Check the preview of 2nd version of this platform being developed by the open MLCommons taskforce on automation and reproducibility as a free, open-source and technology-agnostic on-prem platform.
Live report demonstrating CK-powered multi-objective and multi-dimensional autotuning
Grigori Fursin 1,2, Anton Lokhmotov 2
1 cTuning Foundation, France,   2 dividiti, UK

[ Optimization repository ] [ GCC ] [ LLVM ] [ MILEPOST optimization predictor ]
[ Andrew Ng's ML course ] [ Google ML crash course ] [ Facebook Field Guide to ML ]

This live report demonstrates how CK can help create researchers reproducible and interactive article from reusable components

Note: After GCC 4.9.x we changed parsing script, moved it to CK (see above link), and started parsing optimizations using gcc --help:optimizers, while taking parameters from params.def, hence small variation in number is possible.

Interactive graph of slambench v1.1 cpu autotuning

Reproduce/reuse/replay/discuss via CK (interactive graphs)

    /?/?   ● Samsung ChromeBook 2; Samsung EXYNOS5; ARM Cortex A15/A7; ARM Mali-T628; Ubuntu 12.04; GCC 4.9.2; video 640x480    ● Big green dot: Clang 3.6.0 -O3    ● Big red dot: GCC 4.9.2 -O3    ● Small blue dots: random GCC 4.9.2. optimization flags (with 10% probablitiy of a selection of a given flag    ● Small red dots: Pareto frontier

Non-interactive graph of image corner detection algorith (susan corners) autotuning

Reproduce/reuse/replay/discuss via CK (interactive graphs)

    /?/?   ● LENOVO X240; Intel(R) Core(TM) i5-4210U CPU @ 1.70GHz; Intel HD Graphics 4400; Windows 7 Pro    ● Image image-pgm-0001; 600x450; 270015 bytes /?   ● Black square dot: MingW GCC 4.9.2 -O3    ● Blue dots: random optimization flags (with 10% probablitiy of a selection of a given flag    ● Red line: Pareto frontier


Please log in to add your comments!
If you notice any inapropriate content that should not be here, please report us as soon as possible and we will try to remove it within 48 hours!