Skip to content

Commit

Permalink
Add WGSL derivative and miscellaneous intrinsics
Browse files Browse the repository at this point in the history 
This addresses issue shader-slang#5083.
  • Loading branch information
@aleino-nv
aleino-nv committed Sep 18, 2024
1 parent 51ff5b8 commit b1d6819
Show file tree
Hide file tree
Showing 2 changed files with 46 additions and 17 deletions.
58 changes: 41 additions & 17 deletions source/slang/hlsl.meta.slang
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5710,7 +5710,7 @@ bool all(T x)

__generic<T : __BuiltinType, let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]
bool all(vector<T,N> x)
{
if(N == 1)
Expand Down Expand Up @@ -5747,6 +5747,8 @@ bool all(vector<T,N> x)
OpAll $$bool result %castResult
};
}
case wgsl:
__intrinsic_asm "all";
default:
bool result = true;
for(int i = 0; i < N; ++i)
Expand Down Expand Up @@ -5849,7 +5851,7 @@ bool any(T x)

__generic<T : __BuiltinType, let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]
bool any(vector<T, N> x)
{
if(N == 1)
Expand Down Expand Up @@ -5886,6 +5888,8 @@ bool any(vector<T, N> x)
OpAny $$bool result %castResult
};
}
case wgsl:
__intrinsic_asm "any";
default:
bool result = false;
for(int i = 0; i < N; ++i)
Expand Down Expand Up @@ -7198,7 +7202,7 @@ vector<T,N> cospi(vector<T,N> x)
// Population count
[__readNone]
[ForceInline]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
uint countbits(uint value)
{
__target_switch
Expand All @@ -7214,13 +7218,15 @@ uint countbits(uint value)
__intrinsic_asm "$P_countbits($0)";
case spirv:
return spirv_asm {OpBitCount $$uint result $value};
case wgsl:
__intrinsic_asm "countOneBits";
}
}

__generic <let N : int>
[__readNone]
[ForceInline]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
vector<uint, N> countbits(vector<uint, N> value)
{
__target_switch
Expand All @@ -7233,6 +7239,8 @@ vector<uint, N> countbits(vector<uint, N> value)
__intrinsic_asm "popcount";
case spirv:
return spirv_asm {OpBitCount $$vector<uint, N> result $value};
case wgsl:
__intrinsic_asm "countOneBits";
default:
VECTOR_MAP_UNARY(uint, N, countbits, value);
}
Expand Down Expand Up @@ -7305,7 +7313,7 @@ for (auto xOrY : diffDimensions) {
}}}}
__generic<T : __BuiltinFloatingPointType>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, fragmentprocessing)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]
T dd$(xOrY)(T x)
{
__requireComputeDerivative();
Expand All @@ -7321,12 +7329,14 @@ T dd$(xOrY)(T x)
__intrinsic_asm "dfd$(xOrY)";
case spirv:
return spirv_asm {OpDPd$(xOrY) $$T result $x};
case wgsl:
__intrinsic_asm "dpd$(xOrY)";
}
}

__generic<T : __BuiltinFloatingPointType, let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, fragmentprocessing)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]
vector<T, N> dd$(xOrY)(vector<T, N> x)
{
__requireComputeDerivative();
Expand All @@ -7342,12 +7352,14 @@ vector<T, N> dd$(xOrY)(vector<T, N> x)
__intrinsic_asm "dfd$(xOrY)";
case spirv:
return spirv_asm {OpDPd$(xOrY) $$vector<T, N> result $x};
case wgsl:
__intrinsic_asm "dpd$(xOrY)";
}
}

__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, fragmentprocessing)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]
matrix<T, N, M> dd$(xOrY)(matrix<T, N, M> x)
{
__requireComputeDerivative();
Expand Down Expand Up @@ -8236,7 +8248,7 @@ vector<T,N> faceforward(vector<T,N> n, vector<T,N> i, vector<T,N> ng)

// Find first set bit starting at high bit and working down
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
int firstbithigh(int value)
{
__target_switch
Expand All @@ -8249,12 +8261,13 @@ int firstbithigh(int value)
case spirv: return spirv_asm {
OpExtInst $$int result glsl450 FindSMsb $value
};
case wgsl: __intrinsic_asm "firstLeadingBit";
}
}

__generic<let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
vector<int, N> firstbithigh(vector<int, N> value)
{
__target_switch
Expand All @@ -8265,13 +8278,14 @@ vector<int, N> firstbithigh(vector<int, N> value)
case spirv: return spirv_asm {
OpExtInst $$vector<int, N> result glsl450 FindSMsb $value
};
case wgsl: __intrinsic_asm "firstLeadingBit";
default:
VECTOR_MAP_UNARY(int, N, firstbithigh, value);
}
}

