HLSL Entry Functions


In HLSL, entry functions denote the starting point for shader execution. They must be known at compile time. For all non-library shaders, the compiler assumes the default entry function name main, unless the DXC /E option is provided to specify an alternate entry point. For library shaders entry points are denoted using the [shader(...)] attribute.

All scalar parameters to entry functions must have semantic annotations, and all struct parameters must have semantic annotations on every field in the struct declaration. Additionally if the entry function has a return type, a semantic annotation must be provided for the return type as well.

HLSL entry functions can be called from other parts of the shader, which has implications on code generation.

Implementation Details

In Clang, the DXC /E option is translated to the cc1 flag -hlsl-entry, which in turn applies the HLSLShader attribute to the function with the specified name. This allows code generation for entry functions to always key off the presence of the HLSLShader attribute, regardless of what shader profile you are compiling.

In code generation, two functions are generated. One is the user defined function, which is code generated as a mangled C++ function with internal linkage following normal function code generation.

The actual exported entry function which can be called by the GPU driver is a void(void) function that isn’t name mangled. In code generation we generate the unmangled entry function to serve as the actual shader entry. The shader entry function is annotated with the hlsl.shader function attribute identifying the entry’s pipeline stage.

The body of the unmangled entry function contains first a call to execute global constructors, then instantiations of the user-defined entry parameters with their semantic values populated, and a call to the user-defined function. After the call instruction the return value (if any) is saved using a target-appropriate intrinsic for storing outputs (for DirectX, the llvm.dx.store.output). Lastly, any present global destructors will be called immediately before the return. HLSL does not support C++ atexit registrations, instead calls to global destructors are compile-time generated.


HLSL support in Clang is currently focused on compute shaders, which do not support output semantics. Support for output semantics will not be implemented until other shader profiles are supported.

Below is example IR that represents the planned implementation, subject to change as the llvm.dx.store.output and llvm.dx.load.input intrinsics are not yet implemented.

; Function Attrs: norecurse
define void @main() #1 {
   %0 = call i32 @llvm.dx.load.input.i32(...)
   %1 = call i32 @"?main@@YAXII@Z"(i32 %0)
   call @llvm.dx.store.output.i32(%1, ...)
   ret void