MSL: Fix copy of arrays to/from stage IO variables.

Need to take into account effective storage classes and whether or not
we target stage IO blocks since native arrays are conditionally enabled.

4 files changed, 541 insertions, 0 deletions

diff --git a/reference/opt/shaders-msl/masking/copy-arrays.mask-location-0.msl2.multi-patch.tesc b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-0.msl2.multi-patch.tesc
new file mode 100644
index 00000000..4f9134ee
--- /dev/null
+++ b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-0.msl2.multi-patch.tesc
@@ -0,0 +1,188 @@
+#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 main0_out
+{
+    float4 gl_Position;
+};
+
+struct main0_patchOut
+{
+    spvUnsafeArray<float4, 2> pFoo;
+};
+
+struct main0_in
+{
+    spvUnsafeArray<float4, 2> iFoo;
+    float4 ipFoo;
+};
+
+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(uint3 gl_GlobalInvocationID [[thread_position_in_grid]], device main0_out* spvOut [[buffer(28)]], constant uint* spvIndirectParams [[buffer(29)]], device main0_patchOut* spvPatchOut [[buffer(27)]], device MTLQuadTessellationFactorsHalf* spvTessLevel [[buffer(26)]], device main0_in* spvIn [[buffer(22)]])
+{
+    device main0_out* gl_out = &spvOut[gl_GlobalInvocationID.x - gl_GlobalInvocationID.x % 4];
+    threadgroup float4 spvStorageFoo[8][4][2];
+    threadgroup float4 (&Foo)[4][2] = spvStorageFoo[(gl_GlobalInvocationID.x / 4) % 8];
+    device main0_patchOut& patchOut = spvPatchOut[gl_GlobalInvocationID.x / 4];
+    device main0_in* gl_in = &spvIn[min(gl_GlobalInvocationID.x / 4, spvIndirectParams[1] - 1) * spvIndirectParams[0]];
+    uint gl_InvocationID = gl_GlobalInvocationID.x % 4;
+    uint gl_PrimitiveID = min(gl_GlobalInvocationID.x / 4, spvIndirectParams[1]);
+    gl_out[gl_InvocationID].gl_Position = float4(1.0);
+    spvArrayCopyFromDeviceToThreadGroup1(Foo[gl_InvocationID], gl_in[gl_InvocationID].iFoo.elements);
+    if (gl_InvocationID == 0)
+    {
+        spvUnsafeArray<float4, 2> _56 = spvUnsafeArray<float4, 2>({ gl_in[0].ipFoo, gl_in[1].ipFoo });
+        patchOut.pFoo = _56;
+    }
+}
+
diff --git a/reference/opt/shaders-msl/masking/copy-arrays.mask-location-0.msl2.tesc b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-0.msl2.tesc
new file mode 100644
index 00000000..e9dd68dc
--- /dev/null
+++ b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-0.msl2.tesc
@@ -0,0 +1,191 @@
+#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 main0_out
+{
+    float4 gl_Position;
+};
+
+struct main0_patchOut
+{
+    spvUnsafeArray<float4, 2> pFoo;
+};
+
+struct main0_in
+{
+    float4 iFoo_0 [[attribute(0)]];
+    float4 iFoo_1 [[attribute(1)]];
+    float4 ipFoo [[attribute(2)]];
+};
+
+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(main0_in in [[stage_in]], uint gl_InvocationID [[thread_index_in_threadgroup]], uint gl_PrimitiveID [[threadgroup_position_in_grid]], device main0_out* spvOut [[buffer(28)]], constant uint* spvIndirectParams [[buffer(29)]], device main0_patchOut* spvPatchOut [[buffer(27)]], device MTLQuadTessellationFactorsHalf* spvTessLevel [[buffer(26)]], threadgroup main0_in* gl_in [[threadgroup(0)]])
+{
+    threadgroup float4 Foo[4][2];
+    device main0_out* gl_out = &spvOut[gl_PrimitiveID * 4];
+    device main0_patchOut& patchOut = spvPatchOut[gl_PrimitiveID];
+    if (gl_InvocationID < spvIndirectParams[0])
+        gl_in[gl_InvocationID] = in;
+    threadgroup_barrier(mem_flags::mem_threadgroup);
+    if (gl_InvocationID >= 4)
+        return;
+    gl_out[gl_InvocationID].gl_Position = float4(1.0);
+    spvUnsafeArray<float4, 2> _38 = spvUnsafeArray<float4, 2>({ gl_in[gl_InvocationID].iFoo_0, gl_in[gl_InvocationID].iFoo_1 });
