175 lines
3.7 KiB
Plaintext
175 lines
3.7 KiB
Plaintext
|
#pragma clang diagnostic ignored "-Wmissing-prototypes"
|
||
|
#pragma clang diagnostic ignored "-Wmissing-braces"
|
||
|
|
||
|
#include <metal_stdlib>
|
||
|
#include <simd/simd.h>
|
||
|
|
||
|
using namespace metal;
|
||
|
|
||
|
template<typename T, size_t Num>
|
||
|
struct spvUnsafeArray
|
||
|
{
|
||
|
T elements[Num ? Num : 1];
|
||
|
|
||
|
thread T& operator [] (size_t pos) thread
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
constexpr const thread T& operator [] (size_t pos) const thread
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
|
||
|
device T& operator [] (size_t pos) device
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
constexpr const device T& operator [] (size_t pos) const device
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
|
||
|
constexpr const constant T& operator [] (size_t pos) const constant
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
|
||
|
threadgroup T& operator [] (size_t pos) threadgroup
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
constexpr const threadgroup T& operator [] (size_t pos) const threadgroup
|
||
|
{
|
||
|
return elements[pos];
|
||
|
}
|
||
|
};
|
||
|
|
||
|
struct Block
|
||
|
{
|
||
|
uint2 _m0[2];
|
||
|
uint2 _m1[2];
|
||
|
};
|
||
|
|
||
|
struct SSBO
|
||
|
{
|
||
|
Block _m0[3];
|
||
|
};
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromConstantToStack1(thread T (&dst)[A], constant T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromConstantToThreadGroup1(threadgroup T (&dst)[A], constant T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromStackToStack1(thread T (&dst)[A], thread const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromStackToThreadGroup1(threadgroup T (&dst)[A], thread const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromThreadGroupToStack1(thread T (&dst)[A], threadgroup const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromThreadGroupToThreadGroup1(threadgroup T (&dst)[A], threadgroup const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromDeviceToDevice1(device T (&dst)[A], device const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromConstantToDevice1(device T (&dst)[A], constant T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromStackToDevice1(device T (&dst)[A], thread const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromThreadGroupToDevice1(device T (&dst)[A], threadgroup const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromDeviceToStack1(thread T (&dst)[A], device const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
template<typename T, uint A>
|
||
|
inline void spvArrayCopyFromDeviceToThreadGroup1(threadgroup T (&dst)[A], device const T (&src)[A])
|
||
|
{
|
||
|
for (uint i = 0; i < A; i++)
|
||
|
{
|
||
|
dst[i] = src[i];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
kernel void main0(device SSBO& ssbo [[buffer(0)]])
|
||
|
{
|
||
|
threadgroup uint2 _18[2];
|
||
|
spvUnsafeArray<uint2, 2> _27;
|
||
|
spvArrayCopyFromDeviceToStack1(_27.elements, ssbo._m0[0u]._m1);
|
||
|
spvArrayCopyFromStackToDevice1(ssbo._m0[0u]._m0, _27.elements);
|
||
|
spvArrayCopyFromStackToDevice1(ssbo._m0[0u]._m0, _27.elements);
|
||
|
}
|
||
|
|