Data Parallel Extension for NumPy

https://intelpython.github.io/dpnp/
Admin
Data Parallel Extension for NumPy (DPNP) is a library that provides a data parallel extension for...
Tokens:1,094,037
Snippets:4,608
Trust Score:8.3
Update:9 months ago
Show doc for...
Tokens:
Raw
### Install dpnp using Pip

Source: https://intelpython.github.io/dpnp/quick_start_guide

Installs the dpnp package from Intel's PyPI repository using pip. This command fetches dpnp and its required components.

```bash
python -m pip install --index-url https://software.repos.intel.com/python/pypi dpnp
```

--------------------------------

### Build and Install dpnp with scikit-build (Windows)

Source: https://intelpython.github.io/dpnp/quick_start_guide

Builds and installs the dpnp package from source on Windows using scikit-build and Ninja. It specifies the Intel C/C++ compiler (icx).

```bash
python setup.py install -- -G Ninja -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icx
```

--------------------------------

### Create and Sum Array with dpnp

Source: https://intelpython.github.io/dpnp/quick_start_guide

Demonstrates creating a dpnp array and calculating the sum of its elements. It shows how to check the device allocation of the array and the resulting sum.

```python
import dpnp as np

x = np.asarray([1, 2, 3])
print("Array x allocated on the device:", x.device)

y = np.sum(x)

print("Result y is located on the device:", y.device)  # The same device as x
print("Shape of y is:", y.shape)  # 0-dimensional array
print("y =", y)  # Expect 6

```

--------------------------------

### DPNP arange Function Example

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_arraycreation

Demonstrates the basic usage of the dpnp.arange function to create arrays with evenly spaced values. Shows examples with default parameters and custom start, stop, and step values.

```python
import dpnp as np

# Example 1: Default start and step
print(np.arange(3))

# Example 2: Custom start and default step
print(np.arange(3, 7))

# Example 3: Custom start, stop, and step
print(np.arange(3, 7, 2))
```

--------------------------------

### Build and Install dpnp with scikit-build (Linux)

Source: https://intelpython.github.io/dpnp/quick_start_guide

Builds and installs the dpnp package from source on Linux using scikit-build and Ninja. It specifies the Intel C/C++ compilers (icx, icpx).

```bash
python setup.py install -- -G Ninja -DCMAKE_C_COMPILER:PATH=icx -DCMAKE_CXX_COMPILER:PATH=icpx
```

--------------------------------

### Example: Summing Array Elements in dpnp

Source: https://intelpython.github.io/dpnp/_sources/quick_start_guide.rst

Demonstrates a basic usage of dpnp for creating an array and calculating the sum of its elements. This example is part of the dpnp library's demonstration scripts.

```python
import dpnp

# Example code for summing array elements would go here
# (Actual code not provided in the source text, only reference to file)
```

--------------------------------

### Install dpnp using Conda

Source: https://intelpython.github.io/dpnp/quick_start_guide

Installs the dpnp package and its dependencies, including dpctl and necessary runtimes, from Intel's Conda channel. It also includes packages from conda-forge for compatibility.

```bash
conda install dpnp -c https://software.repos.intel.com/python/conda/ -c conda-forge --override-channels
```

--------------------------------

### Create Conda Build Environment

Source: https://intelpython.github.io/dpnp/quick_start_guide

Sets up a Conda environment named 'build-env' with the necessary tools for building Conda packages, specifically 'conda-build'.

```bash
conda create -n build-env conda-build
conda activate build-env
```

--------------------------------

### Build dpnp for AMD Devices

Source: https://intelpython.github.io/dpnp/quick_start_guide

Builds the dpnp library with support for AMD devices using the `--target-hip=<arch>` argument. The `<arch>` must be a valid architecture code for your AMD GPU, which can be found using `rocminfo`. Only one architecture can be specified at a time.

```bash
python scripts/build_locally.py --target-hip=<arch>

```

```bash
rocminfo | grep 'Name: *gfx.*'

```

--------------------------------

### Develop dpnp Locally using Driver Script (Linux)

Source: https://intelpython.github.io/dpnp/quick_start_guide

Uses a provided driver script to facilitate local development and building of the dpnp package on Linux systems.

```bash
python scripts/build_locally.py
```

--------------------------------

### Create Conda Build Environment for scikit-build (Windows)

Source: https://intelpython.github.io/dpnp/quick_start_guide

Creates a Conda environment named 'build-env' on Windows with all necessary development packages for building dpnp using scikit-build, including DPC++ components and Intel libraries.

```bash
conda create -n build-env dpctl cython dpcpp_win-64 mkl-devel-dpcpp tbb-devel \
      onedpl-devel cmake scikit-build ninja versioneer pytest intel-gpu-ocl-icd-system \
      -c dppy/label/dev -c https://software.repos.intel.com/python/conda/ -c conda-forge --override-channels
conda activate build-env
```

--------------------------------

### DPNP HammingFactory get() Method

Source: https://intelpython.github.io/dpnp/backend_doc/structdpnp_1_1extensions_1_1window_1_1kernels_1_1_hamming_factory

Documentation for the `get()` member function within the `HammingFactory` struct in the `dpnp::extensions::window::kernels` namespace.

```APIDOC
## GET /dpnp/extensions/window/kernels/HammingFactory/get

### Description
Retrieves an instance of the Hamming kernel function.

### Method
GET

### Endpoint
/dpnp/extensions/window/kernels/HammingFactory/get

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
None

### Response
#### Success Response (200)
- **fnT** (function_pointer) - A pointer to the Hamming kernel function.

#### Response Example
```json
{
  "kernel_function": "<pointer_to_hamming_kernel>"
}
```

### Source File
`/github/workspace/dpnp/backend/extensions/window/hamming.hpp`
```

--------------------------------

### Create Array on Specific Device Type with dpnp

Source: https://intelpython.github.io/dpnp/quick_start_guide

Illustrates creating a dpnp array on a specific device type, such as 'gpu', and how subsequent computations, like summation, inherit the device context. Includes error handling for cases where the specified device is unavailable.

```python
import dpnp as np

x = np.empty(3)
try:
    x = np.asarray([1, 2, 3], device="gpu")
except Exception:
    print("GPU device is not available")

print("Array x allocated on the device:", x.device)

y = np.sum(x)

print("Result y is located on the device:", y.device)  # The same device as x
print("Shape of y is:", y.shape)  # 0-dimensional array
print("y=", y)  # Expect 6

```

--------------------------------

### DPNP HanningFactory get() Method

Source: https://intelpython.github.io/dpnp/backend_doc/structdpnp_1_1extensions_1_1window_1_1kernels_1_1_hanning_factory

Provides access to the Hanning window kernel function.

```APIDOC
## GET /dpnp/extensions/window/kernels/HanningFactory/get

### Description
Retrieves the Hanning window kernel function. This is a factory method that returns the actual functor for generating Hanning windows.

### Method
GET

### Endpoint
/dpnp/extensions/window/kernels/HanningFactory/get

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```json
{
  "example": "No request body needed for this GET request."
}
```

### Response
#### Success Response (200)
- **fnT** (function pointer or functor type) - A callable object or function pointer that generates Hanning window values.

#### Response Example
```json
{
  "example": "// Returns a callable Hanning functor or function pointer"
}
```

### Error Handling
- **404**: If the endpoint is not found.
- **500**: Internal server error during kernel retrieval.
```

--------------------------------

### dpnp.asfortranarray Example

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.asfortranarray

Demonstrates how to use dpnp.asfortranarray to create a Fortran-contiguous array from a C-contiguous array. It also shows that calling the function on an already Fortran-contiguous array returns the same object. Examples include creating arrays on different devices or with specified USM types.

```python
>>> import dpnp as np

# Starting with a C-contiguous array:
>>> x = np.ones((2, 3), order='C')
>>> x.flags['C_CONTIGUOUS']
True

# Calling asfortranarray makes a Fortran-contiguous copy:
>>> y = np.asfortranarray(x)
>>> y.flags['F_CONTIGUOUS']
True
>>> x is y
False

# Now, starting with a Fortran-contiguous array:
>>> x = np.ones((2, 3), order='F')
>>> x.flags['F_CONTIGUOUS']
True

# Then, calling asfortranarray returns the same object:
>>> y = np.asfortranarray(x)
>>> x is y
True

# Creating an array on a different device or with a specified usm_type
>>> x0 = np.asarray([1, 2, 3])
>>> x = np.asfortranarray(x0) # default case
>>> x, x.device, x.usm_type
(array([1, 2, 3]), Device(level_zero:gpu:0), 'device')

>>> y = np.asfortranarray(x0, device="cpu")
>>> y, y.device, y.usm_type
(array([1, 2, 3]), Device(opencl:cpu:0), 'device')

>>> z = np.asfortranarray(x0, usm_type="host")
>>> z, z.device, z.usm_type
(array([1, 2, 3]), Device(level_zero:gpu:0), 'host')
```

--------------------------------

### Example: Array Computations on Specific Device in dpnp

Source: https://intelpython.github.io/dpnp/_sources/quick_start_guide.rst

Illustrates how to create an array on a specific device type and perform subsequent computations on it using dpnp. This showcases device-aware array operations.

