[ODE] Some SSE in Quick step

[Ivan Bolcina]

Adam D. Moss wrote:

> Ivan Bolcina wrote:
>
>> I hear that MS had both Intel and AND experts helping write optimized 
>> code for Direc3DX,
>
>
>
> Okay, I hoped I was being clear enough, but:
>
> 1) I'm not specifically interested in vector (as in {X,Y,Z}) ops.
> I asked about generating vectorISED code.  As in, a general
> abstraction for SIMD.  I was hoping there'd be a semistandard set
> of cross-arch intrinsics by now, with the compiler just doing the
> Right Thing w.r.t. the actual target arch.
> 2) When I said 'cross-architecture' I didn't mean 'Intel and AMD'
> ('Oh, we got both kinds. We got country AND western!').
>
> --Adam


Hi,

What do you mean by first point.. something what HLSL or Cg are to GPUs 
you want to have for CPU?

Ok, but isnt it annoying when you integrate several libraries into one 
game, to have to convert matrices,quaternions,... from one to another... 
So my point was, that there should be some standard like OpenGL,OpenAL,  
and also OpenMATHLIBRARY... which could serve as mathematical foundation 
for similar libraries.

In Windows, this OpenMATHLIBRARY could be implemented using D3DX, on 
linux on something else.

Because it is strange to have something so fast,tested and free as D3DX 
and not use it,if it is there. So what if it is not open source. Also 
ATI or NVIDEA are not open source, but you are still using it,right
The only problem with using D3DX I see is the time-clocks the wrapper 
would eat up, but maybe the wrapper could be so thin, this would not 
represent any problems.


BTW,And I don't understand 'Oh, we got both kinds. We got country AND 
western!'. Is this some saying from America....

bye, ivan