// <summary>
/// The mask value applied against a byte array index to access the individual bytes within a <see cref="BufferElement"/>.
/// </summary>
public const int MASK = 3; // SIZE - 1
/// <summary>
/// The left shift value to convert a byte array index to a <see cref="BufferElement"/> array index.
/// </summary>
public const int SHIFT = 2;
uint _data;
/// <summary>
/// A convenience accessor for reading/writing individual bytes within this <see cref="BufferElement"/>.
/// </summary>
/// <param name="offset"></param>
public byte this[int offset]
{
get
{
var i = (offset & MASK) << 3;
return (byte)(_data >> i);
}
set
{
var i = (offset & MASK) << 3;
var d = (uint)value << i;
var di = (uint)0xff << i;
_data = _data & ~di | d;
}
}
/// <summary>
/// Zeros out all bytes of this <see cref="BufferElement"/>.
/// </summary>
public void ClearAll()
{
_data = 0;
}
}
*/
}