[Ffmpeg-devel] [PATCH] qcelp codec supporrt

Michel Bardiaux mbardiaux
Mon Dec 18 11:47:44 CET 2006


Rich Felker wrote:
> On Thu, Dec 14, 2006 at 10:28:33AM +0100, Michel Bardiaux wrote:
[snip]
>> Seymour Cray failing math101?
> 
> I didn't know the man, 
Prime designer of the CDC-Cyber7000 series, CDC-Cyber6000 series, 
Cray-1, Cray-XMP, Cray-2, Cray-YMP. IOW, the leading supercomputers of 
the '70s and '80s. And yes, he had some *very* idiosyncratic ideas, 
implemented on the CDCs, like 1-complement arithmetic, 60-bits 
registers, writeable code memory, and total lack of RAM parity! But it 
also was already using such brilliant ideas as RISC and VLIW and 
Hyperthreading.

> but mathematically speaking "ones complement"
> is utterly idiotic. It's inventing two special cases for something
> which is simpler, more efficient, more functional (1 extra value), and
> non-ambiguous if you do it in number-theoretically correct way, "twos
> complement", which is just arithmetic modulo 2^n where the concent of
> signedness does not even exist.
> 
I dont disagree with your arguments, and indeed starting from the Cray-1 
they moved to 2-complement, 64-bits integer, and SECDED memory. I just 
wonder what the counter-arguments could have been - because Seymour Cray 
*was* a genius, no doubt about it.

-- 
Michel Bardiaux
R&D Director
T +32 [0] 2 790 29 41
F +32 [0] 2 790 29 02
E mailto:mbardiaux at mediaxim.be

Mediaxim NV/SA
Vorstlaan 191 Boulevard du Souverain
Brussel 1160 Bruxelles
http://www.mediaxim.com/




More information about the ffmpeg-devel mailing list