UE5 Niagara Compilation succeeded but certain simulation stage's performance time is N/A

[CheapMeow]

I would like to ask why Niagara is compiled successfully, but the running time of a certain simulation stage is N/A?
The data interface of this simulation stage is grid 2d, and I have initialized its num cells to non-zero values. Logically speaking, there should be elements in it that can be traversed.

If needed, it is what is in the module Draw Depth.
It is module input.

It is module output.

It is custom hlsl in the middle.

const int2 IndexOffsets [ 9 ] = 
{
    int2(-1,-1),
    int2(-1, 0),
    int2(-1, 1),
    int2(0,-1),
    int2(0, 0),
    int2(0, 1),
    int2(1,-1),
    int2(1, 0),
    int2(1, 1)
};

OutDepth = 0;

#if GPU_SIMULATION

int3 Index;
NeighborGrid.UnitToIndex(float3(CellPos, 0), Index.x,Index.y,Index.z);

int3 NumCells;
NeighborGrid.GetNumCells(NumCells.x, NumCells.y, NumCells.z);

int MaxNeighborsPerCell;
NeighborGrid.MaxNeighborsPerCell(MaxNeighborsPerCell);

float minDepth = 1 / 0;
int minDepthIdx = -1;

for(int xxx = 0; xxx < 9; ++xxx)
{
    for(int i = 0; i < MaxNeighborsPerCell; i++)
    {
        const int3 IndexToUse = int3(Index.x + IndexOffsets[xxx].x, Index.y + IndexOffsets[xxx].y, 0);
        
        int NeighborLinearIndex;
        NeighborGrid.NeighborGridIndexToLinear(IndexToUse.x, IndexToUse.y, IndexToUse.z, i, NeighborLinearIndex);
        
        int CurrNeighborIdx;
        NeighborGrid.GetParticleNeighbor(NeighborLinearIndex, CurrNeighborIdx);

        bool myBool; 
        float2 ProjectedParticlePos;
        ParticleReader.GetVector2DByIndex<Attribute="ProjectedPos">(CurrNeighborIdx, myBool, ProjectedParticlePos);

        float ParticleDepth;
        ParticleReader.GetFloatByIndex<Attribute="Depth">(CurrNeighborIdx, myBool, ParticleDepth);
        
        float DistanceFromCellToParticle = length(ProjectedParticlePos - CellPos);
        float ProjectedRenderSpriteSize = RenderSpriteSize/ParticleDepth;

        if(CurrNeighborIdx != -1
        && DistanceFromCellToParticle < ProjectedRenderSpriteSize
        && ParticleDepth != 0)
        {
            minDepth = minDepth < ParticleDepth ? minDepth : ParticleDepth;
            minDepthIdx = minDepth < ParticleDepth ? minDepthIdx : CurrNeighborIdx;
        }
    }
}

OutDepth = minDepth;
#endif

Furthermore, if you want to know what does the module do:

For each pixels in screen space, find particles projected in the screen space position of pixel. To do this, create a new Neighbor Grid 3D, which is used for accelerating searching particles. From the view of occupying space, the Neighbor Grid 3D should occupys full space of screen, so one cell of Neighbor Grid 3D may overlaps with many pixels.
Set Iteration Source as Data Interface, and drag RenderGrid2D to Data Interface, then in the module, if you set namespace of a value in map out as STACKCONTEXT, the value will be stored into Data Interface automatically.
Now, if you want for loop pixels, logically you first traverse Grid 2D, stored your desired value into cells of Grid 2D. After travering finished, you write value from Grid 2D to a render target.

[CheapMeow]

After one day of work, I found the error disappear surprisingly.
From the time I encounter the problem to the time I found the problem disappear, I have done many step, most of them are irrelevant about Grid 2D. The only relevant step is adding a built-in Grid 2D Set Resolution module. But if I delete the module, the performance logging time is still work well. So it is a magic and I don’t want to pay attention to this issue anymore.