Re: MMX version of 4x4 Forward Transform

>       I'm trying to write a MMX version of this function. I write it,
> but I do more test before I commit it. It's faster than C++
> version:

    That's very interesting.  It it helps, each vertical and horizontal row/column is
calculated using the following equations:

    T0 = S0 + S3            // Sx: source data
    T1 = S1 + S2            // Tx: temporals
    T2 = S1 - S2            // Rx: result
    T3 = S0 - S3
    R0 = T0 + T1
    R2 = T0 - T1
    R1 = T2 + ( T3 << 1)
    R3 = T3 - ( T2 << 1)

    Someone commented that the transformation is already quite optimized, and little
gain could be achieved using SIMD.  From your timing results, the speed-up is small
( 1.24), but more detailed profiling should be done to assert that.

    Do you have experience on SSE/SEE2?  Could it be more usefull than MMX?

    /AITOR

> 
> - C++:        2250ms
> - MMX:        1810ms
> 
> This is time showed by openhdot264 when i enable the MMX code. BTW
> This code is not well optimalized... I work on it...
> 
> PS. Sorry for my terrible english....
> 
> -- 
> Best regards
>  monsti                          mailto:[email protected]
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by:
> The Definitive IT and Networking Event. Be There!
> NetWorld+Interop Las Vegas 2003 -- Register today!
> http://ads.sourceforge.net/cgi-bin/redirect.pl?keyn0001en
> _______________________________________________
> Hdot264-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/hdot264-devel
> 



-------------------------------------------------------
This SF.net email is sponsored by: ValueWeb: 
Dedicated Hosting for just $79/mo with 500 GB of bandwidth! 
No other company gives more support or power for your dedicated server
http://click.atdmt.com/AFF/go/sdnxxaff00300020aff/direct/01/