Sign up for our newsletter - Navigating the horizon of business technology​
Technology / Hardware

Quick fix for bit-rot found in new Helium programme

Scientists at MIT have created a computer system called Helium which can quickly fine tune old codes, solving "a billion-dollar problem" faced by software companies known as bit-rot.

Designed by students at MIT, the programme can scan the binary codes called "stencil kernels" that serve as building blocks for more complicated algorithms.

After finding the old and rotted kernels, the system replaces them with newly optimised components, saving software developers time in optimising their software to a new hardware environment.

The programme claimed to address "a billion-dollar problem" as companies like Adobe used to invest a lot of man power and money in order to re-visit their code every few years, manually testing out a bunch of different strategies to try to patch it.

White papers from our partners

MIT professor and Computer Science and Artificial Intelligence Laboratory (CSAIL) researcher Saman Amarasinghe said: "For high-performance code used for image-processing, you have to optimise the heck out of the software.

"The downside is that the code becomes much less effective and much more difficult to understand."

Helium can revamp and fine-tune the computer code faster without requiring the original source code.

With the help of the system, the researchers can lift the "stencil kernels," from the popular software from a stripped binary and restructure them and make them readable in Halide, a programming language designed by CSAIL for image-processing.

The research paper of was presented at the Association for Computing Machinery SIGPLAN conference on Programming Language Design and Implementation (PLDI 2015) held last month in Portland.

Lead author of the paper Charith Mendis said: "Going from binary to high-level languages was a big leap that the team originally didn’t think was doable.

"The order of operations in these optimised binaries is complicated, which means that they can be hard to disentangle.

"Because stencils do the same computation over and over again, we are able to accumulate enough data to recover the original algorithms."

University of Utah’s School of Computing professor Mary Hall said: "We are in an era where computer architectures are changing at a dramatic rate, which makes it important to write code that can work on multiple platforms.

"Helium is an interesting approach that has the potential to facilitate higher-level descriptions of stencil computations that could then be more easily ported to future architectures."
This article is from the CBROnline archive: some formatting and images may not be present.

CBR Staff Writer

CBR Online legacy content.