Views: 6,854,662 | Homepage | Main | Rules/FAQ | Memberlist | Active users | Last posts | Calendar | Stats | Online users | Search | 04-19-24 03:46 AM |
Guest: |
0 users reading GPU rasterization notes | 1 bot |
Main - Development - GPU rasterization notes | Hide post layouts | New reply |
Arisotura |
| ||
Big fire melon magical melon girl Level: 56 Posts: 95/884 EXP: 1334775 Next: 63401 Since: 03-28-17 From: France Last post: 18 hours ago Last view: 5 hours ago |
* we know from edge slopes, that the X offset is calculated with a reciprocal and not a 'proper' division. for example: slope with dx=2 and dy=96. x isn't incremented until y=49 (would be 48 with proper division).
* trying to figure out the reciprocal. its precision, how it's calculated (proper 1/X division? Newton-Raphson? LUT?) * Z interpolation along Y loses precision for whatever reason. Z range 0-0x800000 loses 14 bits, 0-0x400000 loses 13 bits, 0-0x200000 loses 12 bits and so on. except: 0-0x200 still loses 2 bits. weird. * Z interpolation along X seems to have full precision. (losing two bits, but I think fog is losing them) ____________________ Kuribo64 |
Main - Development - GPU rasterization notes | Hide post layouts | New reply |
Page rendered in 0.028 seconds. (2048KB of memory used) MySQL - queries: 28, rows: 75/75, time: 0.017 seconds. Acmlmboard 2.064 (2018-07-20) © 2005-2008 Acmlm, Xkeeper, blackhole89 et al. |