The following content has been provided by the University of Erlangen-Nürnberg.

I think it's a danger to have such a title that people think they learn something new.

For example, when the summer is coming and so on.

But it will be partly also more specific.

So this is not working for one reason or the other.

Which if she uses to work, then I have to do it by hand.

No, if she uses to work.

Oh, now it's doing something.

Okay, so what I would like to do, I would like to...

The microphone.

Ah, the microphone.

Okay, this is all the part for the next centuries.

Okay, so what I would like to do is to say a little bit about temperature, okay?

So we worked since not so long time, maybe two years or so, a little bit on temperature.

I was interested what is the relation between temperature and computing,

and I would like to share a few ideas we had and what we learned about temperature,

and then also how we embedded this into some design flow that may be similar to,

or at least mapping algorithms onto multiprocessors,

and how we can use the knowledge we gained here.

Okay, so I think that's known to everyone, so it's not a power wall we're facing,

it's a temperature wall, and that's the reason why processes are getting slower,

and you have more of them, and therefore we thought well,

and also on power and energy lots of people have been doing work,

and we thought doing something on temperature,

and we learned that temperature is of course different than power and energy, of course,

and we tried to understand what is actually the difference.

In the context is mainly we are in several projects dealing with mapping algorithms onto multiprocessors,

something on 3D electronics where you stack all the chips and have some liquid cooling inside,

where you have microchannels that are trying to cool things,

and we also like many others working on some platforms with a few more processors.

Okay, so we got interested in the question on temperature.

Okay, well in order to do that we tried to get also some results that are where you can prove something,

and you get really an idea of what happens between computation and temperature,

so we are specifically doing, did something on the relation between time and temperature, okay,

because if you are faster then usually will get hotter, and if you are slower it's cooler,

and therefore there is some sort of trade-off,

and we wanted to understand what this trade-off actually is.

So thermal and timing properties and objectives of course must be considered simultaneously,

and this is what we actually tried to understand.

And we started with some questions that at least at that time surprisingly for me didn't have really an answer, okay,

and these are very basic and fundamental questions.

For example, people who are doing something with timing and real-time systems,

they know there is this idea of a critical instance, okay,

there is somehow the sequence of events that arrive so that things in terms of timing are worst, okay,

and of course if you talk about temperature it would be interesting to know,

well, what is the worst thing that can happen so that it gets really hot, okay,

and there was no answer to this question.

So somehow what is the sequence of events and things that can happen so that things are really getting hot,

and therefore this is something we tried to understand,

and related to this is how you can simultaneously guarantee timing and temperature.