Re: MMX emulator ?

H. Peter Anvin (hpa@transmeta.com)
17 Jun 1998 06:49:37 GMT


Followup to: <199806162200.RAA05382@mail.mankato.msus.edu>
By author: Jeffrey Hundstad <jeffrey.hundstad@mankato.msus.edu>
In newsgroup: linux.dev.kernel
>
> no....
>
> If you can assume the MMX functions are there then EACH application
> doesn't have to emulate all of the MMX stuff itself if it doesn't have
> the REAL hardware.
>
> This should be treated EXACTLY like the Floating Point hardware... at
> least in my opinion.
>

Your opinion is, in many ways, flawed. The cost of doing
floating-point on non-FP hardware is such that the cost of emulation
is not a significant overhead. The same doesn't apply for MMX, which
is an integer optimization. Simply put, emulated MMX is too slow to
be used (typically, it will take over 500 cycles to do the two context
switches you need for emulation, and no MMX-instruction-equivalent is
even close to that.)

-hpa

-- 
    PGP: 2047/2A960705 BA 03 D3 2C 14 A8 A8 BD  1E DF FE 69 EE 35 BD 74
    See http://www.zytor.com/~hpa/ for web page and full PGP public key
        I am Bahá'í -- ask me about it or see http://www.bahai.org/
   "To love another person is to see the face of God." -- Les Misérables

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu