In some cases, there are differences in how graphics rendering behaves between different graphics APIs. Most of the time the Unity Editor hides the differences, but there are some situations where the Editor cannot do this for you. These situations, and the actions you need to take if they occur, are listed below.
Vertical Texture coordinate conventions differ between two types of platforms: Direct3D-like and OpenGL-like.
This difference tends not to have any effect on your project, other than when rendering into a Render TextureA special type of Texture that is created and updated at runtime. To use them, first create a new Render Texture and designate one of your Cameras to render into it. Then you can use the Render Texture in a Material just like a regular Texture. More info
See in Glossary. When rendering into a Texture on a Direct3D-like platform, Unity internally flips rendering upside down. This makes the conventions match between platforms, with the OpenGL-like platform convention the standard.
Image Effects and rendering in UV space are two common cases in the ShadersA program that runs on the GPU. More info
See in Glossary where you need to take action to ensure that the different coordinate conventions do not create problems in your project.
When you use Image Effects and anti-aliasing, the resulting source Texture for an Image Effect is not flipped to match the OpenGL-like platform convention. In this case, Unity renders to the screen to get anti-aliasing and then resolves rendering into a Render Texture for further processing with an Image Effect.
If your Image Effect is a simple one that processes one Render Texture at a time, Graphics.Blit deals with the inconsistent coordinates. However, if you’re processing more than one Render Texture together in your Image Effect, the Render Textures are likely to come out at different vertical orientations in Direct3D-like platforms and when you use anti-aliasing. To standardize the coordinates, you need to manually “flip” the screen Texture upside down in your Vertex ShaderA program that runs on each vertex of a 3D model when the model is being rendered. More info
See in Glossary so that it matches the OpenGL-like coordinate standard.
The following code sample demonstrates how to do this:
// Flip sampling of the Texture:
// The main Texture
// texel size will have negative Y).
#if UNITY_UV_STARTS_AT_TOP
if (_MainTex_TexelSize.y < 0)
uv.y = 1-uv.y;
#endif
A similar situation occurs with GrabPass. The resulting render Texture might not actually be turned upside down on Direct3D-like (non-OpenGL-like) platforms. If your Shader code samples GrabPass Textures, use the ComputeGrabScreenPos function from the UnityCG include file.
When rendering in Texture coordinate (UV) space for special effects or tools, you might need to adjust your Shaders so that rendering is consistent between Direct3D-like and OpenGL-like systems. You also might need to adjust your rendering between rendering into the screen and rendering into a Texture. Adjust these by flipping the Direct3D-like projection upside down so its coordinates match the OpenGL-like projection coordinates.
The built-in variable ProjectionParams.x contains a +1 or –1 value. -1 indicates a projection has been flipped upside down to match OpenGL-like projection coordinates, while +1 indicates it hasn’t been flipped.
You can check this value in your Shaders and then perform different actions. The example below checks if a projection has been flipped and, if so, flips and then returns the UV coordinates to match.
float4 vert(float2 uv : TEXCOORD0) : SV_POSITION
{
float4 pos;
pos.xy = uv;
// This example is rendering with upside-down flipped projection,
// so flip the vertical UV coordinate too
if (_ProjectionParams.x < 0)
pos.y = 1 - pos.y;
pos.z = 0;
pos.w = 1;
return pos;
}
Similar to Texture coordinates, the clip space coordinates (also known as post-projection space coordinates) differ between Direct3D-like and OpenGL-like platforms:
Direct3D-like: The clip space depth goes from +1.0 at the near plane to 0.0 at the far plane. This applies to Direct3D, Metal and consoles.
OpenGL-like: The clip space depth goes from –1.0 at the near plane to +1.0 at the far plane. This applies to OpenGL and OpenGL ES.
Inside Shader code, you can use the UNITY_NEAR_CLIP_VALUE built-in macro to get the near plane value based on the platform.
Inside script code, use GL.GetGPUProjectionMatrix to convert from Unity’s coordinate system (which follows OpenGL-like conventions) to Direct3D-like coordinates if that is what the platform expects.
To avoid precision issues, make sure that you test your Shaders on the target platforms. The GPUs in mobile devices and PCs differ in how they treat floating point types. PC GPUs treat all floating point types (float, half and fixed) as the same - they do all calculations using full 32-bit precision, while many mobile device GPUs do not do this.
