mjendruk / mammut

Highspeed-Gravity Racer for Game Programmig Course at HPI
Other
4 stars 0 forks source link

fps drop when hitting a cuboid #94

Open karyon opened 10 years ago

karyon commented 10 years ago

when falling onto and hitting a cuboid, i get this:

total: 16.97 game: 0.08 geom: 0.89 ssao: 5.84 mb: 4.12 fxaa: 0.62 blit: 0.15
total: 16.75 game: 0.07 geom: 0.74 ssao: 6.77 mb: 4.10 fxaa: 0.61 blit: 0.15
total: 17.80 game: 0.09 geom: 0.79 ssao: 7.26 mb: 4.11 fxaa: 0.66 blit: 0.15
total: 19.11 game: 0.10 geom: 0.90 ssao: 9.90 mb: 4.12 fxaa: 0.57 blit: 0.15 <- hit
total: 21.73 game: 0.10 geom: 0.74 ssao: 15.92 mb: 4.23 fxaa: 0.62 blit: 0.15
total: 32.74 game: 0.09 geom: 0.75 ssao: 15.14 mb: 4.16 fxaa: 0.64 blit: 0.15
total: 31.83 game: 0.12 geom: 0.78 ssao: 13.17 mb: 4.16 fxaa: 0.56 blit: 0.15
total: 28.91 game: 0.11 geom: 0.89 ssao: 9.90 mb: 4.18 fxaa: 0.57 blit: 0.15
total: 15.81 game: 0.19 geom: 0.87 ssao: 8.25 mb: 4.18 fxaa: 0.62 blit: 0.15
total: 14.66 game: 0.10 geom: 0.74 ssao: 6.71 mb: 4.16 fxaa: 0.58 blit: 0.15
total: 15.86 game: 0.10 geom: 0.84 ssao: 6.13 mb: 4.11 fxaa: 0.59 blit: 0.15
total: 16.85 game: 0.09 geom: 0.87 ssao: 6.02 mb: 4.11 fxaa: 0.60 blit: 0.15 

i disabled the blurring and got the same results, reducing SSAO iterations by half halved both the usual and the spike times, so i think it's the loop in the ssao shader. does anyone know any reasons why a shader with no early outs or short paths can need such substantially different amounts of time?