[ODE] SIMD ODE code?
tomyeh at CS.UCLA.EDU
Fri Feb 3 18:51:46 MST 2006
I used -mmmx -msse2 -msse -mfpmath=sse
I guess I should try -march=pentium4.
On Wed, 1 Feb 2006, Rodrigo Hernandez wrote:
> There is none that I know off, at least not an official one yet.
> dVector3 is defined as float, and I added the "--with-arch="
> parameter to configure in prevision for SIMD optimization, but
> nothing more is there.
> a good starting point towards SIMD integration would be modifying the
> macros in odemath.h to use SIMD intrinsic functions or inline asm and
> defining dVector3 and dVector4 as __attribute__ ((aligned (16))).
> Also, what flags are you using? in for a Pentium 4 you'd want to use
> -march=pentium4 |-mfpmat=sse and maybe throw in -malign-double.
> Thomas Yeh wrote:
> >Is there a SIMD optimized version of ODE somewhere?
> >When I compile ODE with the SIMD and MMX flags turned on for gcc, the
> >binaries generated seem to run slower on the P4 than the normal release
> >ODE mailing list
> >ODE at q12.org
More information about the ODE