Writing safely to a layer

Thank you for the answer. i will explain in details:

Here is what I'm doing:

Render function:
- create a temporary layer (same size as input layer)
- call iterate function to iterate over all pixels

Here is code snippet:
ERR(in_data->utils->new_world(in_data->effect_ref, input->width,
             input->height, PF_NewWorldFlag_NONE, &tmpLayer));
err = suites.Iterate8Suite()->iterate(in_data,
                                            progress_base,
                                            progress_final,
                                             src,   
                                            area,                   
                                            refcon,                   
                                            pix_fn,
                                             dst);

Here is what the per-pixel function (i.e. pix_fn) does:

1) read from input buffer at (x,y)
2) Clear the temp buffer region (x-width, y-height) to (x+width, y+height) in tmpLayer
3) Render the transformed kira to tmpLayer
4) For every pixel (i,j) in region (x-width, y-height) to (x+width, y+height) in tmpLayer
a) read from tmpLayer location (i,j)
b) do some math using read value
c) write a result value to output buffer location (i,j)

Steps 2 and 3 implemented with the following code:
    dstRect.bottom = (A_short)data->height-1;
    dstRect.top    = 0;
    dstRect.left   = 0;
    dstRect.right  = (A_short)data->width-1;
    ERR(suites.FillMatteSuite1()->fill(
         data->effect_ref,
        NULL,
        &dstRect,
        temp ));
    ERR(suites.WorldTransformSuite1()->transform_world(
        data->effect_ref,
        data->quality,
        PF_MF_Alpha_STRAIGHT,
         PF_Field_FRAME,
        &context->shapeLayerParam.u.ld,
        &compositeMode,
        NULL,
        &mat, 1,
        TRUE,
        &dstRect,
        temp ));

The result is corrupt rendering.

I suspect this is because the iterate function can process two pixels at once.

Between step 2 and step 4c there maybe FillMatteSuite1 that is called by another thread and may result in a corrupted result

I thought about three solutions:

1-allocate the temporary layer in the iterate function.

2-dont use the iterate function and process sequentially pixels

3-find a way to lock the temporary layer to avoid corruption

By using solution 2 and not using the iterate() function and instead processing all the pixels sequentially on my own, the corruption problem was resolved.

I will try the solution 1, hoping AE is doing good caching and memory management to avoid performance hit.

I dont know if there is a way to do the solution 3.

i would be happy to hear If you have any ideas that could help