Grunnur, a generalized API for CUDA and OpenCL

Manual

Grunnur is an abstraction layer on top of PyCUDA/PyOpenCL. Its main purpose is to provide a uniform API for high-level GPGPU algorithms automating some common tasks.

Consider the following example, which is very similar to the one from the index page on PyCUDA documentation:

import numpy
from grunnur import any_api, Context, Queue, Program, Array

N = 256

context = Context.from_devices([any_api.platforms[0].devices[0]])
queue = Queue(context)

program = Program(
    context,
    """
    KERNEL void multiply_them(
        GLOBAL_MEM float *dest,
        GLOBAL_MEM float *a,
        GLOBAL_MEM float *b)
    {
        const SIZE_T i = get_global_id(0);
        dest[i] = a[i] * b[i];
    }
    """)

multiply_them = program.kernel.multiply_them

a = numpy.random.randn(N).astype(numpy.float32)
b = numpy.random.randn(N).astype(numpy.float32)
a_dev = Array.from_host(queue, a)
b_dev = Array.from_host(queue, b)
dest_dev = Array.empty(context, a.shape, a.dtype)

multiply_them(queue, N, None, dest_dev, a_dev, b_dev)
print((dest_dev.get(queue) - a * b == 0).all())

If you are familiar with PyCUDA or PyOpenCL, you will easily understand most of the the steps we have made here. The any_api object returns some API of the ones available (so, depending of whether PyOpenCL or PyCUDA are installed). More precise control over API is available via API discovery functions.

The abstraction from specific C interface of OpenCL or CUDA is achieved by using generic API module on the Python side, and special macros (KERNEL, GLOBAL_MEM, and others) on the kernel side.

The argument of Program constructor can also be a template, which is quite useful for metaprogramming, and also used to compensate for the lack of complex number operations in CUDA and OpenCL. Let us illustrate both scenarios by making the initial example multiply complex arrays. The template engine of choice in grunnur is Mako, and you are encouraged to read about it as it is quite useful. For the purpose of this example all we need to know is that ${python_expression()} is a synthax construction which renders the expression result.

import numpy
from numpy.linalg import norm
import grunnur.dtypes as dtypes
import grunnur.functions as functions
from grunnur import any_api, Context, Queue, Program, Array

context = Context.from_devices([any_api.platforms[0].devices[0]])
queue = Queue(context)

N = 256
dtype = numpy.complex64

program = Program(
    context, """
    KERNEL void multiply_them(
        GLOBAL_MEM ${ctype} *dest,
        GLOBAL_MEM ${ctype} *a,
        GLOBAL_MEM ${ctype} *b)
    {
      const SIZE_T i = get_global_id(0);
      dest[i] = ${mul}(a[i], b[i]);
    }
    """,
    render_globals=dict(
        ctype=dtypes.ctype(dtype),
        mul=functions.mul(dtype, dtype)))

multiply_them = program.kernel.multiply_them

r1 = numpy.random.randn(N).astype(numpy.float32)
r2 = numpy.random.randn(N).astype(numpy.float32)
a = r1 + 1j * r2
b = r1 - 1j * r2
a_dev = Array.from_host(queue, a)
b_dev = Array.from_host(queue, b)
dest_dev = Array.empty(context, a.shape, a.dtype)

multiply_them(queue, N, None, dest_dev, a_dev, b_dev)
print(norm(dest_dev.get(queue) - a * b) / norm(a * b) <= 1e-6)

Here we have passed two values to the template: ctype (a string with C type name), and mul which is a Module object containing a single multiplication function. The object is created by a function mul() which takes data types being multiplied and returns a module that was parametrized accordingly. Inside the template the variable mul is essentially the prefix for all the global C objects (functions, structures, macros etc) from the module. If there is only one public object in the module (which is recommended), it is a common practice to give it the name consisting just of the prefix, so that it could be called easily from the parent code.

For more information on modules, see Tutorial: modules and snippets; the complete list of things available in Grunnur can be found in API reference.

Tutorial: modules and snippets

Modules and snippets are important primitives in Grunnur. Even if you do not write modules yourself, you will most likely use operations from the functions module, which are essentially module factories (callables returning Module objects). Therefore it helps if you know how they work under the hood.

Snippets

Snippets are Mako template defs (essentially functions returning rendered text) with the associated dictionary of render globals. When a snippet is used in a template, the result is quite straightworward: its template function is called, rendering and returning its contents, just as a normal Mako def.

Let us demonstrate it with a simple example. Consider the following snippet:

add = Snippet.from_callable(
    lambda varname: """
    ${varname} + ${num}
    """,
    render_globals=dict(num=1))

Now we can compile a template which uses this snippet:

program = Program(
    context,
    """
    KERNEL void test(GLOBAL_MEM int *arr)
    {
        const SIZE_T idx = get_global_id(0);
        int x = arr[idx];
        arr[idx] = ${add('x')};
    }
    """,
    render_globals=dict(add=add))

As a result, the code that gets compiled is

KERNEL void test(GLOBAL_MEM int *arr)
{
    const SIZE_T idx = get_global_id(0);
    int x = arr[idx];
    arr[idx] = x + 1;
}

If the snippet is used without parentheses (e.g. ${add}), it is equivalent to calling it without arguments (${add()}).

The root code that gets passed to Program() can be viewed as a snippet with an empty signature.

Modules

Modules are quite similar to snippets in a sense that they are also Mako defs with an associated dictionary of render keywords. The difference lies in the way they are processed. Consider a module containing a single function:

add = Module.from_callable(
    lambda prefix, arg: """
    FUNCTION int ${prefix}(int x)
    {
        return x + ${num} + ${arg};
    }
    """,
    name="foobar",
    render_globals=dict(num=1))

Modules contain complete C entities (function, macros, structures) and get rendered in the root level of the source file. In order to avoid name clashes, their def gets a string as a first argument, which it has to use to prefix these entities’ names. If the module contains only one entity that is supposed to be used by the parent code, it is a good idea to set its name to prefix only, to simplify its usage.

