Will Software follow, or even save Moore’s Law?

Software has not kept up with Moore’s Law, and I did not get any disagreement. But will it in the future? Are you bold enough to predict that software will save Moore’s Law? That is what Peder Ulander (VP of Sun Microsystems) and I discussed with the members of the press in a roundtable.

<iframe src="http://intelpr.feedroom.com/linking/index.jsp?skin=oneclip&fr_story=7ea288c7d6fd1e7acfda88d79c86debda4b0191d&rf=ev&hl=true ” width=”320″ height=”277″ scrolling=”no” frameborder=”0″ marginwidth=”0″ marginheight=”0″ >

Things were progressing happily, with processors doubling clock frequency and core complexity every generation, just to feed the voracious appetite of software. Not any more. Frequency and core complexity will barely inch along, but with lots of cores—welcome to the era of multi, and we are now at the mercy of the software to use it effectively. Software, too, now has to follow Moore’s Law. Would you depend on software to follow the law, let alone save it?

There was a lot of skepticism, and I am not surprised. Parallelizing software? Been there, done that, was the overwhelming response. But don’t forget that things are different now. First, there is a compelling reason—there is no other alternative but multi for affordable performance. Second, parallelism is plentiful, but you must go beyond thread level, to task level, and application level to harvest it. And third, there are tremendous advances in software technology in the last ten years. Don’t expect the change to happen over night, it will take years, but we must start now.

We need to look at software a lot differently, as Peder points out, and I completely agree. Look at it beyond a piece of code running on your computer, think about it as a service provider. Look at it as a whole system, integrated with the platform and the hardware. Start influencing Universities with the new outlook to educate future crop of software engineers.

Why go through this trouble, what is the killer app they ask! I think I’d rather attempt to predict the meaning of life before predicting the killer app! Killer apps evolve, they are not planned. You will know it only when it emerges and succeeds. Don’t try predicting it, it will lead to nowhere.