the hot new programming language

...or at least the politician proposing it.

Even better.

The 4004 was slow enough!

I do, often, and call them "breadboards." But Spice has advantages of its own.

Powerbasic:

A$ = B$ + C$

is safe, or, if you need it,

A$ = LEFT$(B$, 16)

works too. And there are tons of other crashproof string functions.

On Sat, 04 Jul 2015 07:20:38 -0700, John Larkin Gave us:

It was also false high-false low error prone slow TTL too.

ANY optical computer at ANY word length will beat ANY silicon machine. Interesting that you missed that aspect.

On Sat, 04 Jul 2015 07:22:22 -0700, John Larkin Gave us:

A breadboard is not an analog computer. The operator of the breadboard who characterizes its behavior is.

Interestingly, that isn't the case. A colleague of mine summed up the main reason: "You can control many electrons with one electron, but you can't control many photons with one photon."

IOW optical logic devices don't have gain. They also generally aren't as fast as transistors--SiGe and InP transistors work up to over 100 GHz.

The switches have to be at least a few wavelengths in size in order to be able to confine the fields, and you can pack a _lot_ of transistors into that space.

Cheers

Phil Hobbs

Ssschhhllllooooowww and opaque by comparison to

strncpy(dest, src, bufsize-1); dest[bufsize-1] = '\0';

But there's no defending the design of strncpy(). Breaking the null-termination rule is just plain stupid.

Cheers

Phil Hobbs

C++ has crashproof strings too: std::string, and they have semantics like that too. They just aren't stdio. The null-termination thing is a minor wart. There are a few things like that strncpy() fail and the fact that you have to remember to free() things created with strdup(), but the output functions of stdio are easy to use.

The parsing functions, especially the scanf() family, are a mess, but as I say I have my own stdio-compatible parsing library that doesn't share the same problems.

Cheers

Phil Hobbs

There are all kinds of 3D structures. The problem is cooling them. For instance, say you're stacking a processor and several planes of memory. The processor generates a lot of heat, so it has to go next to the heat sink, i.e. at the top of the stack. but then all its I/O has to go through the memory chips, so you lose all your area to through-silicon vias (TSVs). Same problem as very tall buildings.

If you put it the other way up, you have to throttle back the CPU to the point that you don't gain anything. Computer speed has been a tradeoff between clock rate and cooling since the 1980s. I remember going to a talk by a system architect in about 1988, where he put up a plot of delay vs. power consumption per gate. It dropped steeply at first, of course but then gradually rose again at high powers, because the chips had to be spaced out in order to cool them, which added time-of-flight delay.

Cheers

Phil Hobbs

You are mistakenly thinking one photon = one transistor.

No, I'm not. Electrons stay where they're put, in general, steering drain current until you put them someplace else. A photon passes by in a picosecond and is lost.

Trust me, I spent seven years in silicon photonics trying to do stuff like that.

Cheers

Phil Hobbs

Aren't the human brain and body 3-D structures? I wonder how they can do what they do?

That assumes a constant in process technology which we all know advances steadily at least if not at the same exponential rates it has been achieving.

You should have studied biology. ;)

No, it doesn't. There are fundamental physical limits involved, like the size of atoms and the conductivity of pure copper. Just the random variations in the local density of dopant atoms causes huge threshold-voltage shifts. Process improvements can help lots of things, but you can't make smaller atoms.

And the brain hardly has the same speed-of-light limits as fast silicon.

Cheers

Phil Hobbs

In 1965 Moore observed that the density of integrated circuits was doubling every year with the most recent observation at a density of 2^6 or about 60 devices, so early days. Then before the term "Moore's Law" was in use, in 1975 he revised his observation to a doubling every two years. The "law" has remained at that state since.

Here is a Moore paper where he discusses this and various aspects of the observation. An interesting read.

None of which is relevant as we are nowhere near those limitations and that is not what we are discussing.

How is that relevant?

I used this to model an antenna system once.

You haven't been keeping up. Copper conductivity has been the main limiting factor in interconnect speed and density for a decade or more--that's why there was all that work in low-K dielectrics, latterly air.

The electric permittivity of vacuum is another of those limits.

And threshold voltage shifts due to fluctuations in local dopant density have been known to be a problem for about that long.

You brought up the brain, not me.

Cheers

Phil Hobbs