You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
We all (should) know that power/energy is best controlled at the system level, ie. in how applications interact. But there's not much guidance on how developers should write software for energy efficiency. One team in Sweden are developing tools to support this analysis and have demonstrated promising results on analysis of data-streaming applications. We'll still need good power management at the hardware level - this should help software designers take better advantage of those features.
It seems to me energy and speed efficient software is possible and can even replace
FPGAs with standard various core CPU type ICs, but I think the key is to provide
tools for people to write the assembly code, i.e. high level assmblers maybe
like the old PL/360, plus flow graph code generators to generate first drafts,
and power measurement or maybe simulation tools. I must admit that
when I mentioned it to some VC types, they laughed at me. Key is that
FPGA design is already 2 state not ASIC style.
I see that as the problem with the US economy. VCs will invest in EDA tools that they
can sell to the big 3 EDA companies, but not ideas that will create new companies. I think
this is the reason there are no more DECs and Hewlett-Packard HP that set up
good engineers in a separate division. The US economy could use some DEC lateral
management instead of a H1B economy.
If only that were true. Unfortunately the level of interest VCs have these days in anything related to EDA is indiscernible. That said, there is some EDA activity around hot domains like security (eg. Tortuga logic)
