A Python library using the Futhark C backend via CFFI
Project description
futhark-ffi
Python library using the Futhark C backend via CFFI
Futhark provides several compiler backends, for example futhark opencl
which is a C backend, and futhark pyopencl
which is a Python
backend based on PyOpenCL. However, the host-side code of the Python
backend is quite slow, leading to a lot of overhead when small,
frequent kernels are used.
A solution to reduce this overhead is to use CFFI to used the C backend from Python, greatly reducing the calling overhead. The OpenCL code is the same, so this is not interesting for long-running kernels.
This library supports the following Futhark backends: c
, opencl
,
multicore
, and cuda
.
Futhark arrays are mapped to and from Numpy arrays. Multiple outputs and multi-dimensional arrays are supported.
Installation
Install Futhark, then simply
pip install futhark-ffi
Usage
Generate a C library, and build a Python binding for it
futhark opencl --library test.fut
build_futhark_ffi test
Use the Python wrapper
import numpy as np
import _test
from futhark_ffi import Futhark
test = Futhark(_test)
res = test.test3(np.arange(10))
test.from_futhark(res)
Example usage
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file futhark_ffi-0.15.0.tar.gz
.
File metadata
- Download URL: futhark_ffi-0.15.0.tar.gz
- Upload date:
- Size: 5.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.1 CPython/3.11.3
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 1279bcf7b3e1276eee1c79db28e62ca0f2dbd023d472f5e105cdad15869b6d61 |
|
MD5 | 4fcda2e60405d7d63c34ec0cf1aad32b |
|
BLAKE2b-256 | a738c6240d315acdc9ed50cf376d3ae9bab8f371e3a4b94c325d3790fd039e96 |
File details
Details for the file futhark_ffi-0.15.0-py3-none-any.whl
.
File metadata
- Download URL: futhark_ffi-0.15.0-py3-none-any.whl
- Upload date:
- Size: 6.5 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.1 CPython/3.11.3
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | e880d17e488842119956d36472e226b018b907d8c7336310c4c7ac821af0316c |
|
MD5 | cfdb2945ae051a074e619ffe10feaefb |
|
BLAKE2b-256 | 0aeadcce87c5c5fa7fe88aafbae12ff4667d4e405fd6b50798aff47bc76c12f9 |