[ODE] Denormal Floating Values

[Bodo Pfeifer]

>We round/truncate all values to a specified precision between each physics step
>(and we use a fixed time step). This is using a non-ODE physics engine, though,
>>but should be the same.

Yes .. thats what I meant by truncating the values. Didn't work that
much.

> Also, setting the floating point control word to use 64-bit registers instead of
> 80-bit registers will be necessary if you want similar operation across architectures.

VC offers a compilerswitch "/Op", that improves the consistency of
floating point operations .. but didn't improve the results.

Is there another option/switch to force 64bit registers?

> Unless you're using StepFast() -- I think StepFast() uses rand() internally
> and thus isn't necessarily consistent without extra work.

Yes, I'm using StepFast .. but I replaced the rand() to get a
deterministic behaviour.


I also tried to set ODE to double precision, this showed some little
improvements, but didn't solve the problem.


Thanks,
Bodo