Monday, January 28, 2008

the superscalar brain

Back in the days of my education, when I was learning how to create microprocessors that were superscalar (roughly: able to dispatch two or more instructions within the same clock to disparate parts of the same CPU - achieving better than one-instruction-per-clock throughput) I can remember thinking about how the brain functions in a superscalar manner as well.

See - I'm kinda old. Not old in the sense of what modern medicine considers old -- I have quite some time before I reach that threshhold -- but measured with Moore's Law, I'm hella old indeed.

This was in the days before multiple core processors. Hell, this was before the days of the much-vaunted Intel Hyperthreaded CPUs. You had one CPU and you had one pipeline into it. However, if you were very clever, you could cheat a little bit. The pipeline into the CPU was wider than many of your instructions needed. You could therefore create a new instruction, a SUPER instruction that was two instructions concatenated together. This super instruction would then also pass data to two separate and functionally isolated parts of the same SINGLE CPU. The FPU and ALU could both be active in the same clock, so it seemed more efficient to try and use them in parallel.

This is the heart of the superscalar architecture and how we were able to design in more throughput than you could get from a single clock normally. This is why AMD Athlons had numbers like 1800+ even though they ran at clock speeds far below 1.8Ghz, which was what the competing Intel clock ran at.

But I digress...

I bring this up because today I'm reading an article (literally.. as I write this, I'm reading it in a different pane, I haven't finished it yet) and listening to some music and monitoring the stock ticker to see when I'm going to be rich enough to buy a house, and checking my email inbox and waiting for an IM to tell me when lunch will be happening and I realize that I have only one brain.

And I'm abusing the hell out of it, using some of those same kind of tricks.

Superscalar thinking is addictive, but not very effective. You can do it, but at the expense of not doing other things very well.

But I remember a time, not all that long ago, when I was not old, when I lived in the countryside and hadn't yet been spoiled by the good food and constant entertainment that the city provides, and when I could _concentrate_ on things.

My ability to concentrate has been damaged. I'm still good at all of the things that I've spent time becoming good at, but I'm not quite as good, and new things are hard. Some new things, like trying to change the way I react to situations, or how I think about certain things, or how I perceive things, are very, VERY hard.

This is the article I'm reading:

Normally I'd wait until I finished it to blog about it, but this seems particularly apropos.

I'll let you know later what I think of it when I'm done.

No comments: