Heterogeneous shaders with multiple arguments

[Checkmate50]

When a __GPU_FOREACH is executed, the resulting function is expected to have exactly three arguments: renderer, gridDims, and buffer. In real programs, this last argument could realistically be several arguments, each corresponding to a separate input argument to the target shader. As a result, the following compute shader header does not have the correct heterogeneous header produced:

void computeMain(uniform RWStructuredBuffer<float> ioBuffer, 
uniform RWStructuredBuffer<float> ioBuffer2, uint3 dispatchThreadID : SV_DispatchThreadID) {...}
...
__GPU_FOREACH(renderer, uint3(4, 1, 1), LAMBDA(uint3 dispatchThreadID) 
{ computeMain(convertBuffer(structuredBuffer), convertBuffer(structuredBuffer2), dispatchThreadID) ; });

In particular, the above header and loop produces the following code:

void computeMain_wrapper(gfx_Renderer_0* renderer, Vector<uint32_t, 3> gridDims, 
	RWStructuredBuffer<float> buffer) {...}
...
computeMain_wrapper(_S12, _S15, _S16, _S17);

Clearly, the generated wrapper is incorrect, while the call to this wrapper is correct (or at least has the correct number of arguments).