shader-slang
`and` intrinsic does not support vector arguments
Currently, Slang does not support vector types for the and intrinsic:
shader.hlsl(14): error 30019: expected an expression of type 'bool', got 'vector<bool,2>'
n = and(m, l);
^
shader.hlsl(14): error 30019: expected an expression of type 'bool', got 'vector<bool,2>'
n = and(m, l);
^
shader.hlsl(15): error 30019: expected an expression of type 'bool', got 'vector<bool,3>'
q = and(o, p);
^
shader.hlsl(15): error 30019: expected an expression of type 'bool', got 'vector<bool,3>'
q = and(o, p);
^
shader.hlsl(16): error 30019: expected an expression of type 'bool', got 'vector<bool,4>'
t = and(r, s);
^
shader.hlsl(16): error 30019: expected an expression of type 'bool', got 'vector<bool,4>'
t = and(r, s);
^
shader.hlsl(19): error 30019: expected an expression of type 'bool', got 'vector<bool,2>'
n = and(n, true);
^
shader.hlsl20): error 30019: expected an expression of type 'bool', got 'vector<bool,3>'
q = and(q, true);
^
shader.hlsl(21): error 30019: expected an expression of type 'bool', got 'vector<bool,4>'
To ease porting, it would be nice to have support for a vectorized and. It's reasonable to assume that the and intrinsic gets a decent amount of usage in cases where clarity is desired. Note that there is a special case for when the arguments are a vector and a scalar bool. The scalar bool is implicitly expanded to a vector. However, there is not any special cases for arguments of vectors of two different dimensions, so no truncation.
Shader*
float4 main() : SV_Target
{
bool a, b, c;
c = and(a, b);
bool1 i, j, k;
bool2 l, m, n;
bool3 o, p, q;
bool4 r, s, t;
k = and(i, j);
n = and(m, l);
q = and(o, p);
t = and(r, s);
k = and(k, true);
n = and(n, true);
q = and(q, true);
t = and(t, true);
bool2 tv = bool2(and(a, b), and(k.x, and(n.x, and(q.x, t.x))));
return tv.x ? float4(0, 0, 0, 0) : float4(1, 1, 1, 1);
}
https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/and--sm4---asm- (reference for task)
https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/and--sm4---asm- (reference for task)
I don't think this is a good reference link. The document shows the assemble code of how the operation should work not HLSL.
The document shows the assemble code of how the operation should work not HLSL.
There is no actual doc page for and/or. I agree the provided link is not very good.
HLSL2021 touch upon the topic, I suppose the following description is fine for implementation purposes although not complete(?): https://github.com/microsoft/DirectXShaderCompiler/wiki/HLSL-2021