For more information, refer to Use 16-bit precision in shaders.
Some Metal devices don’t support the 16-bit half data type. On these devices, Unity shaders represent half as a 32-bit float. Similarly, Unity shaders represent the real data type as a half on mobile devices and as a float on desktop platforms. This section of the documentation is relevant if you want to use the half data type and want to deploy your application to a device that doesn’t support half.
Using half in the correct places can speed up operations, save memory, and save battery power on mobile devices. It’s best practice to use half for values that don’t need high precision and use the information in this section of the documentation to make sure your code handles devices that don’t support half.
For more information about shader data types and the precision they support, refer to Shader data types and precision.
Use of const differs between Microsoft HLSL (see msdn.microsoft.com) and OpenGL’s GLSL (see Wikipedia) Shader language.
Microsoft’s HLSL const has much the same meaning as it does in C# and C++ in that the variable declared is read-only within its scope but can be initialized in any way.
OpenGL’s GLSL const means that the variable is effectively a compile time constant, and so it must be initialized with compile time constraints (either literal values or calculations on other consts).
It is best to follow the OpenGL’s GLSL semantics and only declare a variable as const when it is truly invariant. Avoid initializing a const variable with some other mutableYou can change the contents of a mutable package. This is the opposite of immutable. Only Local packages and Embedded packages are mutable.
See in Glossary values (for example, as a local variable in a function). This also works in Microsoft’s HLSL, so using const in this way avoids confusing errors on some platforms.
If you use buffers to declare variables in your shader, then set values using the data from a GPU compute buffer or graphics buffer, the data layouts might not match depending on the graphics API. This means you might overwrite data or set variables to the wrong values.
For example if you use cbuffer or Unity’s constant buffer macro, depending on the constant buffer’s data layout and the graphics API, a float3 might become a float4, or a float might become a float2.
You can do the following to make sure all graphics APIs compile a buffer with the same data layout:
float4 and float4x4 instead of float3 and float3x3, because float4 variables are the same size on all graphics APIs, while float3 variables can become a different size on some graphics APIs.float4 then float2 then float, so all graphics APIs structure the data in the same way.For example:
cbuffer myConstantBuffer {
float4x4 matWorld;
float4 vObjectPosition; // Uses a float4 instead of a float3
float arrayIndex;
}
To get Shaders working on all platforms, some Shader values should use these semantics:
Vertex Shader output (clip space) position: SV_POSITION. Sometimes Shaders use POSITION semantics to get Shaders working on all platforms. Note that this does not not work on Sony PS4 or with tessellation.
Fragment Shader output color: SV_Target. Sometimes Shaders use COLOR or COLOR0 to get Shaders working on all platforms. Note that this does not work on Sony PS4.
When rendering Meshes as Points, output PSIZE semantics from the vertex Shader (for example, set it to 1). Some platforms, such as OpenGL ES or Metal, treat point size as “undefined” when it’s not written to from the Shader.
See documentation on Shader semantics for more details.
Direct3D platforms use Microsoft’s HLSL Shader compiler. The HLSL compiler is stricter than other compilers about various subtle Shader errors. For example, it doesn’t accept function output values that aren’t initialized properly.
The most common situations that you might run into using this are:
out parameter. Initialize the output like this: void vert (inout appdata_full v, out Input o)
{
**UNITY_INITIALIZE_OUTPUT(Input,o);**
// ...
}
Partially initialized values. For example, a function returns float4 but the code only sets the .xyz values of it. Set all values or change to float3 if you only need three values.
Using tex2D in the Vertex Shader. This is not valid, because UV derivatives don’t exist in the vertex Shader. You need to sample an explicit mip level instead; for example, use tex2Dlod (tex, float4(uv,0,0)). You also need to add #pragma target 3.0 as tex2Dlod is a Shader model 3.0 feature.
Some parts of the Surface Shader compilation pipeline do not understand DirectX 11-specific HLSL (Microsoft’s shader language) syntax.
If you’re using HLSL features like StructuredBuffers, RWTextures and other non-DirectX 9 syntax, wrap them in a DirectX X11-only preprocessor macro as shown in the example below.