Let us now create a kernel that uses this module:

program = Program(
    context,
    """
    KERNEL void test(GLOBAL_MEM int *arr)
    {
        const SIZE_T idx = get_global_id(0);
        int x = arr[idx];
        arr[idx] = ${add(2)}(x);
    }
    """,
    render_globals=dict(add=add))

Before the compilation render keywords are inspected, and if a module object is encountered, the following things happen:

1. This object’s render_globals are inspected recursively and any modules there are rendered in the same way as described here, producing a source file.

2. The module itself gets assigned a new prefix and its template function is rendered with this prefix as the first argument, with the positional arguments given following it. The result is attached to the source file.

3. The corresponding value in the current render_globals is replaced by the newly assigned prefix.

With the code above, the rendered module will produce the code

FUNCTION int _mod_foobar_0_(int x)
{
    return x + 1 + 2;
}

and the add keyword in the render_globals gets its value changed to _mod_foobar_0_. Then the main code is rendered and appended to the previously renderd parts, giving

FUNCTION int _mod_foobar_0_(int x)
{
    return x + 1 + 2;
}

KERNEL void test(GLOBAL_MEM int *arr)
{
    const SIZE_T idx = get_global_id(0);
    int x = arr[idx];
    arr[idx] = _mod_foobar_0_(x);
}

which is then passed to the compiler. If your module’s template def does not take any arguments except for prefix, you can call it in the parent template just as ${add} (without empty parentheses).

Warning

Note that add in this case is not a string, it is an object that has __str__() defined. If you want to concatenate a module prefix with some other string, you have to either call str() explicitly (str(add) + "abc"), or concatenate it inside a template (${add}abc).

Modules can reference snippets in their render_globals, which, in turn, can reference other modules. This produces a tree-like structure with the snippet made from the code passed by user at the root. When it is rendered, it is traversed depth-first, modules are extracted from it and arranged in a flat list in the order of appearance. Their positions in render_globals are replaced by assigned prefixes. This flat list is then rendered, producing a single source file being fed to the compiler.

Note that if the same module object was used without arguments in several other modules or in the kernel itself, it will only be rendered once. Therefore one can create a “root” module with the data structure declaration and then use that structure in other modules without producing type errors on compilation.

Other constructors

If the arguments are not know at compile time, you can use DefTemplate.from_string with a regular constructor:

argnames = ['varname']
add = Snippet(
    DefTemplate.from_string("_func", argnames, "${varname} + ${num}"),
    render_globals=dict(num=1))

Modules can be constructed in a similar way. The only difference is that the template must have at least one positional parameter which will receive the prefix value.

Alternatively, one can create a snippet with no parameters or a module with a single prefix parameter with a from_string() constructor:

add = Module.from_string("""
    FUNCTION int ${prefix}(int x)
    {
        return x + ${num};
    }
    """,
    render_globals=dict(num=1))

Module and snippet discovery

Sometimes you may want to pass a module or a snippet inside a template as an attribute of a custom object. In order for CLUDA to be able to discover and process it without modifying your original object, you need to make your object comply to a discovery protocol. The protocol method takes a processing function and is expected to return a new object of the same class with the processing function applied to all the attributes that may contain a module or a snippet. By default, objects of type tuple, list, and dict are discoverable.

For example:

class MyClass:

    def __init__(self, coeff, mul_module, div_module):
        self.coeff = coeff
        self.mul = mul_module
        self.div = div_module

    def __process_modules__(self, process):
        return MyClass(self.coeff, process(self.mul), process(self.div))

Nontrivial example

Modules were introduced to help split big kernels into small reusable pieces which in CUDA or OpenCL program would be put into different source or header files. For example, a random number generator may be assembled from a function generating random integers, a function transforming these integers into random numbers with a certain distribution, and a parallel computation calling these functions and saving results to global memory. These functions can be extracted into separate modules, so that a user could call them from some custom kernel if he does not need to store the intermediate results.

Going further with this example, one notices that functions that produce randoms with sophisticated distributions are often based on simpler distributions. For instance, the commonly used Marsaglia algorithm for generating Gamma-distributed random numbers requires several uniformly and normally distributed randoms. Normally distributed randoms, in turn, require several uniformly distributed randoms — with the range which differs from the one for uniformly distributed randoms used by the initial Gamma distribution. Instead of copy-pasting the function or setting its parameters dynamically (which in more complicated cases may affect the performance), one just specifies the dependencies between modules and lets the underlying system handle things.

The final render tree may look like:

Snippet(
    PureParallel,
    render_globals={
        base_rng -> Snippet(...)
        gamma -> Snippet(
    }                Gamma,
)                    render_globals = {
                         uniform -> Snippet(...)
                         normal -> Snippet(
                     }                 Normal,
                 )                     render_globals = {
                                           uniform -> Snippet(...)
                                       }
                                   )

Public API

API discovery

In many applications it would be enough to use dynamic module attributes to get an API object:

from grunnur import cuda_api
from grunnur import opencl_api
from grunnur import any_api

For a finer programmatic control one can use the methods of the API class:

class grunnur.API

A generalized GPGPU API.

classmethod all_available() → List[API]

Returns a list of API objects for which backends are available.

classmethod all_by_shortcut(shortcut: Optional[str] = None) → List[API]

If shortcut is a string, returns a list of one API object whose id attribute has its shortcut attribute equal to it (or raises an error if it was not found, or its backend is not available).

If shortcut is None, returns a list of all available API objects.

Parameters

shortcut – an API shortcut to match.

classmethod from_api_id(api_id: grunnur.adapter_base.APIID) → API

Creates an API object out of an identifier.

Parameters

api_id – API identifier.

id: grunnur.adapter_base.APIID

This API’s ID.

property platforms

A list of this API’s Platform objects.

shortcut: str

A shortcut for this API (to use in all_by_shortcut(), usually coming from some kind of a CLI). Equal to id.shortcut.

class grunnur.adapter_base.APIID

An ID of an API object.

shortcut: str

This API’s shortcut.

grunnur.cuda_api_id() → grunnur.adapter_base.APIID

Returns the identifier of CUDA API.

grunnur.opencl_api_id() → grunnur.adapter_base.APIID

Returns the identifier of OpenCL API.

grunnur.all_api_ids() → List[grunnur.adapter_base.APIID]

Returns a list of identifiers for all APIs available.

Platforms

A platform is an OpenCL term, but we use it for CUDA API as well for the sake of uniformity. Naturally, there will always be a single (dummy) platform in CUDA.

class grunnur.Platform

A generalized GPGPU platform.

classmethod all(api: API) → List[Platform]

Returns a list of platforms available for the given API.

Parameters

api – the API to search in.

classmethod all_by_masks(api: API, include_masks: Optional[Sequence[str]] = None, exclude_masks: Optional[Sequence[str]] = None) → List[Platform]

Returns a list of all platforms with names satisfying the given criteria.

Parameters

• api – the API to search in.

• include_masks – a list of strings (treated as regexes), one of which must match with the platform name.

• exclude_masks – a list of strings (treated as regexes), neither of which must match with the platform name.

classmethod from_backend_platform(obj) → Platform

Wraps a backend platform object into a Grunnur platform object.

classmethod from_index(api: API, platform_idx: int) → Platform

Creates a platform based on its index in the list returned by the API.

Parameters

• api – the API to search in.

• platform_idx – the target platform’s index.

api: API

The API object this platform belongs to.

property devices

A list of this device’s Device objects.

name: str

The platform’s name.

vendor: str

The platform’s vendor.

version: str

The platform’s version.

Devices

class grunnur.Device

A generalized GPGPU device.

classmethod all(platform: Platform) → List[Device]

Returns a list of devices available for the given platform.

Parameters

platform – the platform to search in.

classmethod all_by_masks(platform: Platform, include_masks: Optional[Sequence[str]] = None, exclude_masks: Optional[Sequence[str]] = None, unique_only: bool = False, include_pure_parallel_devices: bool = False) → List[Device]

Returns a list of all devices satisfying the given criteria.

Parameters

• platform – the platform to search in.

• include_masks – a list of strings (treated as regexes), one of which must match with the device name.

• exclude_masks – a list of strings (treated as regexes), neither of which must match with the device name.

• unique_only – if True, only return devices with unique names.

• include_pure_parallel_devices – if True, include devices with params.max_total_local_size equal to 1.

classmethod from_backend_device(obj) → Device

Wraps a backend device object into a Grunnur device object.

classmethod from_index(platform: Platform, device_idx: int) → Device

Creates a device based on its index in the list returned by the API.

Parameters

• platform – the API to search in.

• device_idx – the target device’s index.

name: str

This device’s name.

property params

Returns a DeviceParameters object associated with this device.

platform: Platform

The Platform object this device belongs to.

class grunnur.adapter_base.DeviceParameters

An object containing device’s specifications.

abstract property compute_units

The number of multiprocessors (CUDA)/compute units (OpenCL) for the device.

abstract property local_mem_banks

The number of independent channels for shared (CUDA)/local (OpenCL) memory, which can be used from one warp without request serialization.

abstract property local_mem_size

The size of shared (CUDA)/local (OpenCL) memory (in bytes).

abstract property max_local_sizes

The maximum number of threads in one block (CUDA), or work items in one work group (OpenCL) for each of the available dimensions.

abstract property max_num_groups

The maximum number of blocks (CUDA)/work groups (OpenCL) for each of the available dimensions.

abstract property max_total_local_size

The maximum total number of threads in one block (CUDA), or work items in one work group (OpenCL).

abstract property type

Device type.

abstract property warp_size

The number of threads (CUDA)/work items (OpenCL) that are executed synchronously (within one multiprocessor/compute unit).

class grunnur.adapter_base.DeviceType

An enum representing a device’s type.

CPU = 1

CPU type

GPU = 2

GPU type

Device discovery

grunnur.platforms_and_devices_by_mask(api, quantity: Optional[int] = 1, platform_include_masks: Optional[Sequence[str]] = None, platform_exclude_masks: Optional[Sequence[str]] = None, device_include_masks: Optional[Sequence[str]] = None, device_exclude_masks: Optional[Sequence[str]] = None, unique_devices_only: bool = False, include_pure_parallel_devices: bool = False) → List[Tuple[Platform, List[Device]]]

Returns all tuples (platform, list of devices) where the platform name and device names satisfy the given criteria, and there are at least quantity devices in the list.

Parameters

• quantity – the number of devices to find. If None, find all matching devices belonging to a single platform.

• platform_include_masks – passed to Platform.all_by_masks().

• platform_exclude_masks – passed to Platform.all_by_masks().

• device_include_masks – passed to Device.all_by_masks().

• device_exclude_masks – passed to Device.all_by_masks().

• unique_devices_only – passed to Device.all_by_masks().

• include_pure_parallel_devices – passed to Device.all_by_masks().

grunnur.select_devices(api, interactive: bool = False, quantity: Optional[int] = 1, **device_filters) → List[Device]

Using the results from platforms_and_devices_by_mask(), either lets the user select the devices (from the ones matching the criteria) interactively, or takes the first matching list of quantity devices.

Parameters

• interactive – if True, shows a dialog to select the devices. If False, selects the first matching ones.

• quantity – passed to platforms_and_devices_by_mask().

• device_filters – passed to platforms_and_devices_by_mask().

Contexts

class grunnur.Context

GPGPU context.

deactivate()

CUDA API only: deactivates this context, popping all the CUDA context objects from the stack.

classmethod from_backend_contexts(backend_contexts, take_ownership: bool = False) → Context

Creates a context from a single or several backend device contexts. If take_ownership is True, this object will be responsible for the lifetime of backend context objects (important for CUDA backend).

classmethod from_backend_devices(backend_devices) → Context

Creates a context from a single or several backend device objects.

classmethod from_criteria(api: API, interactive: bool = False, devices_num: Optional[int] = 1, **device_filters) → Context

Finds devices matching the given criteria and creates a Context object out of them.

Parameters

• interactive – passed to select_devices().

• devices_num – passed to select_devices() as quantity.

• device_filters – passed to select_devices().

classmethod from_devices(devices: Union[Device, Iterable[Device]]) → Context

Creates a context from a device or an iterable of devices.

Parameters

devices – one or several devices to use.

api: API

The API this context is based on.

devices: Tuple[Device]

Devices in this context.

platform: Platform

The platform this context is based on.

