Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

MSL: added task shader support #2402

Merged
merged 1 commit into from
Nov 1, 2024
Merged

MSL: added task shader support #2402

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 20 additions & 13 deletions reference/opt/shaders-msl/mesh/mesh-shader-basic-lines.msl3.spv14.vk.nocompat.mesh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,6 +89,13 @@ struct BlockOutPrim
float4 b;
};

struct TaskPayload
{
float a;
float b;
int c;
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(2u, 3u, 4u);

struct spvPerVertex
Expand Down Expand Up @@ -116,33 +123,33 @@ struct spvPerPrimitive
using spvMesh_t = mesh<spvPerVertex, spvPerPrimitive, 24, 22, topology::line>;

static inline __attribute__((always_inline))
void _4(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, thread uint& gl_LocalInvocationIndex, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, threadgroup uint2& spvMeshSizes)
void _4(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, thread uint& gl_LocalInvocationIndex, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, const object_data TaskPayload& payload, threadgroup uint2& spvMeshSizes)
{
spvSetMeshOutputsEXT(gl_LocalInvocationIndex, spvMeshSizes, 24u, 22u);
float3 _158 = float3(gl_GlobalInvocationID);
float _159 = _158.x;
gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_Position = float4(_159, _158.yz, 1.0);
float3 _163 = float3(gl_GlobalInvocationID);
float _164 = _163.x;
gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_Position = float4(_164, _163.yz, 1.0);
gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_PointSize = 2.0;
gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_ClipDistance[0] = 4.0;
vOut[gl_LocalInvocationIndex] = float4(_159, _158.yz, 2.0);
vOut[gl_LocalInvocationIndex] = float4(_164, _163.yz, 2.0);
outputs[gl_LocalInvocationIndex].a = float4(5.0);
outputs[gl_LocalInvocationIndex].b = float4(6.0);
threadgroup_barrier(mem_flags::mem_threadgroup);
if (gl_LocalInvocationIndex < 22u)
{
vPrim[gl_LocalInvocationIndex] = float4(float3(gl_WorkGroupID), 3.0);
prim_outputs[gl_LocalInvocationIndex].a = float4(0.0);
prim_outputs[gl_LocalInvocationIndex].b = float4(1.0);
prim_outputs[gl_LocalInvocationIndex].a = float4(payload.a);
prim_outputs[gl_LocalInvocationIndex].b = float4(payload.b);
gl_PrimitiveLineIndicesEXT[gl_LocalInvocationIndex] = uint2(0u, 1u) + uint2(gl_LocalInvocationIndex);
int _206 = int(gl_GlobalInvocationID.x);
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_PrimitiveID = _206;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_Layer = _206 + 1;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_ViewportIndex = _206 + 2;
int _217 = int(gl_GlobalInvocationID.x);
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_PrimitiveID = _217;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_Layer = _217 + 1;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_ViewportIndex = _217 + 2;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_CullPrimitiveEXT = short((gl_GlobalInvocationID.x & 1u) != 0u);
}
}

[[mesh]] void main0(uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], spvMesh_t spvMesh)
[[mesh]] void main0(uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], spvMesh_t spvMesh, const object_data TaskPayload& payload [[payload]])
{
threadgroup uint2 spvMeshSizes;
threadgroup spvUnsafeArray<uint2, 22> gl_PrimitiveLineIndicesEXT;
Expand All @@ -154,7 +161,7 @@ void _4(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, threa
threadgroup spvUnsafeArray<BlockOutPrim, 22> prim_outputs;
threadgroup spvUnsafeArray<float, 16> shared_float;
if (gl_LocalInvocationIndex == 0) spvMeshSizes.y = 0u;
_4(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID, gl_MeshVerticesEXT, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, spvMeshSizes);
_4(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID, gl_MeshVerticesEXT, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, payload, spvMeshSizes);
threadgroup_barrier(mem_flags::mem_threadgroup);
if (spvMeshSizes.y == 0)
{
Expand Down
32 changes: 16 additions & 16 deletions reference/opt/shaders-msl/mesh/mesh-shader-basic-triangle.msl3.spv14.vk.nocompat.mesh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,6 +81,13 @@ struct BlockOutPrim
float4 b;
};

struct TaskPayload
{
float a;
float b;
int c;
};

struct gl_MeshPerPrimitiveEXT
{
uint gl_PrimitiveID [[primitive_id]];
Expand All @@ -91,13 +98,6 @@ struct gl_MeshPerPrimitiveEXT

