MSL: Add native array test for composite array initialization.

This commit is contained in:
Hans-Kristian Arntzen 2020-02-24 13:34:51 +01:00
parent 30343f3e95
commit d91e134500
3 changed files with 226 additions and 0 deletions

View File

@ -0,0 +1,94 @@
#pragma clang diagnostic ignored "-Wmissing-prototypes"
#include <metal_stdlib>
#include <simd/simd.h>
using namespace metal;
struct Data
{
float a;
float b;
};
constant float X_tmp [[function_constant(0)]];
constant float X = is_function_constant_defined(X_tmp) ? X_tmp : 4.0;
struct Data_1
{
float a;
float b;
};
struct SSBO
{
Data_1 outdata[1];
};
constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(2u, 1u, 1u);
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];
}
}
kernel void main0(device SSBO& _53 [[buffer(0)]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], uint3 gl_LocalInvocationID [[thread_position_in_threadgroup]])
{
Data _25[2] = { Data{ 1.0, 2.0 }, Data{ 3.0, 4.0 } };
Data _31[2] = { Data{ X, 2.0 }, Data{ 3.0, 5.0 } };
Data data2[2];
spvArrayCopyFromStackToStack1(data2, _31);
_53.outdata[gl_WorkGroupID.x].a = _25[gl_LocalInvocationID.x].a + data2[gl_LocalInvocationID.x].a;
_53.outdata[gl_WorkGroupID.x].b = _25[gl_LocalInvocationID.x].b + data2[gl_LocalInvocationID.x].b;
}

View File

@ -0,0 +1,104 @@
#pragma clang diagnostic ignored "-Wmissing-prototypes"
#include <metal_stdlib>
#include <simd/simd.h>
using namespace metal;
struct Data
{
float a;
float b;
};
constant float X_tmp [[function_constant(0)]];
constant float X = is_function_constant_defined(X_tmp) ? X_tmp : 4.0;
struct Data_1
{
float a;
float b;
};
struct SSBO
{
Data_1 outdata[1];
};
constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(2u, 1u, 1u);
constant Data _25[2] = { Data{ 1.0, 2.0 }, Data{ 3.0, 4.0 } };
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];
}
}
static inline __attribute__((always_inline))
Data combine(thread const Data& a, thread const Data& b)
{
return Data{ a.a + b.a, a.b + b.b };
}
kernel void main0(device SSBO& _53 [[buffer(0)]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], uint3 gl_LocalInvocationID [[thread_position_in_threadgroup]])
{
Data data[2] = { Data{ 1.0, 2.0 }, Data{ 3.0, 4.0 } };
Data _31[2] = { Data{ X, 2.0 }, Data{ 3.0, 5.0 } };
Data data2[2];
spvArrayCopyFromStackToStack1(data2, _31);
Data param = data[gl_LocalInvocationID.x];
Data param_1 = data2[gl_LocalInvocationID.x];
Data _73 = combine(param, param_1);
_53.outdata[gl_WorkGroupID.x].a = _73.a;
_53.outdata[gl_WorkGroupID.x].b = _73.b;
}

View File

@ -0,0 +1,28 @@
#version 450
layout(local_size_x = 2) in;
struct Data
{
float a;
float b;
};
layout(std430, binding = 0) buffer SSBO
{
Data outdata[];
};
layout(constant_id = 0) const float X = 4.0;
Data data[2] = Data[](Data(1.0, 2.0), Data(3.0, 4.0));
Data data2[2] = Data[](Data(X, 2.0), Data(3.0, 5.0));
Data combine(Data a, Data b)
{
return Data(a.a + b.a, a.b + b.b);
}
void main()
{
outdata[gl_WorkGroupID.x] = combine(data[gl_LocalInvocationID.x], data2[gl_LocalInvocationID.x]);
}