Numpy Intel IPP signal.
Project description
Introduction
Numpy Intel IPP Signal is a Python wrapper for Intel IPP Signal functions.
Principles
Numpy Intel IPP Signal is based on a Setup and Payoff strategy
- Setup : First Numpy data buffers and Intel IPP Signal operations are setup, this step can be slow.
- Payoff : Then operations are executed as fast as possible with Intel IPP Signal or Numpy backend functions.
This strategy suits to deal with stream of data for example.
Example
# Two Numpy data buffers
src1 = numpy.ones(100, dtype=numpy.float32)
src2 = numpy.zeros(100, dtype=numpy.float32)
# A result buffer
dst = numpy.empty(100, dtype=numpy.float32)
# Intel IPP Signal Mul operation
mul = numpy_ipps.Mul(dtype=numpy.float32)
# Unpack Numpy buffer for fast access
with numpy_ipps.utils.context(src1, src2, dst):
mul(src1, src2, dst) # Fast multiplication: dst[n] <- src1[n] * src2[n]
Constructor parameters
Some parameters are often needed by the Setup step:
dtype
To properly select the Intel IPP Signal backend function, the type of the data has to be known.
size
For some operations, the size of the data has to be known to switch between Intel IPP Signal and Numpy backend.
accuracy
Most of Intel IPP Signal operations on float can be operated at three different levels.
List of operations
See more details at ReadTheDocs.io.
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
numpy_ipps-2.2.1.dev0.tar.gz
(62.3 MB
view hashes)