Queues

class grunnur.Queue(context: Context, device_idx: Optional[int] = None)

A queue on a single device.

Parameters

• context – a context on which to create a queue.

• device_idx – device index in the context on which to create a queue. If there is more than one device in the context, it must be specified.

synchronize()

Blocks until sub-queues on all devices are empty.

context: Context

This queue’s context.

device: Device

Device object this queue operates on.

device_idx: int

Device index this queue operates on.

class grunnur.MultiQueue(context: Context, queues: Optional[Iterable[Queue]] = None)

A queue on multiple devices.

Parameters

• context – a context on which to create a queue.

• queues – single-device queues (must belong to distinct devices).

classmethod on_device_idxs(context: Context, device_idxs: Iterable[int]) → MultiQueue

Creates a queue from provided device indexes (in the context).

synchronize()

Blocks until queues on all devices are empty.

context: Context

This queue’s context.

device_idxs: Set[int]

Device indices (in the context) this queue operates on.

devices: Dict[int, Device]

Device objects associated with device indices.

queues: Dict[int, Queue]

Single-device queues associated with device indices.

Buffers and arrays

class grunnur.Buffer

A memory buffer on device.

classmethod allocate(context: Context, size: int, device_idx: Optional[int] = None) → Buffer

Allocate a buffer of size bytes.

Parameters

• context – the context to use.

• size – the buffer’s size in bytes.

• device_idx – the device to allocate on (can be omitted in a single-device context).

get(queue: Queue, host_array: numpy.ndarray, async_: bool = False)

Copy the contents of the buffer to the host array.

Parameters

• queue – the queue to use for the transfer.

• host_array – the destination array.

• async – if True, the transfer is performed asynchronously.

get_sub_region(origin: int, size: int) → Buffer

Return a buffer object describing a subregion of this buffer.

Parameters

• origin – the offset of the subregion.

• size – the size of the subregion.

set(queue: Queue, buf: Union[numpy.ndarray, Buffer], no_async: bool = False)

Copy the contents of the host array or another buffer to this buffer.

Parameters

• queue – the queue to use for the transfer.

• buf – the source - numpy array or a Buffer object.

• no_async – if True, the transfer blocks until completion.

context: Context

Context this buffer is allocated on.

device_idx: int

The index of the device this buffer is allocated on.

property offset

Offset of this buffer (in bytes) from the beginning of the physical allocation it resides in.

property size

This buffer’s size (in bytes).

class grunnur.Array

Array on a single device.

__getitem__(slices) → Array

Returns a view of this array.

classmethod empty(context: Context, shape: Sequence[int], dtype: numpy.dtype, allocator: Optional[Callable[[int, int], Buffer]] = None, device_idx: Optional[int] = None) → Array

Creates an empty array.

Parameters

• shape – array shape.

• dtype – array data type.

• allocator – an optional callable taking two integer arguments (buffer size in bytes, and the device to allocate it on) and returning a Buffer object.

• device_idx – the index of the device on which to allocate the array.

classmethod from_host(queue: Queue, host_arr: numpy.ndarray) → Array

Creates an array object from a host array.

Parameters

• queue – the queue to use for the transfer.

• host_arr – the source array.

get(queue: Queue, dest: Optional[numpy.ndarray] = None, async_: bool = False) → numpy.ndarray

Copies the contents of the array to the host array and returns it.

Parameters

• queue – the queue to use for the transfer.

• dest – the destination array. If None, the target array is created.

• async – if True, the transfer is performed asynchronously.

set(queue: Queue, array: Union[numpy.ndarray, Array], no_async: bool = False)

Copies the contents of the host array to the array.

Parameters

• queue – the queue to use for the transfer.

• array – the source array.

• no_async – if True, the transfer blocks until completion.

context: Context

Context this array is allocated on.

dtype: numpy.dtype

Array item data type.

shape: Tuple[int, …]

Array shape.

strides: Tuple[int, …]

Array strides.

class grunnur.array.BaseSplay

Base class for splay strategies for MultiArray.

abstract __call__(arr: ArrayLike, devices: Dict[int, grunnur.Device]) → Dict[int, ArrayLike]

Creates a dictionary of views of an array-like object for each of the given devices.

Parameters

• arr – an array-like object.

• devices – a dictionary of device indices matched to device objects.

ArrayLike

The type of an array-like object (the one having a shape and supporting views via __getitem__())

alias of TypeVar(‘ArrayLike’)

class grunnur.MultiArray

An array on multiple devices.

class CloneSplay

Copies the given array to each device.

class EqualSplay

Splays the given array equally between the devices using the outermost dimension. The outermost dimension should be larger or equal to the number of devices.

classmethod empty(context: Context, shape: Sequence[int], dtype: numpy.dtype, allocator: Optional[Callable[[int, int], Buffer]] = None, device_idxs: Optional[Iterable[int]] = None, splay: Optional[grunnur.array.BaseSplay] = None) → MultiArray

Creates an empty array.

Parameters

• shape – array shape.

• dtype – array data type.

• allocator – an optional callable taking two integer arguments (buffer size in bytes, and the device to allocate it on) and returning a Buffer object.

• device_idx – the index of the device on which to allocate the array.

• splay – the splay strategy (if None, an EqualSplay object is used).

classmethod from_host(mqueue: MultiQueue, host_arr: numpy.ndarray, splay: Optional[grunnur.array.BaseSplay] = None) → MultiArray

Creates an array object from a host array.

Parameters

• mqueue – the queue to use for the transfer.

• host_arr – the source array.

• splay – the splay strategy (if None, an EqualSplay object is used).

get(mqueue: MultiQueue, dest: Optional[numpy.ndarray] = None, async_: bool = False) → numpy.ndarray

Copies the contents of the array to the host array and returns it.

Parameters

• mqueue – the queue to use for the transfer.

• dest – the destination array. If None, the target array is created.

• async – if True, the transfer is performed asynchronously.

set(mqueue: MultiQueue, array: Union[numpy.ndarray, MultiArray], no_async: bool = False)

Copies the contents of the host array to the array.

Parameters

• mqueue – the queue to use for the transfer.

