PDA

View Full Version : Test my physics simulator



Marco81bs
06-26-2006, 04:18 AM
Hi everyone, I'd like you to test my physics simulator of deformable bodies (by a mass-spring system): it's an university project. It REQUIRES an nVidia card: a GeForceFX (I used PBO) or a better one (faster simulation thanks to FLOAT16 Blending).
Visit: www.webalice.it/altomonte (http://www.webalice.it/altomonte)
Thanks!

k_szczech
06-26-2006, 06:09 AM
Athlon64 3200+
GeForce 7800GT
1024MB / 800MHz
Windows XP with latest beta drivers.

MassSpring.exe:
starts at 50 FPS and grows up to >70 FPS (I guess it's average framerate and first frame executes longer).

MassSpring BLEND16.exe
Hits software path somewhere and runs very slow (one frame per 9 seconds).

My GPU supports 16-bit floating-point blending. Make sure your'e not using 32-bit float blending by mistake. I'm using fp16 blend and filtering in my game and it runs very fast.

Marco81bs
06-26-2006, 07:19 AM
You're right: I started some stability test on my GeForceFX and I forgot to restore 16bit blending.
I uploaded a new version, and www.webalice.it/altomonte/MassSpringUpdate.rar (http://www.webalice.it/altomonte/MassSpringUpdate.rar) with the 2 executables only.
I corrected the FPS counter, but I'm more interested in the simulation time: is it shorter than 1 msec?

k_szczech
06-26-2006, 07:28 AM
It works fine now.
Unfortunately there is something wrong with your simulation time counter. In both applications it can only show 3 values:
0.0000000
1.5000000
1.6000000
Take a look at QueryPerformanceCounter function in Platform SDK - perhaps you should give it a try.
Fps counter shows 80 FPS in NV30 and 81-82 FPS in NV40.

Marco81bs
06-26-2006, 07:58 AM
I used the "clock" function to simplify multiplatform development, but I didn't notice that CLOCK_PER_SEC constant is SO different in Linux and Win. Another little update ( www.webalice.it/altomonte/MassSpringUpdate.rar (http://www.webalice.it/altomonte/MassSpringUpdate.rar) ) with QueryXXX.

k_szczech
06-26-2006, 08:22 AM
I think you misspelled the link or put wrong executables in the archive, because I'm seeing the same thing.