Do you know the meaning of term software rot or bit-rot? It’s the gradual deterioration in the performance of a software and its responsiveness that makes it slow, outdated and unfit for use. Over the time, heaps of code gets accumulated and it becomes less difficult to understand due to hardware upgrade. For about past 12 months, MIT computer scientists and engineers from Adobe are working together to solve this software rot problem faced by all major companies.
MIT professor and a researcher at the Computer and Artificial Intelligence Lab (CSAIL) Saman Amarsinghe calls this “a billion-dollar problem” as companies have to test the code every few years manually and develop “healing” patches for it.
So, the genius minds at MIT and Adobe thought why not make a computer program that would automatically fix old codes.
CSAIL developed Helium, a system that repairs the code in few minutes or hours without needing any manpower or the original source. There’s one particular type of computational kernel popular in proprietary software like Photoshop – the “stencil kernels” – that allow operations on entire areas of pixels. They are the ones responsible for huge consumption resources, and as the environment changes due to the new hardware, their performance degrades.
Also read: Write 1000 Lines Code in Just 50 Lines with New Programming Language “Picture”
The Helium system replaces the rotten software components with optimized ones. It was found that Helium system could perform this task in one day that human engineers took three months to complete. As a result, some Photoshop filters saw 75 percent increase in performance, whereas, the performance of less optimized programs like Microsoft Windows’ IrfanView saw a jump of 400-500 percent.
According to experts, the Helium system is a new and innovative approach to software rot that addresses the high-level descriptions of stencils computations and it could be ported to machine architecture of future.
Comment down your views about Helium system developed to solve the bit-rot problem.
Source: MIT.edu
Want to learn to code? Get started here
