I see a decline of close to 20% for adding mortality. The other big difference between my algorithm (and how I tested it) and the once you posted are that mine are not depending on the length of the buffer and are therefor more generic. The huge disadvantage in my testing is also the fact that I only use one single buffer, causing all 4095 values to be placed on the stack at once. And they all have to be removed when the test is done (something the non-mortalizing version skips).
Still the mortalizing version of my algorithm in C/XS is still 6 times faster than the uu version.
Thanks for the thread, it was fun to untangle.
In reply to Re^3: unpacking 6-bit values
by Tux
in thread unpacking 6-bit values
by BrowserUk
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |