Re: [Discuss-gnuradio] New random number generator

[Marcus Müller]

> I'd say we rather want better
> randomness than faster noise sources
If that's the case, I'd recommend that Stefan uses the
normal_distribution variate from boost (Stefan, see my example), rather
than doing his own "normalization" of the RV, and we use that.

Regarding Boost's mt19937 and the ways Stefan and I make normal
distributed RV out of the uint32_t that this emits:
>From getting "raw" uniform uint32_t instead of normal floats through
boost's normal_distribution(rng) variate is that, to little surprise,
75% of time is spent looking up/interpolating/calculating based on the
uniform integers to normally distributed floats. Blind guess is that
this would be something that someone who really cares about speed might
implement in VOLK with sufficient accuracy, making WGN generation
X3x0-capable.

Cheers,
Marcus

[1] let's say someone who wants to transmit at max has 4GB RAM to spare,
or 2^29 gr_complexes, equal to ~2.6s of complex float noise
[2] http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/ARTICLES/mt.pdf
On 02.09.2015 18:53, Martin Braun wrote:
> On 02.09.2015 05:10, Stefan Wunsch wrote:
>> Now my question: Before doing a pull request, do you have any concerns
>> regarding memory use or processing load? Obviously the new
>> implementation isn't that light-weight as the ten lines of code before.
>> But the current implementation can not be used in any serious simulation
>> or publication, which is highly dependent on good random numbers. Some
>> information about the performance is given on this page: [2]. Look for
>> the generator mt19937 in table 24.5.
> Just a quick reminder that already, our noise source is known to be
> pretty slow, which prompted the fast_noise_source (which precaches some
> random data and reuses that). So, when you need maximum entropy, you use
> the slow one and if you just want some AWGN you use the fast one. Unless
> the load is significantly higher, I'd say we rather want better
> randomness than faster noise sources -- and the fast noise source will
> stay unaffected anyway.
>
> M
>
>
> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio