File: control

package info (click to toggle)
gemmlowp 0.0~git20200506.fda83bd-1
  • links: PTS, VCS
  • area: main
  • in suites: bookworm, bullseye, sid
  • size: 5,724 kB
  • sloc: cpp: 113,766; ansic: 8,943; python: 3,251; sh: 79; objc: 55; makefile: 16
file content (32 lines) | stat: -rw-r--r-- 1,360 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Source: gemmlowp
Section: science
Priority: optional
Maintainer: Debian Science Maintainers <debian-science-maintainers@lists.alioth.debian.org>
Uploaders: Mo Zhou <lumin@debian.org>
Build-Depends: debhelper-compat (= 13),
               cmake,
Standards-Version: 4.5.0
Homepage: https://github.com/google/gemmlowp
Vcs-Browser: https://salsa.debian.org/science-team/gemmlowp
Vcs-Git: https://salsa.debian.org/science-team/gemmlowp.git

Package: libgemmlowp-dev
Section: libdevel
Architecture: any
Multi-Arch: foreign
Depends: ${misc:Depends},
#         ${shlibs:Depends}, # there is no binary file in this package.
Description: small self-contained low-precision GEMM library
 This is not a full linear algebra library, only a GEMM library: it only does
 general matrix multiplication ("GEMM").
 .
 Its performance goals differ from typical GEMM performance goals in the
 following ways:
 1. It cares not only about speed, but also about minimizing power usage.
    It specifically cares about charge usage in mobile/embedded devices.
 2. Most GEMMs are optimized primarily for large dense matrix sizes (>= 1000).
    It does care about large sizes, but it also cares specifically about the
    typically smaller matrix sizes encountered in various mobile applications.
 .
 Keep in mind (previous section) that gemmlowp itself is a pure-headers-only
 library.