All CK components can be found at cKnowledge.io and in one GitHub repository!
This project is hosted by the cTuning foundation.
Introduction
Optimization results to demonstrate compiler autotuning, crowd-tuning and machine learning on RPi3 via customizable Collective Knowledge workflow framework with a portable package manager.
- arXiv report: http://arxiv.org/abs/1801.08024
- interactive CK report: http://cknowledge.org/rpi-crowd-tuning
License
- CC BY 4.0
Prerequisites
- Collective Knowledge framework (@GitHub)
- Python 2.7 or 3.3+
- Python PIP
- Git client
Minimal CK installation
The minimal installation requires:
- Python 2.7 or 3.3+ (limitation is mainly due to unitests)
- Git command line client.
You can install CK in your local user space as follows:
$ git clone http://github.com/ctuning/ck
$ export PATH=$PWD/ck/bin:$PATH
$ export PYTHONPATH=$PWD/ck:$PYTHONPATH
You can also install CK via PIP with sudo to avoid setting up environment variables yourself:
$ sudo pip install ck
CK repository installation
Install the CK repository:
$ ck pull repo --url=https://github.com/dividiti/ck-rpi-optimization
Update all CK repositories at any time
$ ck pull all
Check out report and see related scripts in the following entries:
$ ck ls script:rpi3-*
For example, you can see individual scripts we used to prepare, run and reproduce autotuning experiments via CK for susan corners benchmark in the following entry:
$ cd `ck find:scriptrpi3-susan-autotune`
$ ls
Two CK repositories with additional experimental results in a reproducible form are available at FigShare:
You can download and install them directly via CK as follows (note that each zip is around 150Mb archived and ~1-1.5GB unzipped):
$ ck add repo:ck-rpi-optimization-results-reactions --zip=https://ndownloader.figshare.com/files/10218435 --quiet
$ ck add repo:ck-rpi-optimization-results-reactions-multiple-datasets --zip=https://ndownloader.figshare.com/files/10218441 --quiet
$ ck ls experiment:rpi3-*
We continue gradually documenting all scripts in above entry together with the community - your help is appreciated. Feel free to get in touch with the community via CK mailing list:
Next steps:
- We plan to use reproducible optimization methodology prototyped here to support Pareto-efficient co-design competitions of the whole software and hardware stack for emerging workloads such as deep learning in terms of speed, accuracy, energy and costs: http://cKnowledge.org/request
Notes
We could not build GCC 7.1.0 for RPi3 via CK with Graphite support (outdated libraries and missing deps). This may reduce optimization possibilities during autotuning:
gcc -c -I../ -DCK_HOST_OS_NAME2_LINUX=1 -DCK_HOST_OS_NAME_LINUX=1 -DCK_TARGET_OS_NAME2_LINUX=1 -DCK_TARGET_OS_NAME_LINUX=1 -DXOPENME -I/home/fursin/CK-TOOLS/lib-rtl-xopenme-0.3-gcc-4.9.2-linux-32/include -O3 -fcaller-saves -fcse-follow-jumps -fgcse-lm -fno-gcse-sm -fira-share-save-slots -fno-ira-share-spill-slots -floop-interchange -flto -fmodulo-sched-allow-regmoves -fpeephole -fsched-spec -freciprocal-math -fno-sched-spec-load-dangerous -fselective-scheduling2 -fsel-sched-pipelining-outer-loops -fsignaling-nans -fsplit-ivs-in-unroller -ftree-dominator-opts -fno-tree-fre -ftree-loop-distribute-patterns -ftree-ter ../adler32.c -o adler32.o
../adler32.c:1:0:
sorry, unimplemented: Graphite loop optimizations cannot be used
(isl is not available) (-fgraphite, -fgraphite-identity,
-floop-nest-optimize, -floop-parallelize-all)