```python
# Example code for device-specific array computations would go here
# (Actual code not provided in the source text, only reference to file)
```

--------------------------------

### Install Built Conda Package

Source: https://intelpython.github.io/dpnp/quick_start_guide

Installs a dpnp package that has been previously built locally using Conda. The '-c local' flag points to the local package repository.

```bash
conda install dpnp -c local
```

--------------------------------

### Utility - get_include

Source: https://intelpython.github.io/dpnp/reference/other

The `get_include` function returns the directory containing the *.h header files for the DPNP C++ backend.

```APIDOC
## GET /get_include

### Description
Return the directory that contains *.h header files of dpnp C++ backend.

### Method
GET

### Endpoint
/get_include

### Response
#### Success Response (200)
- **include_path** (string) - The path to the include directory.

#### Response Example
```json
{
  "include_path": "/path/to/dpnp/include"
}
```
```

--------------------------------

### DPNP C++ Backend Kernel Library - dpnp::extensions::blas::DotuContigFactory

Source: https://intelpython.github.io/dpnp/backend_doc/structdpnp_1_1extensions_1_1blas_1_1_dotu_contig_factory-members

Documentation for the `get()` method within the `DotuContigFactory` class, which is part of the DPNP C++ backend kernel library's BLAS extensions.

```APIDOC
## dpnp::extensions::blas::DotuContigFactory< fnT, varT >::get()

### Description
Retrieves an instance of the `DotuContigFactory`.

### Method
This documentation does not specify the HTTP method as it pertains to a C++ library function, not a web API endpoint.

### Endpoint
N/A (C++ library function)

### Parameters
This method does not appear to take any parameters.

### Request Example
N/A (C++ library function)

### Response
#### Success Response
- **Instance** (dpnp::extensions::blas::DotuContigFactory< fnT, varT >) - An instance of the `DotuContigFactory`.

#### Response Example
N/A (C++ library function output)
```

--------------------------------

### Get Upper Triangular Indices with Diagonal Offset using dpnp.triu_indices_from

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.triu_indices_from

This example shows how to use the `k` parameter in `dpnp.triu_indices_from` to retrieve indices for the upper triangular array starting from a specified diagonal offset. It demonstrates slicing the array using these indices.

```python
>>> import dpnp as np
>>> a = np.arange(16).reshape(4, 4)
>>> triuim1 = np.triu_indices_from(a, k=1)
>>> a[triuim1]
array([ 1,  2,  3,  6,  7, 11])
```

--------------------------------

### Create Array with Same Shape and Type (DPNP)

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_arraycreation

Demonstrates creating an array with the same shape and data type as a given array 'a' using dpnp.empty_like. Shows default behavior and how to specify 'cpu' device and 'host' USM type.

```python
>>> import dpnp
>>> import numpy as np

>>> a = np.array([1, 2, 3])

>>> x = np.empty_like(a) # default case
>>> x.shape, x.device, x.usm_type
((3, ), Device(level_zero:gpu:0), 'device')

>>> y = np.empty_like(a, device="cpu")
>>> y.shape, y.device, y.usm_type
((3, ), Device(opencl:cpu:0), 'device')

>>> z = np.empty_like(a, usm_type="host")
>>> z.shape, z.device, z.usm_type
((3, ), Device(level_zero:gpu:0), 'host')
```

--------------------------------

### Create Zero Array with dpnp.zeros_like

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_arraycreation

Demonstrates how to create a new array filled with zeros, matching the shape and data type of an existing array using dpnp.zeros_like. It also shows how to specify the device and usm_type for the resulting array.

