TSMPDecodeByteOutput.cginc

TSMPDecodeByteOutput.cginc is a shader include for codec decode passes. It turns a byte-addressed DecodeByte(index) function into a fragment shader that writes decoded bytes into an RGBA output texture.

Use it when your custom codec can recover one byte by index and you want Core's decoder readback path to receive those bytes in a standard texture layout.

Include path:

#include "Packages/com.kibalab.tsmp.core/Runtime/Codecs/Common/Shaders/cgincs/TSMPDecodeByteOutput.cginc"

Required inputs

The include expects these symbols to exist in the shader before it is included or before the fragment runs.

Symbol	Type	Purpose
v2f	struct	Fragment input type. It must expose float2 uv.
_OutputWidth	numeric uniform	Width of the byte output texture in pixels.
_OutputHeight	numeric uniform	Height of the byte output texture in pixels.
_ByteCount	numeric uniform	Number of valid decoded bytes requested by the decoder.
DecodeByte(int byteIndex)	function	Returns one decoded byte for a byte index.

_OutputWidth, _OutputHeight, and _ByteCount are usually assigned by the decoder or codec material setup.

DecodeByte(int byteIndex) contract

By default the include calls:

int DecodeByte(int byteIndex)

The function should return an integer byte value from 0 to 255.

Rules:

• byteIndex is a zero-based payload byte index.
• The function may be called for indices beyond _ByteCount inside the last RGBA pixel.
• Out-of-range reads should return 0.
• Values outside 0..255 should be clamped or avoided by the codec.
• The function should not rely on filtered sampling for exact byte values.

Custom byte function name

Define TSMP_DECODE_BYTE_FUNC before including the file to use a different function name.

int MyCodecDecodeByte(int byteIndex)
{
    return DecodeMySymbol(byteIndex);
}

#define TSMP_DECODE_BYTE_FUNC MyCodecDecodeByte
#include "Packages/com.kibalab.tsmp.core/Runtime/Codecs/Common/Shaders/cgincs/TSMPDecodeByteOutput.cginc"

The replacement function must have the same call shape: one int index in, one byte-like int out.

Output layout

Each output pixel stores four bytes:

Channel	Byte index
R	baseByte + 0
G	baseByte + 1
B	baseByte + 2
A	baseByte + 3

The base index is calculated as:

baseByte = ((int)pixel.y * (int)_OutputWidth + (int)pixel.x) * 4;

The returned fragment color is:

float4(b0 / 255.0, b1 / 255.0, b2 / 255.0, b3 / 255.0)

If baseByte >= _ByteCount, the fragment returns black transparent zero.

TSMPDecodeByteOutputFragment(v2f i)

float4 TSMPDecodeByteOutputFragment(v2f i)

Main helper function exported by the include. It:

1. Converts i.uv into an output pixel coordinate.
2. Clamps the pixel coordinate to the output texture bounds.
3. Calculates the first byte index for that pixel.
4. Calls TSMP_DECODE_BYTE_FUNC for up to four bytes.
5. Packs those bytes into RGBA channels.

Use this function when your shader already has its own frag entry point.

Default frag(v2f i)

Unless suppressed, the include defines:

float4 frag(v2f i) : SV_Target
{
    return TSMPDecodeByteOutputFragment(i);
}

This is convenient for a simple decode material where the pass only writes the byte output texture.

TSMP_SUPPRESS_DEFAULT_FRAG

Define TSMP_SUPPRESS_DEFAULT_FRAG before including the file if you want to provide your own fragment function.

#define TSMP_SUPPRESS_DEFAULT_FRAG
#include "Packages/com.kibalab.tsmp.core/Runtime/Codecs/Common/Shaders/cgincs/TSMPDecodeByteOutput.cginc"

float4 frag(v2f i) : SV_Target
{
    return TSMPDecodeByteOutputFragment(i);
}

This is useful when the pass needs extra debug output, conditional logic, or a different entry point name.

Minimal shader pass example

struct v2f
{
    float4 pos : SV_POSITION;
    float2 uv : TEXCOORD0;
};

float _OutputWidth;
float _OutputHeight;
float _ByteCount;

int DecodeByte(int byteIndex)
{
    return byteIndex & 255;
}

#include "Packages/com.kibalab.tsmp.core/Runtime/Codecs/Common/Shaders/cgincs/TSMPDecodeByteOutput.cginc"

This example writes a predictable byte pattern into the output texture. A real codec should decode from the TSMP source frame instead.

Common mistakes

Symptom	Likely cause
Output is all zero	_ByteCount is zero or DecodeByte always returns zero.
Last bytes contain garbage	DecodeByte does not guard indices beyond payload length.
Every fourth byte is wrong	RGBA channel order does not match the output layout.
Works in editor but fails in runtime	Shader property names or material setup differ in the runtime path.
Stale bytes after a smaller frame	The output texture is not cleared and the decoder reads more than _ByteCount.