1
$\begingroup$

Digital signal processors can represent irrational numbers using fixed-point or floating-point data types.

Fixed point means fixed number of decimal places, and floating point means floating number of decimal places.

There are advantages to each and it isn't always best to use a floating-point processor.

Still, where numerical precision matters, floating-point has a leg up, and I've been assuming a launch vehicle like the space shuttle or Saturn V would have used exactly that in its GNC systems.

But have I been assuming right? Or might it have been fixed-point instead?

Not a computer scientist. I know just enough about data types to get me in trouble. Please hold down your bazookas, kind people :D

Update

Just found the text below suggesting the space shuttle in fact used fixed-point arithmetic. Now wondering if modern rockets use fixed-point too... and can someone confirm that the excerpt below really means 480,000 instructions per second for fixed-point against 7,000 for floating-point? That difference is huuuge and unexpected to me. Big thanks if you can confirm!

enter image description here

$\endgroup$
17
  • $\begingroup$ That quote says the shuttle GPC used both. Where do you see 70,000? It quotes 7000 for the Gemini computer ... what are you asking? $\endgroup$ – Organic Marble Jun 6 at 0:53
  • 2
    $\begingroup$ This isn't an answer, but I can give an explanation of why floating point is so much slower. Consider addition: in fixed point, it works a lot like integer addition. Everything lines up nicely; you add the digits, handle any overflow, and you're done. In floating point, you basically have two numbers represented in (more or less) scientific notation. That means you first have to scale them so the exponents are the same before you can perform addition, then you have to normalize the result. Floating point ops are quite a few more steps. $\endgroup$ – Tristan Jun 7 at 14:39
  • 1
    $\begingroup$ @uhoh: there's more to the question than the bit I added in my update. Will close the question soon if no more info arrives, though :) $\endgroup$ – user39728 Jun 7 at 15:08
  • 1
    $\begingroup$ I'm not sure that "Digital Signal Processor" means what you think it means. "Digital signal processing" refers to numerical techniques for analyzing images, sounds, and other "continuous phennomena", and DSP hardware is computing machinery that is highly specialized for that task. The AP-101 that you cite was functionally a general-purpose computer, even though it was physically specialized for aerospace applications. $\endgroup$ – Solomon Slow Jun 7 at 18:54
  • 1
    $\begingroup$ @SolomonSlow Agreed! NASA even referred to it as the General Purpose Computer (GPC) nasa.gov/mission_pages/shuttle/flyout/… $\endgroup$ – Organic Marble Jun 7 at 20:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy