Subject: Re: highest priority Re: IC's mystery patent claims
From: Thomas Lord <lord@emf.net>
Date: Fri, 29 Sep 2006 12:34:08 -0700

Thomas Lord wrote:
>>   However, his method
>> requires a (log2 n)**2 preprocessing phase, where our SIMD methods
>> requires only log2 n, a substantial improvement (from 49 down
>> to 7) for 128 bits.
>>   
>
>
> The Smalltalk-80 algorithm is log2(n) ** 2  -- log2(n) in bitblits.   
> It's no small
> feat to adapt for these processors -- plenty of instruction-bumming to 
> do (somwhat
> simplified by specializing for a fixed-size matrix).   It's in chapter 
> 20, on "Display
> Object" in "Smalltalk-80: The Language and its Implementation" by 
> "Goldberg &
> Robson".
>


Sorry, that was wildly unclear.    The point is that the Smalltalk-80 
algorithm
likely has the same complexity as yours and my guess is that the two are
closely related.

-t