• array – the source array.

• no_async – if True, the transfer blocks until completion.

context: Context

Context this array is allocated on.

dtype: numpy.dtype

Array item data type.

shape: Tuple[int, …]

Array shape.

shapes: Dict[int, Tuple[int, …]]

Sub-array shapes matched to device indices.

Programs and kernels

class grunnur.Program(context: Context, template_src: Union[str, Callable[[…], str], DefTemplate, Snippet], device_idxs: Optional[Sequence[int]] = None, no_prelude: bool = False, fast_math: bool = False, render_args: Union[List, Tuple] = [], render_globals: Dict = {}, compiler_options: Iterable[str] = [], keep: bool = False, constant_arrays: Mapping[str, Tuple[int, numpy.dtype]] = {})

A compiled program on device(s).

Parameters

• context – context to compile the program on.

• template_src – a string with the source code, or a Mako template source to render.

• device_idxs – a list of device numbers to compile on. If None, compile on all context’s devices.

• no_prelude – do not add prelude to the rendered source.

• fast_math – compile using fast (but less accurate) math functions.

• render_args – a list of positional args to pass to the template.

• render_globals – a dictionary of globals to pass to the template.

• compiler_options – a list of options to pass to the backend compiler.

• keep – keep the intermediate files in a temporary directory.

• constant_arrays – (CUDA only) a dictionary name: (size, dtype) of global constant arrays to be declared in the program.

set_constant_array(queue: Queue, name: str, arr: Union[Array, numpy.ndarray])

Uploads a constant array to the context’s devices (CUDA only).

Parameters

• queue – the queue to use for the transfer.

• name – the name of the constant array symbol in the code.

• arr – either a device or a host array.

context: Context

The context this program was compiled for.

kernel: grunnur.program.KernelHub

An object whose attributes are Kernel objects with the corresponding names.

sources: Dict[int, str]

Source files used for each device.

class grunnur.program.KernelHub

An object providing access to the host program’s kernels.

__getattr__(kernel_name: str) → Kernel

Returns a Kernel object for a function (CUDA)/kernel (OpenCL) with the name kernel_name.

class grunnur.program.Kernel

A kernel compiled for multiple devices.

__call__(queue: Union[grunnur.Queue, grunnur.MultiQueue], global_size: Union[int, Sequence[int], Dict[int, Union[int, Sequence[int]]]], local_size: Union[int, Sequence[int], None, Dict[int, Optional[Union[int, Sequence[int]]]]] = None, *args, **kwds)

A shortcut for Kernel.prepare() and subsequent PreparedKernel.__call__(). See their doc entries for details.

prepare(global_size: Union[int, Sequence[int], Dict[int, Union[int, Sequence[int]]]], local_size: Union[int, Sequence[int], None, Dict[int, Optional[Union[int, Sequence[int]]]]] = None) → PreparedKernel

Prepares the kernel for execution.

If local_size or global_size are integer, they will be treated as 1-tuples.

One can pass specific global and local sizes for each device using dictionaries keyed with device indices. This achieves another purpose: the kernel will only be prepared for those devices, and not for all devices available in the context.

Parameters

• global_size – the total number of threads (CUDA)/work items (OpenCL) in each dimension (column-major). Note that there may be a maximum size in each dimension as well as the maximum number of dimensions. See DeviceParameters for details.

• local_size – the number of threads in a block (CUDA)/work items in a work group (OpenCL) in each dimension (column-major). If None, it will be chosen automatically.

property max_total_local_sizes

The maximum possible number of threads in a block (CUDA)/work items in a work group (OpenCL) for this kernel.

class grunnur.program.PreparedKernel

A kernel specialized for execution on a set of devices with all possible preparations and checks performed.

__call__(queue: Union[grunnur.Queue, grunnur.MultiQueue], *args, **kwds)

Enqueues the kernel on the devices in the given queue. The kernel must have been prepared for all of these devices.

If an argument is a Array or Buffer object, it must belong to the device on which the kernel is being executed (so queue must only have one device).

If an argument is a MultiArray, it should have subarrays on all the devices from the given queue.

If an argument is a numpy scalar, it will be passed to the kernel directly.

If an argument is a integer-keyed dict, its values corresponding to the device indices the kernel is executed on will be passed as kernel arguments.

Parameters

• args – kernel arguments.

• kwds – backend-specific keyword parameters.

Returns

a list of Event objects for enqueued kernels in case of PyOpenCL.

Static kernels

class grunnur.StaticKernel(context: Context, template_src: Union[str, Callable[[…], str], DefTemplate, Snippet], name: str, global_size: Union[int, Sequence[int], Dict[int, Union[int, Sequence[int]]]], local_size: Union[int, Sequence[int], None, Dict[int, Optional[Union[int, Sequence[int]]]]] = None, render_globals: Dict = {}, constant_arrays: Mapping[str, Tuple[int, numpy.dtype]] = {}, **kwds)

An object containing a GPU kernel with fixed call sizes.

The globals for the source template will contain an object with the name static of the type VsizeModules containing the id/size functions to be used instead of regular ones.

Parameters

• context – context to compile the kernel on.

• template_src – a string with the source code, or a Mako template source to render.

• name – the kernel’s name.

• global_size – see prepare().

• local_size – see prepare().

• render_globals – a dictionary of globals to pass to the template.

• constant_arrays – (CUDA only) a dictionary name: (size, dtype) of global constant arrays to be declared in the program.

__call__(queue, *args)

Execute the kernel. In case of the OpenCL backend, returns a pyopencl.Event object.

Parameters

• queue – the multi-device queue to use.

• args – kernel arguments. See grunnur.program.PreparedKernel.__call__().

set_constant_array(queue: Queue, name: str, arr: Union[Array, numpy.ndarray])

Uploads a constant array to the context’s devices (CUDA only).

Parameters

• queue – the queue to use for the transfer.

• name – the name of the constant array symbol in the code.

