weird: chopping (and rounding) 3 or 4 bits off a double may be more accurate than 2 bits. 3 seems best most often. this was unexpected...
-
-
@RichFelker I don't actually know why or how. was skeptical at first, but it seems to hold up in my tests. seems odd in any case... -
@RichFelker best rounding biases in test: 2 bits=2, 3 bits=3, 4 bits=7. seems specific to use of rounding. mostly applicable to type-tagging -
@RichFelker ADD: follow-up test: for float, the 1 and 2 bit cases did best (for evaluating tagged vec2f: 2x 30b vs 2x 28b, float30 wins).
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.