| Re: [AD] Poor Ming performance |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
> if the screen is locked before calling the circle routine, then the cost
> of locking a (already locked) bitmap should be close to zero. in this
> case, hardware operations should be faster than software ones.
> problem solved i think :)
Yes, except that's exactly the opposite !
Once the screen is locked (once per primitive), the software operations (at
least hlines) are much faster than the hardware ones.
In the economical way: software rendering has a huge entry cost (locking)
but a very low marginal cost per hline after that, whereas hardware
rendering has no entry cost but a higher marginal cost per hline.
It's then obvious that the former solution is cheaper than the latter for
high volumes :-)
--
Eric Botcazou
ebotcazou@xxxxxxxxxx