Home > Archive > Compression > November 2007 > Parallel White Noise Generation on a GPU via Cryptographic Hash
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Parallel White Noise Generation on a GPU via Cryptographic Hash
|
|
| Mark Nelson 2007-11-05, 7:56 am |
| A TR from the evil empire on the topic shown above. Probably a lot
better than the standard LCR generators we're used to using for random
numbers.
http://research.microsoft.com/resea...+report&id=1384
No mention of intellectual property claims.
Let me go ahead and write up the usual responses in advance so I can
save time and bandwidth:
1) Microsoft sucks, everything in this paper will be patented so it's
pretty much useless to me.
2) Microsoft sucks, this is all old stuff that Jim Leonard did on his
Atari in 1986. As usual all they are doing is mining prior art and
calling it their own.
3) Microsoft sucks, they're just trying to create a new standard based
on their proprietary technology.
4) I hate Windows. OS/X|Linux rulez.
|
| Mark Nelson - http://marknelson.us
|
| |
| Phil Carmody 2007-11-05, 7:56 am |
| Mark Nelson <snorkelman@gmail.com> writes:
> A TR from the evil empire on the topic shown above. Probably a lot
> better than the standard LCR generators we're used to using for random
> numbers.
>
> http://research.microsoft.com/resea...+report&id=1384
>
> No mention of intellectual property claims.
>
> Let me go ahead and write up the usual responses in advance so I can
> save time and bandwidth:
>
> 1) Microsoft sucks, everything in this paper will be patented so it's
> pretty much useless to me.
>
> 2) Microsoft sucks, this is all old stuff that Jim Leonard did on his
> Atari in 1986. As usual all they are doing is mining prior art and
> calling it their own.
>
> 3) Microsoft sucks, they're just trying to create a new standard based
> on their proprietary technology.
>
> 4) I hate Windows. OS/X|Linux rulez.
I vote "2". Or at least that's closest. Their "Parallel Computation"
claim on the first page is just plain wrong. They're assuming what
they want to prove, for a start. The "Statistical Randomness" claim
looks like an overgeneralisation (some conventional methods are bad,
therefore our method is better than conventional methods). The
"Order Independence" seems to be assuming that there is always a
measurable statistical bias (otherwise you wouldn't need to avoid
it). It also looks like their corpus of prior PRNG implementations
is woefully inadequate.
Oh, and Jim for President!
Phil
--
Dear aunt, let's set so double the killer delete select all.
-- Microsoft voice recognition live demonstration
| |
| Thomas Richter 2007-11-05, 6:56 pm |
| Phil Carmody schrieb:
> Mark Nelson <snorkelman@gmail.com> writes:
>
> I vote "2". Or at least that's closest. Their "Parallel Computation"
> claim on the first page is just plain wrong. They're assuming what
> they want to prove, for a start. The "Statistical Randomness" claim
> looks like an overgeneralisation (some conventional methods are bad,
> therefore our method is better than conventional methods). The
> "Order Independence" seems to be assuming that there is always a
> measurable statistical bias (otherwise you wouldn't need to avoid
> it). It also looks like their corpus of prior PRNG implementations
> is woefully inadequate.
Concerning 2), there's a nice related work by Toffoli and Margolus ("CAM - Cellular Automaton Machines",
Cambridge University Press) where the authors construct a Pseudo-RNG from a cellular automaton,
back then in hardware. Clearly, the same machine can be implemented on a GPU nowadways.
So long,
Thomas
| |
| Phil Carmody 2007-11-05, 6:56 pm |
| Thomas Richter <thor@math.tu-berlin.de> writes:
> Phil Carmody schrieb:
[color=darkred]
[color=darkred]
>
> Concerning 2), there's a nice related work by Toffoli and Margolus ("CAM - Cellular Automaton Machines",
> Cambridge University Press) where the authors construct a Pseudo-RNG from a cellular automaton,
> back then in hardware. Clearly, the same machine can be implemented on a GPU nowadways.
I'm not a big fan of CA PRNGs. Sure, the parallelisability
is great, but I'm not convinced you get any more randomness,
and in fact suspect you get less, than from a similarly-
sized LFSR-based design. (Unless the CA is reversible or
infinite in domain, it's a contraction of the phase-space.)
Phil
--
Dear aunt, let's set so double the killer delete select all.
-- Microsoft voice recognition live demonstration
| |
| Jim Leonard 2007-11-05, 6:56 pm |
| On Nov 5, 7:34 am, Phil Carmody <thefatphil_demun...@yahoo.co.uk>
wrote:
> Oh, and Jim for President!
Hey now, you guys leave me out of this :-)
|
|
|
|
|