.. _aliasing: ======================================================= Understanding Memory Aliasing for Speed and Correctness ======================================================= The aggressive reuse of memory is one of the ways through which PyTensor makes code fast, and it is important for the correctness and speed of your program that you understand how PyTensor might alias buffers. This section describes the principles based on which PyTensor handles memory, and explains when you might want to alter the default behaviour of some functions and methods for faster performance. The Memory Model: Two Spaces ============================ There are some simple principles that guide PyTensor's handling of memory. The main idea is that there is a pool of memory managed by PyTensor, and PyTensor tracks changes to values in that pool. - PyTensor manages its own memory space, which typically does not overlap with the memory of normal Python variables that non-PyTensor code creates. - PyTensor functions only modify buffers that are in PyTensor's memory space. - PyTensor's memory space includes the buffers allocated to store ``shared`` variables and the temporaries used to evaluate functions. - Physically, PyTensor's memory space may be spread across the host, a GPU device(s), and in the future may even include objects on a remote machine. - The memory allocated for a ``shared`` variable buffer is unique: it is never aliased to another ``shared`` variable. - PyTensor's managed memory is constant while PyTensor functions are not running and PyTensor's library code is not running. - The default behaviour of a function is to return user-space values for outputs, and to expect user-space values for inputs. The distinction between PyTensor-managed memory and user-managed memory can be broken down by some PyTensor functions (e.g. ``shared``, ``get_value`` and the constructors for ``In`` and ``Out``) by using a ``borrow=True`` flag. This can make those methods faster (by avoiding copy operations) at the expense of risking subtle bugs in the overall program (by aliasing memory). The rest of this section is aimed at helping you to understand when it is safe to use the ``borrow=True`` argument and reap the benefits of faster code. Borrowing when Creating Shared Variables ======================================== A ``borrow`` argument can be provided to the shared-variable constructor. .. testcode:: borrow import numpy, pytensor np_array = numpy.ones(2, dtype='float32') s_default = pytensor.shared(np_array) s_false = pytensor.shared(np_array, borrow=False) s_true = pytensor.shared(np_array, borrow=True) By default (``s_default``) and when explicitly setting ``borrow=False``, the shared variable we construct gets a (deep) copy of ``np_array``. So changes we subsequently make to ``np_array`` have no effect on our shared variable. .. testcode:: borrow np_array += 1 # now it is an array of 2.0 s print(s_default.get_value()) print(s_false.get_value()) print(s_true.get_value()) .. testoutput:: borrow [ 1. 1.] [ 1. 1.] [ 2. 2.] If we are running this with the CPU as the device, then changes we make to ``np_array`` right away will show up in ``s_true.get_value()`` because NumPy arrays are mutable, and ``s_true`` is using the ``np_array`` object as it's internal buffer. However, this aliasing of ``np_array`` and ``s_true`` is not guaranteed to occur, and may occur only temporarily even if it occurs at all. It is not guaranteed to occur because if PyTensor is using a GPU device, then the ``borrow`` flag has no effect. It may occur only temporarily because if we call an PyTensor function that updates the value of ``s_true`` the aliasing relationship may or may not be broken (the function is allowed to update the ``shared`` variable by modifying its buffer, which will preserve the aliasing, or by changing which buffer the variable points to, which will terminate the aliasing). Take home message: It is a safe practice (and a good idea) to use ``borrow=True`` in a ``shared`` variable constructor when the ``shared`` variable stands for a large object (in terms of memory footprint) and you do not want to create copies of it in memory. It is not a reliable technique to use ``borrow=True`` to modify ``shared`` variables through side-effect, because with some devices (e.g. GPU devices) this technique will not work. Borrowing when Accessing Value of Shared Variables ================================================== Retrieving ---------- A ``borrow`` argument can also be used to control how a ``shared`` variable's value is retrieved. .. testcode:: borrow s = pytensor.shared(np_array) v_false = s.get_value(borrow=False) # N.B. borrow default is False v_true = s.get_value(borrow=True) When ``borrow=False`` is passed to ``get_value``, it means that the return value may not be aliased to any part of PyTensor's internal memory. When ``borrow=True`` is passed to ``get_value``, it means that the return value might be aliased to some of PyTensor's internal memory. But both of these calls might create copies of the internal memory. The reason that ``borrow=True`` might still make a copy is that the internal representation of a ``shared`` variable might not be what you expect. When you create a ``shared`` variable by passing a NumPy array for example, then ``get_value()`` must return a NumPy array too. That's how PyTensor can make the GPU use transparent. But when you are using a GPU (or in the future perhaps a remote machine), then the numpy.ndarray is not the internal representation of your data. If you really want PyTensor to return its internal representation and never copy