https://github.com/marshallward/optiflop
Optiflop measures the optimally achievable FLOPs for mathematical operations on various platforms.
https://github.com/marshallward/optiflop
avx avx2 avx512 cuda roofline vectorization x86
Last synced: 29 days ago
JSON representation
Optiflop measures the optimally achievable FLOPs for mathematical operations on various platforms.
- Host: GitHub
- URL: https://github.com/marshallward/optiflop
- Owner: marshallward
- Created: 2016-07-14T04:30:55.000Z (almost 10 years ago)
- Default Branch: main
- Last Pushed: 2024-09-12T14:44:47.000Z (over 1 year ago)
- Last Synced: 2024-09-13T03:06:09.464Z (over 1 year ago)
- Topics: avx, avx2, avx512, cuda, roofline, vectorization, x86
- Language: C
- Homepage:
- Size: 619 KB
- Stars: 12
- Watchers: 4
- Forks: 2
- Open Issues: 5
-
Metadata Files:
- Readme: README.rst
Awesome Lists containing this project
README
========
Optiflop
========
A tool for measuring peak performance and roofline analysis.
.. figure:: doc/readme_flops.svg
.. figure:: doc/readme_roofline.svg
About Optiflop
==============
Optiflop is a program for measuring the peak computational performance
(in FLOPs per second) and memory bandwidth on a compute node. Tests are
written in C with a goal of portability.
Quick usage guide
=================
This is still a development branch, so the build process requires an extra step
or two.
1. Generate the ``configure`` script and ``Makefile``, then ``make``.::
$ autoreconf
$ ./configure
$ make
To run for a default vector size (3200)::
./optiflop
For further options, ``optiflop --help``.