In 2015, we released the 4th version of our cTuning repository - a brand new, open-source, customizable Collective Knowledge Repository. It aggregates all our past developments, ideas and techniques, and allows users to share, cross-link and reference any object and knowledge (workloads, data sets, tools, optimization results, predictive models, etc.) as a reusable component with a unified JSON API via GitHub or BitBucket!
You can easily check out, reuse and improve the following workloads/artifacts and modules shared by the community! You can also find various results from experiment crowdsourcing in the CK live repository and even participate in crowd-tuning yourself.
With our background in physics and AI, and a practical need to get efficient computer systems to solve our tasks, we are shocked by the current state of computer engineering. It is rare to find publications with reproducible and statistically meaningful results, systematized knowledge, open-source data and tools. One of the main reason is that sharing of data and tools simply doesn't pay off while the main focus is to publish as many papers as possible and to share as little data or tools as possible to avoid competition. However, since our aim to build efficient computer systems in terms of performance, power and reliability is to continue innovation in science, since 1996 we started a long-term and painful process of systematization of knowledge about design and optimization of computer systems through collaborative open-source tools and repositories. The first public repository have been used to enable machine learning self-tuning compiler - users ('crowd') have been searching effective predictive models to explain the behavior of computer systems preserved in the cTuning database. These models have been integrated with the MILEPOST GCC compiler or could be used as plugins for dynamic program adaptation (UNIDAPT framework). Eventually, we initiated new research methodology and publication model where experimental results (data, tools, models) are continuously shared and validated by the community. More information is available here.
# | Year | CK | Description |
[R1] | 2017-01 - cur. | Open repository of reusable, customizable, portable and optimized AI artifacts to accelerate AI research and boost innovation
|
|
[R2] | 2014-11 - cur. | Collective Knowledge public repository (CK aka cTuning4) to continue improving whole experimental setup sharing (code, data, dependencies, experimental results, models) along with interactive articles
|
|
[R3] | 2011-09 - cur. | Collective Mind public repository (cM aka cTuning3) to start collaborative systematization of analysis, design and optimization of computer systems based on extensible public repositories of knowledge, crowdsourcing, online tuning and machine learning, and to initiate new publication model where all research artifacts are continuously shared, validated and extended by the community
|
|
[R4] | 2010-03 - 2011-08 | In-house Codelet Tuning Repository for Intel Exascale Lab (aka cTuning2) to decompose large applications into codelets for continuous characterization and tuning |
|
[R5] | 2006-01 - cur. | cTuning.org public repository (aka cTuning1) to start collaborative systematization of analysis, design and optimization of computer systems based on extensible public repositories of knowledge, crowdsourcing, online tuning and machine learning
|
|
[R6] | 2004-06 - 2006-06 | In-house collaborative optimization repository for research on multi-objective program and architecture autotuning and co-design combined with machine learning |
|
[R7] | 1999-02 - 2006-06 | In-house collaborative optimization repository for research on multi-objective program and architecture characterization, optimization and co-design with first experiments on predictive modeling |
|
[R8] | 1993-03 - 1999-02 | In-house Experimental Repository for research, development and experimentation on novel, semiconductor neural networks, and on providing unified access to HPC resources as a web service |