• arr – either a device or a host array.

queue: Queue

The queue this static kernel was compiled and prepared for.

sources: Dict[int, str]

Source files used for each device.

class grunnur.vsize.VsizeModules(local_id, local_size, group_id, num_groups, global_id, global_size, global_flat_id, global_flat_size, begin)

A collection of modules passed to grunnur.StaticKernel. Should be used instead of regular group/thread id functions.

begin: grunnur.Module

Provides the statement ${begin} that should be used at the start of a static kernel function.

global_flat_id: grunnur.Module

Provides the function VSIZE_T ${global_flat_id}() returning the global id of the current thread with all dimensions flattened.

global_flat_size: grunnur.Module

Provides the function VSIZE_T ${global_flat_size}(). returning the global size of with all dimensions flattened.

global_id: grunnur.Module

Provides the function VSIZE_T ${global_id}(int dim) returning the global id of the current thread.

global_size: grunnur.Module

Provides the function VSIZE_T ${global_size}(int dim) returning the global size along dimension dim.

group_id: grunnur.Module

Provides the function VSIZE_T ${group_id}(int dim) returning the group id of the current thread.

local_id: grunnur.Module

Provides the function VSIZE_T ${local_id}(int dim) returning the local id of the current thread.

local_size: grunnur.Module

Provides the function VSIZE_T ${local_size}(int dim) returning the size of the current group.

num_groups: grunnur.Module

Provides the function VSIZE_T ${num_groups}(int dim) returning the number of groups in dimension dim.

Utilities

class grunnur.Template(mako_template: mako.template.Template)

A wrapper for mako Template objects.

classmethod from_associated_file(filename: str) → Template

Returns a Template object created from the file which has the same name as filename and the extension .mako. Typically used in computation modules as Template.from_associated_file(__file__).

classmethod from_string(template_source: str)

Returns a Template object created from source.

get_def(name: str) → DefTemplate

Returns the template def with the name name.

class grunnur.DefTemplate(name: str, mako_def_template: mako.template.DefTemplate, source: str)

A wrapper for Mako DefTemplate objects.

classmethod from_callable(name: str, callable_obj: Callable[[…], str]) → DefTemplate

Creates a template def from a callable returning a string. The parameter list of the callable is used to create the pararameter list of the resulting template def; the callable should return the body of a Mako template def regardless of the arguments it receives.

classmethod from_string(name: str, argnames: Iterable[str], source: str) → DefTemplate

Creates a template def from a string with its body and a list of argument names.

render(*args, **globals_) → str

Renders the template def with given arguments and globals.

class grunnur.RenderError(exception: Exception, args: tuple, globals_: dict, source: str)

A custom wrapper for Mako template render errors, to facilitate debugging.

exception: Exception

The original exception thrown by Mako’s render().

globals: dict

The globals used to render the template.

source: str

The source of the template.

class grunnur.Snippet(template: DefTemplate, render_globals: Mapping = {})

Contains a source snippet - a template function that will be rendered in place, with possible context that can include other Snippet or Module objects.

Creates a snippet out of a prepared template.

Parameters

• template –

• render_globals –

classmethod from_callable(callable_obj: Callable[[…], str], name: str = '_snippet', render_globals: Mapping = {}) → Snippet

Creates a snippet from a callable returning a string. The parameter list of the callable is used to create the pararameter list of the resulting template def; the callable should return the body of a Mako template def regardless of the arguments it receives.

Parameters

• callable_obj – a callable returning the template source.

• name – the snippet’s name (will simplify debugging)

• render_globals – a dictionary of “globals” to be used when rendering the template.

classmethod from_string(source: str, name: str = '_snippet', render_globals: Mapping = {}) → Snippet

Creates a snippet from a template source, treated as a body of a template def with no arguments.

Parameters

• source – a string with the template source.

• name – the snippet’s name (will simplify debugging)

• render_globals – a dictionary of “globals” to be used when rendering the template.

class grunnur.Module(template: DefTemplate, render_globals: Mapping = {})

Contains a source module - a template function that will be rendered at root level, and the place where it was called will receive its unique identifier (prefix), which is used to prefix all module’s functions, types and macros in the global namespace.

Creates a module out of a prepared template.

Parameters

• template –

• render_globals –

classmethod from_callable(callable_obj: Callable[[…], str], name: str = '_module', render_globals: Mapping = {}) → Module

Creates a module from a callable returning a string. The parameter list of the callable is used to create the pararameter list of the resulting template def; the callable should return the body of a Mako template def regardless of the arguments it receives.

The prefix will be passed as the first argument to the template def on render.

Parameters

• callable_obj – a callable returning the template source.

• name – the module’s name (will simplify debugging)

• render_globals – a dictionary of “globals” to be used when rendering the template.

classmethod from_string(source: str, name: str = '_module', render_globals: Mapping = {}) → Module

Creates a module from a template source, treated as a body of a template def with a single argument (prefix).

Parameters

• source – a string with the template source.

• name – the module’s name (will simplify debugging)

• render_globals – a dictionary of “globals” to be used when rendering the template.

Data type utilities

C interop

grunnur.dtypes.ctype(dtype: numpy.dtype) → Union[str, grunnur.Module]

Returns an object that can be passed as a global to Program() and used to render a C equivalent of the given numpy dtype. If there is a built-in C equivalent, the object is just a string with the type name; otherwise it is a Module object containing the corresponding struct declaration.

Note

If dtype is a struct type, it needs to be aligned (see ctype_struct() and align()).

Parameters

dtype –

grunnur.dtypes.ctype_struct(dtype: Union[Type, numpy.dtype], ignore_alignment: bool = False) → grunnur.Module

For a struct type, returns a Module object with the typedef of a struct corresponding to the given dtype (with its name set to the module prefix).

The structure definition includes the alignment required to produce field offsets specified in dtype; therefore, dtype must be either a simple type, or have proper offsets and dtypes (the ones that can be reporoduced in C using explicit alignment attributes, but without additional padding) and the attribute isalignedstruct == True. An aligned dtype can be produced either by standard means (aligned flag in numpy.dtype constructor and explicit offsets and itemsizes), or created out of an arbitrary dtype with the help of align().

If ignore_alignment is True, all of the above is ignored. The C structures produced will not have any explicit alignment modifiers. As a result, the the field offsets of dtype may differ from the ones chosen by the compiler.