```python
import dpnp as np
x0 = np.arange(6)

# Default case
x = np.zeros_like(x0)
print(f"Array: {x}, Device: {x.device}, USM Type: {x.usm_type}")

# Specify CPU device
y = np.zeros_like(x0, device="cpu")
print(f"Array: {y}, Device: {y.device}, USM Type: {y.usm_type}")

# Specify host usm_type
z = np.zeros_like(x0, usm_type="host")
print(f"Array: {z}, Device: {z.device}, USM Type: {z.usm_type}")
```

--------------------------------

### Build dpnp for AMD Devices

Source: https://intelpython.github.io/dpnp/_sources/quick_start_guide.rst

Builds the dpnp library with support for AMD devices using the `--target-hip=<arch>` argument. The architecture must be specified, and only one can be provided at a time. The architecture code can be found using `rocminfo`.

```bash
python scripts/build_locally.py --target-hip=<arch>
python scripts/build_locally.py --target-hip=gfx90a
```

--------------------------------

### dpnp.argsort Example Usage

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.argsort

Demonstrates how to use dpnp.argsort to get the indices that sort a 1D array. This function is a direct equivalent to numpy.argsort.

```python
>>> import dpnp as np
>>> x = np.array([3, 1, 2])
>>> np.argsort(x)
array([1, 2, 0])
```

--------------------------------

### dpnp.asanyarray Example

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_arraycreation

Demonstrates the basic usage of dpnp.asanyarray to convert a list to a dpnp.ndarray. It also shows how to create an array on a different device or with a specified USM type.

```python
>>> import dpnp as np
>>> np.asanyarray([1, 2, 3])
array([1, 2, 3])

Creating an array on a different device or with a specified usm_type

>>> x = np.asanyarray([1, 2, 3]) # default case
>>> x, x.device, x.usm_type
(array([1, 2, 3]), Device(level_zero:gpu:0), 'device')

>>> y = np.asanyarray([1, 2, 3], device="cpu")
>>> y, y.device, y.usm_type
(array([1, 2, 3]), Device(opencl:cpu:0), 'device')

>>> z = np.asanyarray([1, 2, 3], usm_type="host")
>>> z, z.device, z.usm_type
(array([1, 2, 3]), Device(level_zero:gpu:0), 'host')
```

--------------------------------

### DPNP ndarray argsort Example

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.ndarray.argsort

Demonstrates how to use the dpnp.ndarray.argsort method to get the indices that would sort an array. This method is useful for reordering array elements based on their sorted positions.

```python
import dpnp as np
a = np.array([3, 1, 2])
a.argsort()

```

```python
import dpnp as np
a = np.array([[0, 3], [2, 2]])
a.argsort(axis=0)

```

--------------------------------

### Get Array Data

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_array

Retrieves the USM (Unified Shared Memory) object pointing to the start of the array's data allocation. This is useful for direct memory access or interoperability with other libraries that use USM.

```python
return dpm.create_data(self._array_obj)
```

--------------------------------

### dpnp.pad: Edge Padding Example

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_manipulation

Illustrates using dpnp.pad with the 'edge' mode to pad an array by repeating the edge values.

```python
import dpnp as np
a = np.array([1, 2, 3, 4, 5])
print(np.pad(a, (2, 3), 'edge'))
# Expected output: array([1, 1, 1, 2, 3, 4, 5, 5, 5, 5])
```

--------------------------------

### Extract Diagonals from 3D Array with dpnp.diagonal

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.diagonal

Illustrates extracting diagonals from a 3D DPNP array by specifying axes. This example shows how to get diagonals by selecting specific axes for sub-array formation.

```python
>>> a = np.arange(8).reshape(2, 2, 2)
>>> a
array([[[0, 1],
        [2, 3]],
       [[4, 5],
        [6, 7]]])
>>> a.diagonal(0,  # Main diagonals of two arrays created by skipping
...            0,  # across the outer(left)-most axis last and
...            1)  # the "middle" (row) axis first.
array([[0, 6],
       [1, 7]])
```

--------------------------------

### dpnp.pad: Constant Padding Example

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_manipulation

Demonstrates how to use dpnp.pad with the 'constant' mode to pad an array with specific constant values at the beginning and end.

```python
import dpnp as np
a = np.array([1, 2, 3, 4, 5])
print(np.pad(a, (2, 3), 'constant', constant_values=(4, 6)))
# Expected output: array([4, 4, 1, 2, 3, 4, 5, 6, 6, 6])
```

--------------------------------

### Partition Array (dpnp_partition_c)

Source: https://intelpython.github.io/dpnp/backend_doc/group___b_a_c_k_e_n_d___a_p_i

Returns a partitioned copy of an array based on a specified element index (kth). It takes SYCL queue reference, input arrays, an output array, the partitioning index, shape and dimension information, and dependency SYCL events.

