#version 460


#define BARRIER() \
    memoryBarrierShared(); \
    barrier();

#extension GL_EXT_mesh_shader : enable

layout(local_size_x = 32, local_size_y=1, local_size_z=1) in;

// test use of shared memory in task shaders:
layout(binding=0) writeonly uniform image2D uni_image;
uniform block0 {
    uint uni_value;
};
shared vec4 mem[10];

// test use of task memory in task shaders:
struct Task {
    vec2 dummy;
    vec2 submesh[3];
};

taskPayloadSharedEXT Task mytask;
void main()
{
    uint iid = gl_LocalInvocationID.x;
    uint gid = gl_WorkGroupID.x;

    // 1. shared memory load and stores
    for (uint i = 0; i < 10; ++i) {
        mem[i] = vec4(i + uni_value);
    }
    imageStore(uni_image, ivec2(iid), mem[gid]);
    imageStore(uni_image, ivec2(iid), mem[gid+1]);

    BARRIER();

    // 2. task memory stores

    mytask.dummy      = vec2(30.0, 31.0);
    mytask.submesh[0] = vec2(32.0, 33.0);
    mytask.submesh[1] = vec2(34.0, 35.0);
    mytask.submesh[2] = mytask.submesh[gid%2];

    BARRIER();

    // 3. emit task count under uniform control flow
    EmitMeshTasksEXT(3U, 1U, 1U);
}

#extension GL_KHR_shader_subgroup_basic: enable
void basic_works (void)
{
  gl_SubgroupSize;
  gl_SubgroupInvocationID;
  subgroupBarrier();
  subgroupMemoryBarrier();
  subgroupMemoryBarrierBuffer();
  subgroupMemoryBarrierImage();
  subgroupElect();
  gl_NumSubgroups;                  // allowed in task
  gl_SubgroupID;                    // allowed in task
  subgroupMemoryBarrierShared();    // allowed in task
}

#extension GL_KHR_shader_subgroup_ballot: enable
void ballot_works(vec4 f4) {
  gl_SubgroupEqMask;
  gl_SubgroupGeMask;
  gl_SubgroupGtMask;
  gl_SubgroupLeMask;
  gl_SubgroupLtMask;
  subgroupBroadcast(f4, 0);
  subgroupBroadcastFirst(f4);
  uvec4 ballot = subgroupBallot(false);
  subgroupInverseBallot(uvec4(0x1));
  subgroupBallotBitExtract(ballot, 0);
  subgroupBallotBitCount(ballot);
  subgroupBallotInclusiveBitCount(ballot);
  subgroupBallotExclusiveBitCount(ballot);
  subgroupBallotFindLSB(ballot);
  subgroupBallotFindMSB(ballot);
}

#extension GL_KHR_shader_subgroup_vote: enable
void vote_works(vec4 f4)
{
  subgroupAll(true);
  subgroupAny(false);
  subgroupAllEqual(f4);
}

#extension GL_KHR_shader_subgroup_shuffle: enable
#extension GL_KHR_shader_subgroup_shuffle_relative: enable
void shuffle_works(vec4 f4)
{
  subgroupShuffle(f4, 0);
  subgroupShuffleXor(f4, 0x1);
  subgroupShuffleUp(f4, 1);
  subgroupShuffleDown(f4, 1);
}

#extension GL_KHR_shader_subgroup_arithmetic: enable
void arith_works(vec4 f4)
{
  uvec4 ballot;
  subgroupAdd(f4);
  subgroupMul(f4);
  subgroupMin(f4);
  subgroupMax(f4);
  subgroupAnd(ballot);
  subgroupOr(ballot);
  subgroupXor(ballot);
  subgroupInclusiveAdd(f4);
  subgroupInclusiveMul(f4);
  subgroupInclusiveMin(f4);
  subgroupInclusiveMax(f4);
  subgroupInclusiveAnd(ballot);
  subgroupInclusiveOr(ballot);
  subgroupInclusiveXor(ballot);
  subgroupExclusiveAdd(f4);
  subgroupExclusiveMul(f4);
  subgroupExclusiveMin(f4);
  subgroupExclusiveMax(f4);
  subgroupExclusiveAnd(ballot);
  subgroupExclusiveOr(ballot);
  subgroupExclusiveXor(ballot);
}

#extension GL_KHR_shader_subgroup_clustered: enable
void clustered_works(vec4 f4)
{
  uvec4 ballot = uvec4(0x55,0,0,0);
  subgroupClusteredAdd(f4, 2);
  subgroupClusteredMul(f4, 2);
  subgroupClusteredMin(f4, 2);
  subgroupClusteredMax(f4, 2);
  subgroupClusteredAnd(ballot, 2);
  subgroupClusteredOr(ballot, 2);
  subgroupClusteredXor(ballot, 2);
}

#extension GL_KHR_shader_subgroup_quad: enable
void quad_works(vec4 f4)
{
  subgroupQuadBroadcast(f4, 0);
  subgroupQuadSwapHorizontal(f4);
  subgroupQuadSwapVertical(f4);
  subgroupQuadSwapDiagonal(f4);
}