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(2u, 3u, 4u);

struct TaskPayload
{
float a;
float b;
int c;
};

struct spvPerVertex
{
float4 gl_Position [[position]];
Expand All @@ -123,7 +123,7 @@ struct spvPerPrimitive
using spvMesh_t = mesh<spvPerVertex, spvPerPrimitive, 24, 22, topology::triangle>;

static inline __attribute__((always_inline))
void _4(threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, thread uint& gl_LocalInvocationIndex, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, threadgroup spvUnsafeArray<uint3, 22>& gl_PrimitiveTriangleIndicesEXT, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, threadgroup uint2& spvMeshSizes)
void _4(threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, thread uint& gl_LocalInvocationIndex, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, const object_data TaskPayload& payload, threadgroup spvUnsafeArray<uint3, 22>& gl_PrimitiveTriangleIndicesEXT, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, threadgroup uint2& spvMeshSizes)
{
spvSetMeshOutputsEXT(gl_LocalInvocationIndex, spvMeshSizes, 24u, 22u);
float3 _27 = float3(gl_GlobalInvocationID);
Expand All @@ -138,18 +138,18 @@ void _4(threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT,
if (gl_LocalInvocationIndex < 22u)
{
vPrim[gl_LocalInvocationIndex] = float4(float3(gl_WorkGroupID), 3.0);
prim_outputs[gl_LocalInvocationIndex].a = float4(0.0);
prim_outputs[gl_LocalInvocationIndex].b = float4(1.0);
prim_outputs[gl_LocalInvocationIndex].a = float4(payload.a);
prim_outputs[gl_LocalInvocationIndex].b = float4(payload.b);
gl_PrimitiveTriangleIndicesEXT[gl_LocalInvocationIndex] = uint3(0u, 1u, 2u) + uint3(gl_LocalInvocationIndex);
int _116 = int(gl_GlobalInvocationID.x);
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_PrimitiveID = _116;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_Layer = _116 + 1;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_ViewportIndex = _116 + 2;
int _123 = int(gl_GlobalInvocationID.x);
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_PrimitiveID = _123;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_Layer = _123 + 1;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_ViewportIndex = _123 + 2;
gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_CullPrimitiveEXT = short((gl_GlobalInvocationID.x & 1u) != 0u);
}
}

[[mesh]] void main0(uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], spvMesh_t spvMesh)
[[mesh]] void main0(uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], spvMesh_t spvMesh, const object_data TaskPayload& payload [[payload]])
{
threadgroup uint2 spvMeshSizes;
threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24> gl_MeshVerticesEXT;
Expand All @@ -161,7 +161,7 @@ void _4(threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT,
threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22> gl_MeshPrimitivesEXT;
threadgroup spvUnsafeArray<float, 16> shared_float;
if (gl_LocalInvocationIndex == 0) spvMeshSizes.y = 0u;
_4(gl_MeshVerticesEXT, gl_LocalInvocationIndex, gl_GlobalInvocationID, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, gl_PrimitiveTriangleIndicesEXT, gl_MeshPrimitivesEXT, spvMeshSizes);
_4(gl_MeshVerticesEXT, gl_LocalInvocationIndex, gl_GlobalInvocationID, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, payload, gl_PrimitiveTriangleIndicesEXT, gl_MeshPrimitivesEXT, spvMeshSizes);
threadgroup_barrier(mem_flags::mem_threadgroup);
if (spvMeshSizes.y == 0)
{
Expand Down
22 changes: 22 additions & 0 deletions reference/opt/shaders-msl/task/task-basic.msl3.spv14.vk.nocompat.task
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
#include <metal_stdlib>
#include <simd/simd.h>

using namespace metal;

struct TaskPayload
{
float a;
float b;
int c;
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(1u);

[[object]] void main0(mesh_grid_properties spvMgp, object_data TaskPayload& payload [[payload]])
{
payload.a = 1.2000000476837158203125;
payload.b = 2.2999999523162841796875;
payload.c = 3;
spvMgp.set_threadgroups_per_grid(uint3(1u, 2u, 3u));
}

23 changes: 23 additions & 0 deletions reference/opt/shaders-msl/task/task-func.msl3.spv14.vk.nocompat.task
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
#include <metal_stdlib>
#include <simd/simd.h>

using namespace metal;

struct TaskPayload
{
float a;
float b;
int c;
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(1u);