```cpp
template<typename _DataType >
DPCTLSyclEventRef dpnp_partition_c  | ( | DPCTLSyclQueueRef | _q_ref_ ,   
|  | void * | _array_ ,   
|  | void * | _array2_ ,   
|  | void * | _result_ ,   
|  | const size_t | _kth_ ,   
|  | const shape_elem_type * | _shape_ ,   
|  | const size_t | _ndim_ ,   
|  | const DPCTLEventVectorRef | _dep_event_vec_ref_ )  
```

--------------------------------

### Create Array of Ones (dpnp_ones_like_c)

Source: https://intelpython.github.io/dpnp/backend_doc/group___b_a_c_k_e_n_d___a_p_i

Implements the ones_like function to create an output array filled with ones, matching the data type and size of an input. It requires a SYCL queue reference, an output buffer, the size of the output array, and a vector of dependency SYCL events.

```cpp
template<typename _DataType >
DPCTLSyclEventRef dpnp_ones_like_c  | ( | DPCTLSyclQueueRef | _q_ref_ ,   
|  | void * | _result_ ,   
|  | size_t | _size_ ,   
|  | const DPCTLEventVectorRef | _dep_event_vec_ref_ )  
```

--------------------------------

### dpnp.rollaxis Function Example

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.rollaxis

Demonstrates how to use the dpnp.rollaxis function to change the position of an array axis. It takes an input array, the axis to roll, and an optional start position. The function returns an array with the specified axis moved.

```python
import dpnp as np
a = np.ones((3, 4, 5, 6))
print(np.rollaxis(a, 3, 1).shape)
print(np.rollaxis(a, 2).shape)
print(np.rollaxis(a, 1, 4).shape)
```

--------------------------------

### Get dpnp.ndarray.size - Python

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.ndarray.size

Demonstrates how to retrieve the total number of elements in a DPNP array using the `.size` property. This is useful for understanding the scale of operations or memory requirements. Requires the `dpnp` library to be installed.

```python
>>> import dpnp as np
>>> x = np.zeros((3, 5, 2), dtype=np.complex64)
>>> x.size
30
```

--------------------------------

### DPNP arange Initialization and Device Info

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_arraycreation

Demonstrates the initialization of an array using dpnp.arange with a specified USM type ('host') and retrieves the array, its associated device, and USM type. This showcases basic array creation and property inspection within DPNP.

```python
>>> z = np.arange(3, usm_type="host")
>>> z, z.device, z.usm_type
(array([0, 1, 2]), Device(level_zero:gpu:0), 'host')
```

--------------------------------

### Build dpnp Conda Package from Source

Source: https://intelpython.github.io/dpnp/quick_start_guide

Builds the dpnp package from its source code within a Conda environment. It specifies Intel's Conda channel and conda-forge for dependencies.

```bash
conda build conda-recipe -c https://software.repos.intel.com/python/conda/ -c conda-forge --override-channels
```

--------------------------------

### RegistryWindow Class Documentation

Source: https://intelpython.github.io/dpnp/backend_doc/structstatistics_1_1sliding__window1d_1_1_registry_window

Documentation for the RegistryWindow class, including its template parameters and type aliases.

```APIDOC
## Type Alias: SizeT

### Description
Alias for the size type of the registry data.

### Template Parameters
* **T** - The data type.
* **Size** - The window size.

### Definition
`template<typename T , uint32_t Size>
using statistics::sliding_window1d::RegistryWindow< T, Size >::SizeT = typename RegistryData<T, Size>::SizeT;`

### File
`sliding_window1d.hpp`
```

--------------------------------

### Create Array with Full Value on Different Devices using dpnp.full_like

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.full_like

Illustrates how to use dpnp.full_like to create an array filled with a specified value on different devices (e.g., CPU) or with specific USM (Unified Shared Memory) types. This showcases the flexibility in managing data placement and memory for parallel operations.

```python
>>> import dpnp as np
>>> a = np.arange(6)
>>> x = np.full_like(a, 1) # default case
>>> x, x.device, x.usm_type
(array([1, 1, 1, 1, 1, 1]), Device(level_zero:gpu:0), 'device')

>>> y = np.full_like(a, 1, device="cpu")
>>> y, y.device, y.usm_type
(array([1, 1, 1, 1, 1, 1]), Device(opencl:cpu:0), 'device')

>>> z = np.full_like(a, 1, usm_type="host")
>>> z, z.device, z.usm_type
(array([1, 1, 1, 1, 1, 1]), Device(level_zero:gpu:0), 'host')
```

--------------------------------

### Run dpnp Python Test Suites

Source: https://intelpython.github.io/dpnp/quick_start_guide

Executes the available Python test suites for dpnp using the pytest framework. The `-s` flag allows for stdout and stderr to be printed during test execution.

