4 ms·

Fortran predictably overflows the result, in contrast to fptoui which gives a poison value. I agree that even a predictable overflow can bite you on the behind.

Fortran predictably overflows the result, in contrast to fptoui which gives a poison value. I agree that even a predictable overflow can bite you on the behind. But it's better than undefined behavior. I'm not a fan of saturated cast, but having a defined behavior is for sure a great improvement.

nitrogen 6y ago
When working with audio as an example, saturation is much less obnoxious than wraparound. You can potentially destroy speakers and your hearing that way.
- pjmlp 6y ago
  Wraparound is also the bug that turns Ghandi into a huge dictator in Civilization, and killed a couple of people in cancer treatment devices.
  - nindalf 6y ago
    For those not in the know, computer controlled Civs would be assigned behaviour attributes based on historical personality. Each attribute would be 1-10, stored in an 8 bit integer. These attributes change in response to events in the game. When a civilization would reach the modern era, it's aggression attribute would be reduced by 2. Gandhi's aggression was reduced from 1 to 255 (underflow). He would rain down nukes on everything in sight, somewhat uncharacteristic of him. Fans liked this so much that Firaxis maintained the same behaviour in later games.
    - paavohtl 6y ago
      It's a cool story, but I don't think there is evidence that it actually happened (in Civ 1, certainly in later games as a homage to this potentially urban legend). The YouTube channel People Make Games did an investigation[1] about this a year ago, including interviews with the original developers. [1] https://www.youtube.com/watch?v=Ur3SdgkW8W4 https://www.youtube.com/watch?v=Ur3SdgkW8W4
      - nindalf 6y ago
        TIL that I am fake news.
- sjwright 6y ago
  Why would you want math errors to be less obnoxious?
  - baseballdork 6y ago
    I think they just prefer them to be less destructive.
    - sjwright 6y ago
      “Destructive”, regardless how minor, should be intentional or at the very least conscious. This is one of the various reasons why serious audio processing is done as 32 bit float: no (practical) risk of unexpected overflows.
  - frabert 6y ago
    They're not math errors, it's the most sensible thing to do: when you overload a speaker, you get clipping, not some weird thing where the cone snaps to the opposite side of the range.
    - sjwright 6y ago
      I’m talking about the basis of the analogy, not the analogy. If you’re processing audio signals, saturation (i.e. clipping) is of course preferable to the alternative.
- tialaramex 6y ago
  Saturated addition is indeed the correct way to add two PCM streams together. Especially back in the 1980s and 1990s you'd get awful code that did things like averaging the two streams because wrapping sounds awful and the authors were ignorant of the theory and/or unaware that saturated addition is a thing. You can tell when somebody did this because it means playing silence makes everything else quieter, or worse there's an arbitrary limit on how many streams are played and playing any one thing is very quiet because it's attenuated so as to never wrap. Haiku the 1990s-style operating system did this for years, as did various Amiga music software.
  - diffuse_l 6y ago
    At least on x86, saturated adds didn't become a thing until MMX was released. Doing saturated adds was simply too expensive on older machines, and usually required a branch.
jabl 6y ago
> Fortran predictably overflows the result Not sure what you mean here, and I don't have the standard at hand ATM, but I'm quite sure this is undefined behaviour in Fortran. But yes, I agree defined behaviour is good. Undefined behaviour is occasionally good for optimization, at the cost of gray hairs for users.
- milancurcic 6y ago
  Standard? True, undefined. Compilers, however, happily overflow unless you tell them to catch it at run time. So, my original statement is at least half false. :)

nitrogen 6y ago

When working with audio as an example, saturation is much less obnoxious than wraparound. You can potentially destroy speakers and your hearing that way.

pjmlp 6y ago
Wraparound is also the bug that turns Ghandi into a huge dictator in Civilization, and killed a couple of people in cancer treatment devices.
- nindalf 6y ago
  For those not in the know, computer controlled Civs would be assigned behaviour attributes based on historical personality. Each attribute would be 1-10, stored in an 8 bit integer. These attributes change in response to events in the game. When a civilization would reach the modern era, it's aggression attribute would be reduced by 2. Gandhi's aggression was reduced from 1 to 255 (underflow). He would rain down nukes on everything in sight, somewhat uncharacteristic of him. Fans liked this so much that Firaxis maintained the same behaviour in later games.
  - paavohtl 6y ago
    It's a cool story, but I don't think there is evidence that it actually happened (in Civ 1, certainly in later games as a homage to this potentially urban legend). The YouTube channel People Make Games did an investigation[1] about this a year ago, including interviews with the original developers. [1] https://www.youtube.com/watch?v=Ur3SdgkW8W4 https://www.youtube.com/watch?v=Ur3SdgkW8W4
    - nindalf 6y ago
      TIL that I am fake news.
sjwright 6y ago
Why would you want math errors to be less obnoxious?
- baseballdork 6y ago
  I think they just prefer them to be less destructive.
  - sjwright 6y ago
    “Destructive”, regardless how minor, should be intentional or at the very least conscious. This is one of the various reasons why serious audio processing is done as 32 bit float: no (practical) risk of unexpected overflows.
- frabert 6y ago
  They're not math errors, it's the most sensible thing to do: when you overload a speaker, you get clipping, not some weird thing where the cone snaps to the opposite side of the range.
  - sjwright 6y ago
    I’m talking about the basis of the analogy, not the analogy. If you’re processing audio signals, saturation (i.e. clipping) is of course preferable to the alternative.
tialaramex 6y ago
Saturated addition is indeed the correct way to add two PCM streams together. Especially back in the 1980s and 1990s you'd get awful code that did things like averaging the two streams because wrapping sounds awful and the authors were ignorant of the theory and/or unaware that saturated addition is a thing. You can tell when somebody did this because it means playing silence makes everything else quieter, or worse there's an arbitrary limit on how many streams are played and playing any one thing is very quiet because it's attenuated so as to never wrap. Haiku the 1990s-style operating system did this for years, as did various Amiga music software.
- diffuse_l 6y ago
  At least on x86, saturated adds didn't become a thing until MMX was released. Doing saturated adds was simply too expensive on older machines, and usually required a branch.

jabl 6y ago

> Fortran predictably overflows the result Not sure what you mean here, and I don't have the standard at hand ATM, but I'm quite sure this is undefined behaviour in Fortran. But yes, I agree defined behaviour is good. Undefined behaviour is occasionally good for optimization, at the cost of gray hairs for users.

milancurcic 6y ago
Standard? True, undefined. Compilers, however, happily overflow unless you tell them to catch it at run time. So, my original statement is at least half false. :)