Modules are cached, and the function returns a single module instance for equal dtype’s. Therefore inside a kernel it will be rendered with the same prefix everywhere it is used. This results in a behavior characteristic for a structural type system, same as for the basic dtype-ctype conversion.

Parameters

• dtype –

• ignore_alignment –

Warning

As of numpy 1.8, the isalignedstruct attribute is not enough to ensure a mapping between a dtype and a C struct with only the fields that are present in the dtype. Therefore, ctype_struct will make some additional checks and raise ValueError if it is not the case.

grunnur.dtypes.complex_ctr(dtype: numpy.dtype) → str

Returns name of the constructor for the given dtype.

Parameters

dtype –

grunnur.dtypes.c_constant(val, dtype: Optional[numpy.dtype] = None) → str

Returns a C-style numerical constant. If val has a struct dtype, the generated constant will have the form { ... } and can be used as an initializer for a variable.

Parameters

• val –

• dtype –

grunnur.dtypes.align(dtype: numpy.dtype) → numpy.dtype

Returns a new struct dtype with the field offsets changed to the ones a compiler would use (without being given any explicit alignment qualifiers). Ignores all existing explicit itemsizes and offsets.

Parameters

dtype –

Struct helpers

grunnur.dtypes.c_path(path: List[Union[str, int]]) → str

Returns a string corresponding to the path to a struct element in C. The path is the sequence of field names/array indices returned from flatten_dtype().

Parameters

path –

grunnur.dtypes.flatten_dtype(dtype: numpy.dtype) → List[Tuple[List[Union[str, int]], numpy.dtype]]

Returns a list of tuples (path, dtype) for each of the basic dtypes in a (possibly nested) dtype. path is a list of field names/array indices leading to the corresponding element.

Parameters

dtype –

grunnur.dtypes.extract_field(arr: numpy.ndarray, path: List[Union[str, int]]) → numpy.ndarray

Extracts an element from an array of struct dtype. The path is the sequence of field names/array indices returned from flatten_dtype().

Parameters

• arr –

• path –

Data type checks and conversions

grunnur.dtypes.normalize_type(dtype: Union[Type, numpy.dtype]) → numpy.dtype

Numpy’s dtype shortcuts (e.g. numpy.int32) are type objects and have slightly different properties from actual numpy.dtype objects. This function converts the former to numpy.dtype and keeps the latter unchanged.

Parameters

dtype –

grunnur.dtypes.is_complex(dtype: numpy.dtype) → bool

Returns True if dtype is complex.

Parameters

dtype –

grunnur.dtypes.is_double(dtype: numpy.dtype) → bool

Returns True if dtype is double precision floating point.

Parameters

dtype –

grunnur.dtypes.is_integer(dtype: numpy.dtype) → bool

Returns True if dtype is an integer.

Parameters

dtype –

grunnur.dtypes.is_real(dtype: numpy.dtype) → bool

Returns True if dtype is a real number (but not complex).

Parameters

dtype –

grunnur.dtypes.result_type(*dtypes: numpy.dtype) → numpy.dtype

Wrapper for numpy.result_type() which takes into account types supported by GPUs.

Parameters

dtypes –

grunnur.dtypes.min_scalar_type(val, force_signed: bool = False) → numpy.dtype

Wrapper for numpy.min_scalar_dtype() which takes into account types supported by GPUs.

If force_signed is True, a signed type will be returned even if val is positive.

grunnur.dtypes.detect_type(val) → numpy.dtype

Returns the data type of val.

grunnur.dtypes.complex_for(dtype: numpy.dtype) → numpy.dtype

Returns complex dtype corresponding to given floating point dtype.

Parameters

dtype –

grunnur.dtypes.real_for(dtype: numpy.dtype) → numpy.dtype

Returns floating point dtype corresponding to given complex dtype.

Parameters

dtype –

grunnur.dtypes.cast(dtype: numpy.dtype) → Callable[[Any], Any]

Returns function that takes one argument and casts it to dtype.

Parameters

dtype –

Function modules

This module contains Module factories which are used to compensate for the lack of complex number operations in OpenCL, and the lack of C++ synthax which would allow one to write them.

grunnur.functions.add(*in_dtypes: numpy.dtype, out_dtype: Optional[numpy.dtype] = None) → grunnur.Module

Returns a Module with a function of len(in_dtypes) arguments that adds values of types in_dtypes. If out_dtype is given, it will be set as a return type for this function.

This is necessary since on some platforms complex numbers are based on 2-vectors, and therefore the + operator for a complex and a real number works in an unexpected way (returning (a.x + b, a.y + b) instead of (a.x + b, a.y)).

Parameters

• in_dtypes –

• out_dtype –

grunnur.functions.cast(in_dtype: numpy.dtype, out_dtype: numpy.dtype) → grunnur.Module

Returns a Module with a function of one argument that casts values of in_dtype to out_dtype.

Parameters

• in_dtype –

• out_dtype –

grunnur.functions.conj(dtype: numpy.dtype) → grunnur.Module

Returns a Module with a function of one argument that conjugates the value of type dtype (if it is not a complex data type, the value will not be modified).

Parameters

dtype –

grunnur.functions.div(dividend_dtype: numpy.dtype, divisor_dtype: numpy.dtype, out_dtype: Optional[numpy.dtype] = None) → grunnur.Module

Returns a Module with a function of two arguments that divides a value of type dividend_dtype by a value of type divisor_dtype. If out_dtype is given, it will be set as a return type for this function.

Parameters

• dividend_dtype –

• divisor_dtype –

• out_dtype –

grunnur.functions.exp(dtype: numpy.dtype) → grunnur.Module

Returns a Module with a function of one argument that exponentiates the value of type dtype (must be a real or a complex data type).

Parameters

dtype –

grunnur.functions.mul(*in_dtypes: numpy.dtype, out_dtype: Optional[numpy.dtype] = None) → grunnur.Module

Returns a Module with a function of len(in_dtypes) arguments that multiplies values of types in_dtypes. If out_dtype is given, it will be set as a return type for this function.

Parameters

• in_dtypes –

• out_dtype –

grunnur.functions.norm(dtype: numpy.dtype) → grunnur.Module

Returns a Module with a function of one argument that returns the 2-norm of the value of type dtype (product by the complex conjugate if the value is complex, square otherwise).