it then you should use the ``return_internal_type=True`` argument to ``get_value``. It will never cast the internal object (always return in constant time), but might return various datatypes depending on contextual factors (e.g. the compute device, the dtype of the NumPy array). .. testcode:: borrow v_internal = s.get_value(borrow=True, return_internal_type=True) It is possible to use ``borrow=False`` in conjunction with ``return_internal_type=True``, which will return a deep copy of the internal object. This is primarily for internal debugging, not for typical use. For the transparent use rewrites, there is the policy that ``get_value()`` always return by default the same object type it received when the ``shared`` variable was created. So if you created manually data on the gpu and create a ``shared`` variable on the gpu with this data, ``get_value`` will always return gpu data even when ``return_internal_type=False``. Take home message: It is safe (and sometimes much faster) to use ``get_value(borrow=True)`` when your code does not modify the return value. Do not use this to modify a ``shared`` variable by side-effect because it will make your code device-dependent. Modification of GPU variables through this sort of side-effect is impossible. Assigning --------- ``Shared`` variables also have a ``set_value`` method that can accept an optional ``borrow=True`` argument. The semantics are similar to those of creating a new ``shared`` variable - ``borrow=False`` is the default and ``borrow=True`` means that PyTensor may reuse the buffer you provide as the internal storage for the variable. A standard pattern for manually updating the value of a ``shared`` variable is as follows: .. testsetup:: borrow def some_inplace_fn(v): return v .. testcode:: borrow s.set_value( some_inplace_fn(s.get_value(borrow=True)), borrow=True) This pattern works regardless of the computing device, and when the latter makes it possible to expose PyTensor's internal variables without a copy, then it proceeds as fast as an in-place update. .. When ``shared`` variables are allocated on the GPU, the transfers to and from the GPU device memory can be costly. Here are a few tips to ensure fast and efficient use of GPU memory and bandwidth: * Prior to PyTensor 0.3.1, ``set_value`` did not work in-place on the GPU. This meant that, sometimes, GPU memory for the new value would be allocated before the old memory was released. If you're running near the limits of GPU memory, this could cause you to run out of GPU memory unnecessarily. *Solution*: update to a newer version of PyTensor. * If you are going to swap several chunks of data in and out of a ``shared`` variable repeatedly, you will want to reuse the memory that you allocated the first time if possible - it is both faster and more memory efficient. *Solution*: upgrade to a recent version of PyTensor (>0.3.0) and consider padding your source data to make sure that every chunk is the same size. * It is also worth mentioning that, current GPU copying routines support only contiguous memory. So PyTensor must make the value you provide *C-contiguous* prior to copying it. This can require an extra copy of the data on the host. *Solution*: make sure that the value you assign to a GpuArraySharedVariable is *already* *C-contiguous*. .. _borrowfunction: Borrowing when Constructing Function Objects ============================================ A ``borrow`` argument can also be provided to the ``In`` and ``Out`` objects that control how ``pytensor.function`` handles its argument[s] and return value[s]. .. testcode:: import pytensor import pytensor.tensor as pt from pytensor.compile.io import In, Out x = pt.matrix() y = 2 * x f = pytensor.function([In(x, borrow=True)], Out(y, borrow=True)) Borrowing an input means that PyTensor will treat the argument you provide as if it were part of PyTensor's pool of temporaries. Consequently, your input may be reused as a buffer (and overwritten!) during the computation of other variables in the course of evaluating that function (e.g. ``f``). Borrowing an output means that PyTensor will not insist on allocating a fresh output buffer every time you call the function. It will possibly reuse the same one as on a previous call, and overwrite the old content. Consequently, it may overwrite old return values through side-effect. Those return values may also be overwritten in the course of evaluating another compiled function (for example, the output may be aliased to a ``shared`` variable). So be careful to use a borrowed return value right away before calling any more PyTensor functions. The default is of course to not borrow internal results. It is also possible to pass a ``return_internal_type=True`` flag to the ``Out`` variable which has the same interpretation as the ``return_internal_type`` flag to the ``shared`` variable's ``get_value`` function. Unlike ``get_value()``, the combination of ``return_internal_type=True`` and ``borrow=True`` arguments to ``Out()`` are not guaranteed to avoid copying an output value. They are just hints that give more flexibility to the compilation and rewriting of the graph. Take home message: When an input ``x`` to a function is not needed after the function returns and you would like to make it available to PyTensor as additional workspace, then consider marking it with ``In(x, borrow=True)``. It may make the function faster and reduce its memory requirement. When a return value ``y`` is large (in terms of memory footprint), and you only need to read from it once, right away when it's returned, then consider marking it with an ``Out(y, borrow=True)``.