[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
uint firstbithigh(uint value)
{
__target_switch
Expand All @@ -8284,12 +8298,13 @@ uint firstbithigh(uint value)
case spirv: return spirv_asm {
OpExtInst $$uint result glsl450 FindUMsb $value
};
case wgsl: __intrinsic_asm "firstLeadingBit";
}
}

__generic<let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
vector<uint,N> firstbithigh(vector<uint,N> value)
{
__target_switch
Expand All @@ -8300,14 +8315,15 @@ vector<uint,N> firstbithigh(vector<uint,N> value)
case spirv: return spirv_asm {
OpExtInst $$vector<uint,N> result glsl450 FindUMsb $value
};
case wgsl: __intrinsic_asm "firstLeadingBit";
default:
VECTOR_MAP_UNARY(uint, N, firstbithigh, value);
}
}

// Find first set bit starting at low bit and working up
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
int firstbitlow(int value)
{
__target_switch
Expand All @@ -8320,12 +8336,13 @@ int firstbitlow(int value)
case spirv: return spirv_asm {
OpExtInst $$int result glsl450 FindILsb $value
};
case wgsl: __intrinsic_asm "firstTrailingBit";
}
}

__generic<let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
vector<int,N> firstbitlow(vector<int,N> value)
{
__target_switch
Expand All @@ -8336,13 +8353,14 @@ vector<int,N> firstbitlow(vector<int,N> value)
case spirv: return spirv_asm {
OpExtInst $$vector<int,N> result glsl450 FindILsb $value
};
case wgsl: __intrinsic_asm "firstTrailingBit";
default:
VECTOR_MAP_UNARY(int, N, firstbitlow, value);
}
}

[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
uint firstbitlow(uint value)
{
__target_switch
Expand All @@ -8355,12 +8373,13 @@ uint firstbitlow(uint value)
case spirv: return spirv_asm {
OpExtInst $$uint result glsl450 FindILsb $value
};
case wgsl: __intrinsic_asm "firstTrailingBit";
}
}

__generic<let N : int>
[__readNone]
[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]
[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]
vector<uint,N> firstbitlow(vector<uint,N> value)
{
__target_switch
Expand All @@ -8371,6 +8390,7 @@ vector<uint,N> firstbitlow(vector<uint,N> value)
case spirv: return spirv_asm {
OpExtInst $$vector<uint,N> result glsl450 FindILsb $value
};
case wgsl: __intrinsic_asm "firstTrailingBit";
default:
VECTOR_MAP_UNARY(uint, N, firstbitlow, value);
}
Expand Down Expand Up @@ -8722,7 +8742,7 @@ matrix<T, N, M> frexp(matrix<T, N, M> x, out matrix<int, N, M, L> exp)
// Texture filter width
__generic<T : __BuiltinFloatingPointType>
[__readNone]
[require(glsl_hlsl_metal_spirv, fragmentprocessing)]
[require(glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]
T fwidth(T x)
{
__requireComputeDerivative();
Expand All @@ -8739,12 +8759,14 @@ T fwidth(T x)
{
OpFwidth $$T result $x;
};
case wgsl:
__intrinsic_asm "fwidth($0)";
}
}

__generic<T : __BuiltinFloatingPointType, let N : int>
[__readNone]
[require(glsl_hlsl_spirv, fragmentprocessing)]
[require(glsl_hlsl_spirv_wgsl, fragmentprocessing)]
vector<T, N> fwidth(vector<T, N> x)
{
__requireComputeDerivative();
Expand All @@ -8759,6 +8781,8 @@ vector<T, N> fwidth(vector<T, N> x)
{
OpFwidth $$vector<T, N> result $x;
};
case wgsl:
__intrinsic_asm "fwidth($0)";
}
}

Expand Down
5 changes: 5 additions & 0 deletions source/slang/slang-capabilities.capdef
View file
Original file line number Diff line number Diff line change
Expand Up @@ -346,6 +346,10 @@ alias cuda_hlsl_spirv = cuda | hlsl | spirv;
/// [Compound]
alias glsl_hlsl_spirv = glsl | hlsl | spirv;

/// GLSL, HLSL, SPIRV and WGSL code-gen targets
/// [Compound]
alias glsl_hlsl_spirv_wgsl = glsl | hlsl | spirv | wgsl;

/// GLSL, HLSL, Metal, and SPIRV code-gen targets
/// [Compound]
alias glsl_hlsl_metal_spirv = glsl | hlsl | metal | spirv;
Expand Down Expand Up @@ -1709,6 +1713,7 @@ alias fragmentprocessing = fragment + _sm_5_0
| fragment + metal
| fragment + cpp
| fragment + cuda
| fragment + wgsl
;
/// Capabilities required to use fragment derivative operations (with GLSL derivativecontrol)
/// [Compound]
Expand Down

0 comments on commit b1d6819

Please sign in to comment.