Is It Time to Move to Multicore?
Is it time to move to multicore? It's inevitable, but there are certainly many caveats.
I'm heading up a panel at the Multicore Expo that will examine the typical elements which make up the ROI analysis which all software teams should undertake when assessing migrating to a larger number of cores. How much parallelism is easily accessible in the existing software application? How much work will it be to unlock that concurrency? What are the risks of inserting new problems into the code during migration?
When I first started assembling this ROI panel topic, I wasn't really thinking desktop, server, or any other standard type of platform. I was thinking embedded. And with embedded, what exactly is the application that you're porting over to multicore? Automotive? Smartphone? Networking? Video? Signal processing? The list goes on, and each application unveils different challenges and opportunities for multicore developers.
Even within the application itself, there are many variations to consider. SMP or SoC? Dual-core or manycore? How about a mixture of SMP within a heterogeneous SoC? From my perspective, these are all forms of multicore. In short, a few weeks ago, I started a similar discussion on this ROI analysis topic on the LinkedIn Group: "Multicore & Parallel Computing." Based on the amount and type of feedback I received, I'll conclude that my panel discussion will only scratch the surface of this topic.