+    spvArrayCopyFromStackToThreadGroup1(Foo[gl_InvocationID], _38.elements);
+    if (gl_InvocationID == 0)
+    {
+        spvUnsafeArray<float4, 2> _56 = spvUnsafeArray<float4, 2>({ gl_in[0].ipFoo, gl_in[1].ipFoo });
+        patchOut.pFoo = _56;
+    }
+}
+
diff --git a/reference/opt/shaders-msl/masking/copy-arrays.mask-location-1.msl2.multi-patch.tesc b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-1.msl2.multi-patch.tesc
new file mode 100644
index 00000000..a2ad0102
--- /dev/null
+++ b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-1.msl2.multi-patch.tesc
@@ -0,0 +1,79 @@
+#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 main0_out
+{
+    spvUnsafeArray<float4, 2> Foo;
+    float4 gl_Position;
+};
+
+struct main0_patchOut
+{
+    spvUnsafeArray<float4, 2> pFoo;
+};
+
+struct main0_in
+{
+    spvUnsafeArray<float4, 2> iFoo;
+    float4 ipFoo;
+};
+
+kernel void main0(uint3 gl_GlobalInvocationID [[thread_position_in_grid]], device main0_out* spvOut [[buffer(28)]], constant uint* spvIndirectParams [[buffer(29)]], device main0_patchOut* spvPatchOut [[buffer(27)]], device MTLQuadTessellationFactorsHalf* spvTessLevel [[buffer(26)]], device main0_in* spvIn [[buffer(22)]])
+{
+    device main0_out* gl_out = &spvOut[gl_GlobalInvocationID.x - gl_GlobalInvocationID.x % 4];
+    device main0_patchOut& patchOut = spvPatchOut[gl_GlobalInvocationID.x / 4];
+    device main0_in* gl_in = &spvIn[min(gl_GlobalInvocationID.x / 4, spvIndirectParams[1] - 1) * spvIndirectParams[0]];
+    uint gl_InvocationID = gl_GlobalInvocationID.x % 4;
+    uint gl_PrimitiveID = min(gl_GlobalInvocationID.x / 4, spvIndirectParams[1]);
+    gl_out[gl_InvocationID].gl_Position = float4(1.0);
+    gl_out[gl_InvocationID].Foo = gl_in[gl_InvocationID].iFoo;
+    if (gl_InvocationID == 0)
+    {
+        spvUnsafeArray<float4, 2> _56 = spvUnsafeArray<float4, 2>({ gl_in[0].ipFoo, gl_in[1].ipFoo });
+        patchOut.pFoo = _56;
+    }
+}
+
diff --git a/reference/opt/shaders-msl/masking/copy-arrays.mask-location-1.msl2.tesc b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-1.msl2.tesc
new file mode 100644
index 00000000..3da1d18c
--- /dev/null
+++ b/reference/opt/shaders-msl/masking/copy-arrays.mask-location-1.msl2.tesc
@@ -0,0 +1,83 @@
+#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 main0_out
+{
+    spvUnsafeArray<float4, 2> Foo;
+    float4 gl_Position;
+};
+
+struct main0_patchOut
+{
+    spvUnsafeArray<float4, 2> pFoo;
+};
+
+struct main0_in
+{
+    float4 iFoo_0 [[attribute(0)]];
+    float4 iFoo_1 [[attribute(1)]];
+    float4 ipFoo [[attribute(2)]];
+};
+
+kernel void main0(main0_in in [[stage_in]], uint gl_InvocationID [[thread_index_in_threadgroup]], uint gl_PrimitiveID [[threadgroup_position_in_grid]], device main0_out* spvOut [[buffer(28)]], constant uint* spvIndirectParams [[buffer(29)]], device main0_patchOut* spvPatchOut [[buffer(27)]], device MTLQuadTessellationFactorsHalf* spvTessLevel [[buffer(26)]], threadgroup main0_in* gl_in [[threadgroup(0)]])
+{
+    device main0_out* gl_out = &spvOut[gl_PrimitiveID * 4];
+    device main0_patchOut& patchOut = spvPatchOut[gl_PrimitiveID];
+    if (gl_InvocationID < spvIndirectParams[0])
+        gl_in[gl_InvocationID] = in;
+    threadgroup_barrier(mem_flags::mem_threadgroup);
+    if (gl_InvocationID >= 4)
+        return;
+    gl_out[gl_InvocationID].gl_Position = float4(1.0);
+    spvUnsafeArray<float4, 2> _38 = spvUnsafeArray<float4, 2>({ gl_in[gl_InvocationID].iFoo_0, gl_in[gl_InvocationID].iFoo_1 });
+    gl_out[gl_InvocationID].Foo = _38;
+    if (gl_InvocationID == 0)
+    {
+        spvUnsafeArray<float4, 2> _56 = spvUnsafeArray<float4, 2>({ gl_in[0].ipFoo, gl_in[1].ipFoo });
+        patchOut.pFoo = _56;
+    }
+}
+