concure, put meat behind comments some ties I agree with you sometimes i don't, but no meat to learn or teach to.
show reasoning so as to be educated or to educate
we arent shlouches here, perhaps "lazy" (reference lazy article)
There is no doubt that there will be more microprocessors in the world in 2020 than today. Lower power consumption CPU will be in high demand than ever. However, I can't agree to the 1986 technology of compilers. Compilers improve a lot since 1986. Binary generated today run more efficiently and with smaller footprint.
Agile is a software development method. HTML5 is hardly a programming language. Software engineers and developers know what UML is.
There are 2 schools of thought. On one hand, there is no future for hardware. Software will be leading the industry in the foreseeable future. On the other hand, hardware is the future. It is leading the software. I believe neither. Rather, I think hardware and software go hand in hand. Only if you understand both, you will be able to bring the best system by utilizing the potential of both worlds.
>> On the other hand, hardware is the future. It is leading the software. I believe neither. Rather, I think hardware and software go hand in hand
The problem is that software makes a better business now. It is easier to distribute and sell than hardware. You innovate faster with lower risks. Making things is hard and risky. And if you have to choose, go with software. Yet, you need hardware to run software. The good news is that you may not have to do it yourself.
Processors just based on high performance is not the only road ahead. We used to think of Processor for a computer only, 10-15years back. But every device has a processor now. In years ahead, the amount of diversity we will see in the devices is far more. We will see all combinations of performance, power, cost etc based on type of devices it will get into..
When it comes to a low level world of software, there is not much change definitely. It is still the same old C/assembly language, compilers. May be there no need to improve beyond, else it would have definitely seen a change..
for non wake sleep applications show the combination graph of delay*energy with 1/3 and 2/3 weighting of one vs the other. This will give a good gut feel for power and delay in server and super computer use profile applications. Also would be useful to see it with delayed log vs no log graph.
Can this low headroom subthreshold be made dynamic depending on system level (think SNR)or weighted by distance from activity?
So when just your timing island is powered up and no nearby activity (its the only thing on) it could be lowerd just above the operating point.
Power mode change-other section wakes up dynamically and premtivly raise hedroom on subthreshold operation.
Yes I know its patentable but you IC companies won't even look at my resume causE I don't hAVE ic EXPERIENCE. so this one is now open source