Parameters

dtype –

grunnur.functions.polar(dtype: numpy.dtype) → grunnur.Module

Returns a Module with a function of two arguments that returns the complex-valued rho * exp(i * theta) for values rho, theta of type dtype (must be a real data type).

Parameters

dtype –

grunnur.functions.polar_unit(dtype: numpy.dtype) → grunnur.Module

Returns a Module with a function of one argument that returns a complex number exp(i * theta) == (cos(theta), sin(theta)) for a value theta of type dtype (must be a real data type).

Parameters

dtype –

grunnur.functions.pow(base_dtype: numpy.dtype, exponent_dtype: Optional[numpy.dtype] = None, out_dtype: Optional[numpy.dtype] = None) → grunnur.Module

Returns a Module with a function of two arguments that raises the first argument of type base_dtype to the power of the second argument of type exponent_dtype (an integer or real data type).

If exponent_dtype or out_dtype are not given, they default to base_dtype. If base_dtype is not the same as out_dtype, the input is cast to out_dtype before exponentiation. If exponent_dtype is real, but both base_dtype and out_dtype are integer, a ValueError is raised.

Parameters

• base_dtype –

• exponent_dtype –

• out_dtype –

Virtual buffers

Often one needs temporary buffers that are only used in one place in the code, but used many times. Allocating them each time they are used may involve too much overhead; allocating real buffers and storing them increases the program’s memory requirements. A possible middle ground is using virtual allocations, where several of them can use the samy physical allocation. The virtual allocation manager will make sure that two virtual buffers that are used simultaneously (as declared by the user) will not share the same physical space.

class grunnur.virtual_alloc.VirtualManager(context: grunnur.Context)

Base class for a manager of virtual allocations.

Parameters

context – an instance of Context.

allocator(dependencies=None) → grunnur.virtual_alloc.VirtualAllocator

Create a callable to use for Array creation.

Parameters

dependencies – can be a Array instance (the ones containing persistent allocations will be ignored), an iterable with valid values, or an object with the attribute __virtual_allocations__ which is a valid value (the last two will be processed recursively).

pack(queue)

Packs the real allocations possibly reducing total memory usage. This process can be slow and may synchronize the base queue.

statistics() → VirtualAllocationStatistics

Returns allocation statistics.

class grunnur.virtual_alloc.TrivialManager(*args, **kwds)

Trivial manager — allocates a separate buffer for each allocation request.

class grunnur.virtual_alloc.ZeroOffsetManager(*args, **kwds)

Tries to assign several allocation requests to a single real allocation, if dependencies allow that. All virtual allocations start from the beginning of real allocations.

class grunnur.virtual_alloc.VirtualAllocator(manager: VirtualManager, dependencies: Set[int])

A helper callable object to use as an allocator for Array creation. Encapsulates the dependencies (as identifiers, doesn’t hold references for actual objects).

class grunnur.virtual_alloc.VirtualAllocationStatistics

Virtual allocation details.

real_num: int

The number of physical allocations.

real_size_total: int

The total size of physical allocations (in bytes).

real_sizes: Dict[int, int]

A dictionary size: count with the counts for physical allocations of each size.

virtual_num: int

The number of virtual allocations.

virtual_size_total: int

The total size of virtual allocations (in bytes).

virtual_sizes: Dict[int, int]

A dictionary size: count with the counts for virtual allocations of each size.

Kernel toolbox

There is a set of macros attached to any kernel depending on the API it is being compiled for:

GRUNNUR_CUDA_API

If defined, specifies that the kernel is being compiled for CUDA API.

GRUNNUR_OPENCL_API

If defined, specifies that the kernel is being compiled for CUDA API.

GRUNNUR_FAST_MATH

If defined, specifies that the compilation for this kernel was requested with fast_math == True.

LOCAL_BARRIER

Synchronizes threads inside a block.

FUNCTION

Modifier for a device-only function declaration.

KERNEL

Modifier for a kernel function declaration.

GLOBAL_MEM

Modifier for a global memory pointer argument.

LOCAL_MEM_DECL

Modifier for a statically allocated local memory variable.

LOCAL_MEM_DYNAMIC

Modifier for a dynamically allocated local memory variable (CUDA only).

LOCAL_MEM

Modifier for a local memory argument in device-only functions.

CONSTANT_MEM_DECL

Modifier for a statically allocated constant memory variable.

CONSTANT_MEM

Modifier for a constant memory argument in device-only functions.

INLINE

Modifier for inline functions.

SIZE_T

The type of local/global IDs and sizes. Equal to unsigned int for CUDA, and size_t for OpenCL (which can be 32- or 64-bit unsigned integer, depending on the device).

SIZE_T get_local_id(unsigned int dim)

SIZE_T get_group_id(unsigned int dim)

SIZE_T get_global_id(unsigned int dim)

SIZE_T get_local_size(unsigned int dim)

SIZE_T get_num_groups(unsigned int dim)

SIZE_T get_global_size(unsigned int dim)

Local, group and global identifiers and sizes. In case of CUDA mimic the behavior of corresponding OpenCL functions.

VSIZE_T

The type of local/global IDs in the virtual grid. It is separate from SIZE_T because the former is intended to be equivalent to what the backend is using, while VSIZE_T is a separate type and can be made larger than SIZE_T in the future if necessary.

ALIGN(int)

Used to specify an explicit alignment (in bytes) for fields in structures, as

typedef struct {
    char ALIGN(4) a;
    int b;
} MY_STRUCT;

Version history

0.2.0 (10 Mar 2021)

• (CHANGED) Arrays don’t hold queues any more; they are passed explicitly to get() or set().

• (CHANGED) Prepared kernels don’t hold queues any more; they are passed on call.

• (CHANGED) Queue now stands for a single-device queue only; multi-device queues are extracted into MultiQueue.

• (ADDED) MultiArray to simplify simultaneous kernel execution on multiple devices.

0.1.1 (9 Oct 2020)

Package build fixed.

0.1.0 (9 Oct 2020)

Initial version

Indices and tables

• Index

• Module Index

• Search Page