#ifdef SHADER_API_D3D11
// DirectX11-specific code, for example
StructuredBuffer<float4> myColors;
RWTexture2D<float4> myRandomWriteTexture;
#endif
Some GPUs (most notably PowerVR-based ones on iOS) allow you to do a form of programmable blending by providing current fragment color as input to the Fragment Shader (see EXT_shader_framebuffer_fetch on khronos.org).
It is possible to write Shaders in Unity that use the framebuffer fetch functionality. To do this, use the inout color argument when you write a Fragment Shader in either HLSL (Microsoft’s shading language - see msdn.microsoft.com) or Cg (the shading language by Nvidia - see nvidia.co.uk).
The example below is in Cg.
CGPROGRAM
// only compile Shader for platforms that can potentially
// do it (currently gles,gles3,metal)
#pragma only_renderers framebufferfetch
void frag (v2f i, inout half4 ocol : SV_Target)
{
// ocol can be read (current framebuffer color)
// and written into (will change color to that one)
// ...
}
ENDCG
Depth (Z) direction differs on different Shader platforms.
DirectX 11, DirectX 12, Metal: Reversed direction
The depth (Z) buffer is 1.0 at the near plane, decreasing to 0.0 at the far plane.
Clip space range is [near,0] (meaning the near plane distance at the near plane, decreasing to 0.0 at the far plane).
Other platforms: Traditional direction
The depth (Z) buffer value is 0.0 at the near plane and 1.0 at the far plane.
Note that reversed direction depth (Z), combined with a floating point depth bufferA memory store that holds the z-value depth of each pixel in an image, where the z-value is the depth for each rendered pixel from the projection plane. More info
See in Glossary, significantly improves depth buffer precision against the traditional direction. The advantages of this are less conflict for Z coordinates and better shadows, especially when using small near planes and large far planes.
So, when you use Shaders from platforms with the depth (Z) reversed:
_CameraDepth Texture texture range is 1 (near) to 0 (far).However, the following macros and functions automatically work out any differences in depth (Z) directions:
Linear01Depth(float z)LinearEyeDepth(float z)If you are fetching the depth (Z) buffer value manually, you might want to check the buffer direction. The following is an example of this:
float z = tex2D(_CameraDepthTexture, uv);
#if defined(UNITY_REVERSED_Z)
z = 1.0f - z;
#endif
If you are using clip space (Z) depth manually, you might also want to abstract platform differences by using the following macro:
float clipSpaceRange01 = UNITY_Z_0_FAR_FROM_CLIPSPACE(rawClipSpace);
Note: This macro does not alter clip space on OpenGL or OpenGL ES platforms, so it returns within “-near”1 (near) to far (far) on these platforms.
GL.GetGPUProjectionMatrix() returns a z-reverted matrix if you are on a platform where the depth (Z) is reversed. However, if you’re composing from projection matrices manually (for example, for custom shadows or depth rendering), you need to revert depth (Z) direction yourself where it applies via script.
An example of this is below:
var shadowProjection = Matrix4x4.Ortho(...); //shadow camera projection matrix
var shadowViewMat = ... //shadow camera view matrix
var shadowSpaceMatrix = ... //from clip to shadowMap texture space
//'m_shadowCamera.projectionMatrix' is implicitly reversed
//when the engine calculates device projection matrix from the camera projection
m_shadowCamera.projectionMatrix = shadowProjection;
//'shadowProjection' is manually flipped before being concatenated to 'm_shadowMatrix'
//because it is seen as any other matrix to a Shader.
if(SystemInfo.usesReversedZBuffer)
{
shadowProjection[2, 0] = -shadowProjection[2, 0];
shadowProjection[2, 1] = -shadowProjection[2, 1];
shadowProjection[2, 2] = -shadowProjection[2, 2];
shadowProjection[2, 3] = -shadowProjection[2, 3];
}
m_shadowMatrix = shadowSpaceMatrix * shadowProjection * shadowViewMat;
Unity automatically deals with depth (Z) bias to ensure it matches Unity’s depth (Z) direction. However, if you are using a native code rendering plugin, you need to negate (reverse) depth (Z) bias in your C or C++ code.