Bullet compiler options (MSCV2008)

[Pierre]

Is there any reason why the default compiler options for Bullet are, let's say, sub-optimal in Release mode? In particular, why not using the /arch::SSE or /arch::SSE2 flag?

I got a nice little speed boost by tweaking the compile flags, and it's a bit of a shame that people may not realize the default setup is not the 'fastest'.

Just wondering...

[Erwin Coumans]

Hi Pierre,

The cmake generated project files have SSE enabled, but the premake ones not yet.

I've been mainly experimenting with GPU/OpenCL in a different repostory (http://github.com/erwincoumans/experiments) and neglected the regular CPU version a bit.

I just enabled SSE for the premake projects, so the project files of the next release has it enabled:
https://code.google.com/p/bullet/source/detail?r=2624
(or simply click on build/vs2010.bat to generate them)

Thanks!
Erwin

[SynapticBytes]

I'm using bullet source code directly in a VS2010 project for my ShiVa plugin, and don't know a lot about setting compiler options. I've just run a speedtest comparison of 1,000 spheres in a matrix, falling onto a plan using the SHiVa integrated (ODE) engine vs Bullet via my plugin. Worst framerate dropped to about 18 in Bullet vs about 30 in ShiVa.

So what should I be looking for in the source code, or Visual Studio compile options to make sure I'm building the most optimal .dll?

Thanks.