Higher resolution timestamps

[olofson]

Given that even phones have 64 bit CPUs at this point, it might be time to switch to 64 bit timestamps, with higher sub-sample resolution. A 48:16 timestamp format would have a number of advantages over the current 24:8 format:

• Wrapping at 192 kHz would occur after 46 years, as opposed to every 87 seconds, so we can just ignore that, meaning...
  • No special care (macros) needed for timestamp arithmetic and comparisons
  • Long delays (event scheduling, envelope times etc) will no longer be problematic
• Accuracy should be enough for implementing granular/BLIT oscillators with no audible artifacts due to timing errors
• 64 bits is plenty of space for alternative timestamp formats, such as measure:beat:tick, SMPTE etc, perhaps with the format encoded in the high bits
  • On the downside, that would of course reintroduce the need for special care when comparing and operating on timestamps

Other formats may be of interest, for a bit of extra resolution, but the wrap time quickly becomes a real issue;

• 32:32 wraps ever 6.2 hours, which is problematic in common applications
• 40:24 wraps every 66 days, which would be a problem in embedded systems, arcade games and the like