Cheating? and Speed Gains… Correlated?

Modern computers are fast but usually developers overload them with all the stuff they can find. I know I do that. And then they start to slow down again. On this topic I’ve found two interesting stories lately:

  • The first one is about Browser speeds. It seems that IE 9 might be “cheating” the SunSpider Javascript test. It could be that IE9 was only doing deadcode analysis before trying to run it a la Haskell, but some tests showed that it might not be the case. Go figure!
  • The other news has to do with speeding up the Linux Desktop, by a small change (~200 lines) in the kernel scheduler source code. The rationality behind this patch is that groups tasks according the TTY they are attached to. Some videos showed that this new approach can increase dramatically the perceived responsiveness of the system.

So, why have I put this two items in one post? The main reason is that both have to do with the PERCEIVED computational speed of pieces of software. In the first case, IE9 did some tweeking to it’s Javascript so it would look good in benchmarks (after all the reading this is my opinion), and in the second case the tweeking to the kernel scheduler was done to look good for the user. This two thing are at different morality levels, but in the end they show something:

(PERCEIVED) SPEED MATTERS! A LOT!