Information Technology News.


Scientists at MIT improve code optimisation for multi-core processors

Share on Twitter.

Sponsered ad: Get a Linux Enterprise server with 92 Gigs of RAM, 16 CPUs and 8 TB of storage at our liquidation sale. Only one left in stock.

Sponsered ad: Order the best SMTP service for your business. Guaranteed or your money back.

November 9, 2016

Scientists at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Stony Brook University say they've discovered a new method to automatically adapt dynamic optimisation code for multi-core processors, potentially eliminating the need for hand-coding of some tricky manoeuvres when designing new IT systems.

As explained by the CSAIL and in the whitepaper Deriving Divide-and-Conquer Dynamic Programming Algorithms using Solver-Aided Transformations, dynamic optimization or dynamic programming code is simply a new coding technique that breaks complex problems down into lots of little sub-issues, then stores the results of each sub-problem so they only need to be solved once.

However, the method often involves solving the sub-problems in parallel, a twist to which multi-core CPUs are already very well suited, the scientists asserted us.

But the kind of people who design dynamic programming tend to be domain experts. For its part, MIT mentions that biologists and economists who already possess some coding experience probably have not been trained in the nuances of coding for multi-core CPUs.

Or they might just not have time to code for multi-core CPUs. MIT says “hand-optimized, parallel version of a dynamic-programming algorithm is typically about ten times as long as the single-core version, and that the individual lines of code are more complex, making things a bit worse.”

The technique described in the whitepaper described as “Bellmania” addresses that complexity by simply automating the various process of optimizing dynamic programming algorithms for parallel operations on multi-core CPUs.

The paper says that overall tests of the technique produce “performance comparable to that of the best manually optimized code.”

You could say that this is good for those who use dynamic programming, but perhaps less so for system developers who currently make good money helping them to make their code more efficient.

The whitepaper is more optimistic however, with the authors claiming that their results offer hope that end-users with some good mathematical background will be able to utilize the system without the steep learning curve that is usually associated with proof assistants.

This can be a valuable tool for algorithms research, some say. Whatever the outcome is, it will be interesting to see at that speed (if any) at which the industry will be willing to tackle this new design method.

Source: The Massachusetts Institute of Technology.


Sponsered ad: Get a Linux Enterprise server with 92 Gigs of RAM, 16 CPUs and 8 TB of storage at our liquidation sale. Only one left in stock.

Sponsered ad: Order the best SMTP service for your business. Guaranteed or your money back.

Share on Twitter.

IT News Archives | Site Search | Advertise on IT Direction | Contact | Home

All logos, trade marks or service marks on this site are the property of their respective owners.

Sponsored by Sure Mail™, Avantex and
by Montreal Server Colocation.

       © IT Direction. All rights reserved.