[require]
shader model >= 5.0

[uav 1]
format r32 typeless
size (raw_buffer, 4)

1 0 0 0

[compute shader todo]
RWByteAddressBuffer u : register(u1);
groupshared uint m;

    [numthreads(32, 1, 1)]
void main(uint local_idx : SV_GroupIndex, uint group_id : SV_GroupID)
{
    if (!local_idx)
        m = group_id.x;
    GroupMemoryBarrierWithGroupSync();
    InterlockedAdd(m, group_id.x);
    GroupMemoryBarrierWithGroupSync();
    if (!local_idx)
        u.Store(4 * group_id.x, m);
}

[test]
todo(sm<6) dispatch 4 1 1
probe uav 1 (0) rui (0)
probe uav 1 (1) rui (33)
probe uav 1 (2) rui (66)
probe uav 1 (3) rui (99)


[uav 1]
format r32 typeless
size (raw_buffer, 4)

1 0 0 0

[compute shader todo]
RWByteAddressBuffer u : register(u1);
groupshared int m;

    [numthreads(32, 1, 1)]
void main(uint local_idx : SV_GroupIndex, uint group_id : SV_GroupID)
{
    if (!local_idx)
        m = group_id.x;
    GroupMemoryBarrierWithGroupSync();
    InterlockedAdd(m, -group_id.x);
    GroupMemoryBarrierWithGroupSync();
    if (!local_idx)
        u.Store(4 * group_id.x, m);
}

[test]
todo(sm<6) dispatch 4 1 1
probe uav 1 (0) ri (0)
probe uav 1 (1) ri (-31)
probe uav 1 (2) ri (-62)
probe uav 1 (3) ri (-93)


[uav 1]
format r32 float
size (buffer, 8)

1 1 1 1 0 0 0 0

[uav 2]
format r32 sint
size (buffer, 8)

1 1 1 1 0 0 0 0

[compute shader todo]
uniform uint idx;

#define GROUP_SIZE 4

struct data
{
    float f;
    uint u;
};

RWBuffer<float> u : register(u1);
RWBuffer<uint> u2 : register(u2);
groupshared data m[GROUP_SIZE];

    [numthreads(GROUP_SIZE, 1, 1)]
void main(uint local_idx : SV_GroupIndex, uint group_id : SV_GroupID,
        uint thread_id : SV_DispatchThreadID)
{
    uint i;
    if (!local_idx)
    {
        for (i = 0; i < GROUP_SIZE; ++i)
        {
            m[i].f = group_id.x;
            m[i].u = group_id.x;
        }
    }
    GroupMemoryBarrierWithGroupSync();
    InterlockedAdd(m[0].u, 2);
    InterlockedAdd(m[idx].u, 1);
    GroupMemoryBarrierWithGroupSync();
    for (i = 0; i < local_idx; ++i)
    {
        m[local_idx].f += group_id.x;
        m[local_idx].u += group_id.x;
    }
    u[thread_id.x] = m[local_idx].f;
    u2[thread_id.x] = m[local_idx].u;
}

[test]
uniform 0 uint 1
todo(sm<6) dispatch 2 1 1
probe uav 1 (0)  r (0.0)
probe uav 1 (1)  r (0.0)
probe uav 1 (2)  r (0.0)
probe uav 1 (3)  r (0.0)
probe uav 1 (4)  r (1.0)
probe uav 1 (5)  r (2.0)
probe uav 1 (6)  r (3.0)
probe uav 1 (7)  r (4.0)
probe uav 2 (0)  ri (8)
probe uav 2 (1)  ri (4)
probe uav 2 (2)  ri (0)
probe uav 2 (3)  ri (0)
probe uav 2 (4)  ri (9)
probe uav 2 (5)  ri (6)
probe uav 2 (6)  ri (3)
probe uav 2 (7)  ri (4)


[uav 1]
format r32 typeless
size (raw_buffer, 1)

0

[compute shader todo]
RWByteAddressBuffer u : register(u1);
groupshared uint m;

    [numthreads(32, 1, 1)]
void main(uint local_idx : SV_GroupIndex)
{
    uint orig;
    if (!local_idx)
        m = 7;
    GroupMemoryBarrierWithGroupSync();
    InterlockedCompareExchange(m, local_idx, local_idx + 32, orig);
    GroupMemoryBarrierWithGroupSync();
    if (!local_idx)
        u.Store(0, m);
}

[test]
todo(sm<6) dispatch 1 1 1
probe uav 1 (0) rui (39)