```bash
pytest -s tests

```

--------------------------------

### Linear Algebra (`dpnp.scipy.linalg`)

Source: https://intelpython.github.io/dpnp/reference/scipy

Documentation for linear algebra routines within the SciPy-compatible subset of DPNP. This includes various decomposition methods.

```APIDOC
## Linear algebra (`dpnp.scipy.linalg`)

### Description
This subsection details the linear algebra routines available through `dpnp.scipy.linalg`. It includes support for various matrix decompositions.

### Method
N/A (This is a documentation section, not an API endpoint)

### Endpoint
N/A

### Parameters
N/A

### Request Example
N/A

### Response
N/A

```

--------------------------------

### Build dpnp for Multi-Target (CUDA and AMD)

Source: https://intelpython.github.io/dpnp/quick_start_guide

Enables support for both CUDA and AMD devices simultaneously when building dpnp. This is achieved by combining the `--target-cuda` and `--target-hip=<arch>` arguments in the build script.

```python
python scripts/build_locally.py --target-cuda --target-hip=gfx90a

```

--------------------------------

### Get Lower Triangular Indices with dpnp.tril_indices

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.tril_indices

This snippet demonstrates how to use dpnp.tril_indices to obtain indices for the lower triangular part of a 4x4 array. It shows how to generate indices starting from the main diagonal and two diagonals to the right. The output indices can be used for array indexing and assignment.

```python
import dpnp as np
il1 = np.tril_indices(4)
print(il1)
a = np.arange(16).reshape(4, 4)
print(a[il1])
a[il1] = -1
print(a)
il2 = np.tril_indices(4, 2)
a[il2] = -10
print(a)
```

--------------------------------

### Create Array of Zeros Like (dpnp_zeros_like_c)

Source: https://intelpython.github.io/dpnp/backend_doc/group___b_a_c_k_e_n_d___a_p_i

Implements the zeros_like function to create an output array filled with zeros, matching the data type and size of an input. It requires a SYCL queue reference, an output buffer, the size of the output array, and a vector of dependency SYCL events.

```cpp
template<typename _DataType >
DPCTLSyclEventRef dpnp_zeros_like_c  | ( | DPCTLSyclQueueRef | _q_ref_ ,   
|  | void * | _result_ ,   
|  | size_t | _size_ ,   
|  | const DPCTLEventVectorRef | _dep_event_vec_ref_ )  
```

--------------------------------

### DPNP reduce_hypot Example

Source: https://intelpython.github.io/dpnp/_modules/dpnp/dpnp_iface_trigonometric

Demonstrates the usage of dpnp.reduce_hypot to compute the hypotenuse of elements in an array. It also shows an equivalent calculation using np.sqrt and np.sum.

```python
import dpnp as np
a = np.ones(10)
print(np.reduce_hypot(a))
print(np.sqrt(np.sum(np.square(a))))
```

--------------------------------

### Create Conda Build Environment for scikit-build (Linux)

Source: https://intelpython.github.io/dpnp/quick_start_guide

Creates a Conda environment named 'build-env' on Linux with all necessary development packages for building dpnp using scikit-build, including DPC++ components and Intel libraries.

```bash
conda create -n build-env dpctl cython dpcpp_linux-64 mkl-devel-dpcpp tbb-devel \
      onedpl-devel cmake scikit-build ninja versioneer pytest intel-gpu-ocl-icd-system \
      -c dppy/label/dev -c https://software.repos.intel.com/python/conda/ -c conda-forge --override-channels
conda activate build-env
```

--------------------------------

### Get Real Part of Array - dpnp

Source: https://intelpython.github.io/dpnp/reference/generated/dpnp.ndarray.real

This snippet demonstrates how to access the real part of a complex-valued array using the `.real` property in dpnp. It requires importing dpnp as np and creating a complex array, for example, by taking the square root of complex numbers.

```python
import dpnp as np
x = np.sqrt(np.array([1+0j, 0+1j]))
x.real
```
Tokens:
Raw