2290cc0f79e9f9c255e10bd3775c711591c34e99,python/tvm/topi/cuda/scatter.py,,gen_ir_2d,#Any#Any#Any#Any#Any#Any#,89

Before Change

    out_ptr = ib.buffer_ptr(out)
    data_ptr = ib.buffer_ptr(data)

    with ib.new_scope():
        bx = te.thread_axis("blockIdx.x")
        ib.scope_attr(bx, "thread_extent", n)
        tx = te.thread_axis("threadIdx.x")
        ib.scope_attr(tx, "thread_extent", warp_size)
        with ib.for_range(0, ceil_div(c, warp_size), name="j") as j_:
            j = j_ * warp_size + tx
            with ib.if_scope(j < c):
                idx = bx * c + j
                out_ptr[idx] = data_ptr[idx]

    indices_ptr = ib.buffer_ptr(indices)
    updates_ptr = ib.buffer_ptr(updates)

    ni = indices.shape[0]

After Change

    n = data.shape[0]
    c = data.shape[1]

    ib = tvm.tir.ir_builder.create()

    out_ptr = ib.buffer_ptr(out)
    data_ptr = ib.buffer_ptr(data)

    _memcpy_ir(ib, out_ptr, data_ptr, data.shape)

    indices_ptr = ib.buffer_ptr(indices)
    updates_ptr = ib.buffer_ptr(updates)

Instances