OOB SharedMemory access even in checked mode

[0x7CFE]

CubeCl revision: 4372c41

#[cube(launch)]
pub fn test_kernel(out: &mut Tensor<u32>) {
    if ABSOLUTE_POS_X >= CUBE_COUNT_X {
        return;
    }

    // Deliberately insufficient allocation of 1 element vs 32 actually needed
    let mut memory = SharedMemory::<u32>::new(1);

    memory[UNIT_POS_X * 2] = ABSOLUTE_POS_X;
    memory[UNIT_POS_X * 2 + 1] = UNIT_POS_X;

    sync_units();

    out[ABSOLUTE_POS_X * 2] = memory[UNIT_POS_X * 2];
    out[ABSOLUTE_POS_X * 2 + 1] = memory[UNIT_POS_X * 2 + 1];

    // Try this as well
    // out[ABSOLUTE_POS_X * 2] = memory[CUBE_DIM_X * 2 - UNIT_POS_X * 2 - 2];
    // out[ABSOLUTE_POS_X * 2 + 1] = memory[CUBE_DIM_X * 2 - UNIT_POS_X * 2 - 1];

    // Just to make sure indices are adequate
    // out[ABSOLUTE_POS_X * 2] = ABSOLUTE_POS_X;
    // out[ABSOLUTE_POS_X * 2 + 1] = UNIT_POS_X;
}

#[test]
fn topology() {
    type MyBackend = burn::backend::wgpu::JitBackend<WgpuRuntime, f32, i32, u32>;
    let device = Default::default();

    let client = cubecl::wgpu::WgpuRuntime::client(&Default::default());
    let output = Tensor::<MyBackend, 2, Int>::zeros([50, 2], &device);

    // unsafe 
    {
        kernel::test_kernel::launch(
            &client,
            CubeCount::Static(50, 1, 1),
            CubeDim::new(16, 1, 1),
            output.clone().into_primitive().as_tensor_arg::<u32>(1)
        );
    };

    println!("{output}")
}

The output is:

Tensor {
  data:
[[0, 0],
 [1, 1],
 [2, 2],
 [3, 3],
 [4, 4],
 [5, 5],
 [6, 6],
 [7, 7],
 [8, 8],
 [9, 9],
 [10, 10],
 [11, 11],
 [12, 12],
 [13, 13],
 [14, 14],
 [15, 15],
 [16, 0],
 [17, 1],
 [18, 2],
 [19, 3],
 [20, 4],
 [21, 5],
 [22, 6],
 [23, 7],
 [24, 8],
 [25, 9],
 [26, 10],
 [27, 11],
 [28, 12],
 [29, 13],
 [30, 14],
 [31, 15],
 [32, 0],
 [33, 1],
 [34, 2],
 [35, 3],
 [36, 4],
 [37, 5],
 [38, 6],
 [39, 7],
 [40, 8],
 [41, 9],
 [42, 10],
 [43, 11],
 [44, 12],
 [45, 13],
 [46, 14],
 [47, 15],
 [48, 0],
 [49, 1]],
  shape:  [50, 2],
  device:  DefaultDevice,
  backend:  "jit<wgpu<wgsl>>",
  kind:  "Int",
  dtype:  "i32",
}