mirror of
https://github.com/gentoo-mirror/gentoo.git
synced 2025-12-23 21:57:33 -08:00
30c6893d5c
Add CPU_FLAGS_* for amd64, arm, arm64, ppc64 and x86 to explicitly control minimal CPU intrinsic use. According to upstream, this should grant some performance benefit compared to the default use of dynamic dispatching based on CPU detection at runtime. Since upstream is heavily chaining flags one upon another, we respect upstream chains by stopping when a dependent flag is missing. Presumably, this should be less confusing than implicitly enabling it, and more user friendly than REQUIRED_USE. Also add USE=cpudetection to make it possible to disable dynamic dispatch for CPUs with more instruction sets. This should especially be useful when we are deliberately disabling some instruction sets to test NumPy without them. Signed-off-by: Michał Górny <mgorny@gentoo.org> Part-of: https://github.com/gentoo/gentoo/pull/43190 Closes: https://github.com/gentoo/gentoo/pull/43190 Signed-off-by: Michał Górny <mgorny@gentoo.org>
33 lines
1.2 KiB
XML
33 lines
1.2 KiB
XML
<?xml version="1.0" encoding="UTF-8"?>
|
|
<!DOCTYPE pkgmetadata SYSTEM "https://www.gentoo.org/dtd/metadata.dtd">
|
|
<pkgmetadata>
|
|
<maintainer type="project">
|
|
<email>sci@gentoo.org</email>
|
|
<name>Gentoo Science Project</name>
|
|
</maintainer>
|
|
<maintainer type="project">
|
|
<email>python@gentoo.org</email>
|
|
<name>Python</name>
|
|
</maintainer>
|
|
<longdescription lang="en">
|
|
NumPy is a general-purpose array-processing Python package designed to
|
|
efficiently manipulate large multi-dimensional arrays of arbitrary
|
|
records without sacrificing too much speed for small multi-dimensional
|
|
arrays. There are also basic facilities for discrete fourier transform,
|
|
basic linear algebra and random number generation.
|
|
It is the successor of Numeric and numarray.
|
|
</longdescription>
|
|
<upstream>
|
|
<remote-id type="github">numpy/numpy</remote-id>
|
|
<remote-id type="pypi">numpy</remote-id>
|
|
</upstream>
|
|
<use>
|
|
<flag name="cpudetection">
|
|
Enable dynamic dispatch to additional CPU extensions not covered
|
|
by enabled CPU_FLAGS_*. This permits NumPy to benefit from improved
|
|
performance when CPUs support more instruction sets, while preserving
|
|
compatibility with the baseline set by CPU_FLAGS_*.
|
|
</flag>
|
|
</use>
|
|
</pkgmetadata>
|