[[object]] void main0(mesh_grid_properties spvMgp, object_data TaskPayload& payload [[payload]])
{
payload.a = 1.2000000476837158203125;
payload.b = 2.2999999523162841796875;
payload.c = 3;
threadgroup_barrier(mem_flags::mem_threadgroup);
spvMgp.set_threadgroups_per_grid(uint3(1u, 2u, 3u));
}

18 changes: 9 additions & 9 deletions reference/shaders-msl/mesh/mesh-shader-basic-lines.msl3.spv14.vk.nocompat.mesh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,15 +89,15 @@ struct BlockOutPrim
float4 b;
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(2u, 3u, 4u);

struct TaskPayload
{
float a;
float b;
int c;
};

constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(2u, 3u, 4u);

struct spvPerVertex
{
float4 gl_Position [[position]];
Expand Down Expand Up @@ -133,7 +133,7 @@ void main3(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, th
}

static inline __attribute__((always_inline))
void main2(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, thread uint& gl_LocalInvocationIndex, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, threadgroup uint2& spvMeshSizes)
void main2(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, thread uint& gl_LocalInvocationIndex, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, const object_data TaskPayload& payload, threadgroup uint2& spvMeshSizes)
{
spvSetMeshOutputsEXT(gl_LocalInvocationIndex, spvMeshSizes, 24u, 22u);
gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_Position = float4(float3(gl_GlobalInvocationID), 1.0);
Expand All @@ -146,19 +146,19 @@ void main2(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, th
if (gl_LocalInvocationIndex < 22u)
{
vPrim[gl_LocalInvocationIndex] = float4(float3(gl_WorkGroupID), 3.0);
prim_outputs[gl_LocalInvocationIndex].a = float4(0.0);
prim_outputs[gl_LocalInvocationIndex].b = float4(1.0);
prim_outputs[gl_LocalInvocationIndex].a = float4(payload.a);
prim_outputs[gl_LocalInvocationIndex].b = float4(payload.b);
main3(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID);
}
}

static inline __attribute__((always_inline))
void _4(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, thread uint& gl_LocalInvocationIndex, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, threadgroup uint2& spvMeshSizes)
void _4(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, thread uint& gl_LocalInvocationIndex, threadgroup spvUnsafeArray<gl_MeshPerPrimitiveEXT, 22>& gl_MeshPrimitivesEXT, thread uint3& gl_GlobalInvocationID, threadgroup spvUnsafeArray<gl_MeshPerVertexEXT, 24>& gl_MeshVerticesEXT, threadgroup spvUnsafeArray<float4, 24>& vOut, threadgroup spvUnsafeArray<BlockOut, 24>& outputs, threadgroup spvUnsafeArray<float4, 22>& vPrim, thread uint3& gl_WorkGroupID, threadgroup spvUnsafeArray<BlockOutPrim, 22>& prim_outputs, const object_data TaskPayload& payload, threadgroup uint2& spvMeshSizes)
{
main2(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID, gl_MeshVerticesEXT, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, spvMeshSizes);
main2(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID, gl_MeshVerticesEXT, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, payload, spvMeshSizes);
}

[[mesh]] void main0(uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], spvMesh_t spvMesh)
[[mesh]] void main0(uint gl_LocalInvocationIndex [[thread_index_in_threadgroup]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]], uint3 gl_WorkGroupID [[threadgroup_position_in_grid]], spvMesh_t spvMesh, const object_data TaskPayload& payload [[payload]])
{
threadgroup uint2 spvMeshSizes;
threadgroup spvUnsafeArray<uint2, 22> gl_PrimitiveLineIndicesEXT;
Expand All @@ -170,7 +170,7 @@ void _4(threadgroup spvUnsafeArray<uint2, 22>& gl_PrimitiveLineIndicesEXT, threa
threadgroup spvUnsafeArray<BlockOutPrim, 22> prim_outputs;
threadgroup spvUnsafeArray<float, 16> shared_float;
if (gl_LocalInvocationIndex == 0) spvMeshSizes.y = 0u;
_4(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID, gl_MeshVerticesEXT, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, spvMeshSizes);
_4(gl_PrimitiveLineIndicesEXT, gl_LocalInvocationIndex, gl_MeshPrimitivesEXT, gl_GlobalInvocationID, gl_MeshVerticesEXT, vOut, outputs, vPrim, gl_WorkGroupID, prim_outputs, payload, spvMeshSizes);
threadgroup_barrier(mem_flags::mem_threadgroup);
if (spvMeshSizes.y == 0)
{
Expand Down
Loading
Loading