Re: [boost] [random] new threefry random engine

Subject: Re: [boost] [random] new threefry random engine
From: John Bytheway (jbytheway+boost_at_[hidden])
Date: 2014-05-01 22:58:11

On 2014-04-22 10:51, John Salmon wrote:
> On Mon, Apr 21, 2014 at 7:51 PM, Steven Watanabe <watanabesj_at_[hidden]>wrote:
> 
>> AMDG
>>
>> On 04/21/2014 02:28 PM, John Salmon wrote:
>>> <snip>
>>> I would love to see a boost implementation that captured and exposed
>>> this capability for general use.  Thus, I think a better approach is
>>> as follows:
>>>
>>> - define a new "RandomFunction" concept that dscribes/enforces the
>>> essential, common features of the randomizing function in Threefry,
>>> Philox, ARS and other counter-based random generators.
>>>
>>
>> I don't think this is a particularly useful
>> abstraction.  Most users should not care at
>> all about this, since they should be using
>> the predefined engine typedefs.  If more
>> similar generators are added later, then the
>> implementation can be adjusted to factor
>> out the common boilerplate, but I really
>> don't see the point of creating a whole new
>> concept for this.
>>
> 
> 
>> I don't think this is a particularly useful
>> abstraction.  Most users should not care at
>> all about this, since they should be using
>> the predefined engine typedefs.  If more
>> similar generators are added later, then the
>> implementation can be adjusted to factor
>> out the common boilerplate, but I really
>> don't see the point of creating a whole new
>> concept for this.
> 
> It may well be true that most users should use a predefined generator
> because the Engine concept is well matched to the style of random
> number generator that has been in use for 65 years.
> 
> But the question is not whether a RandomFunction abstraction should or
> would be used by *most* users.  The question is whether it's useful to
> *enough* users.

In case it helps motivate this abstraction, I will mention that I have
wanted it in the past.  I wanted to generate unpredictable Perlin noise
in a game (predictability would in this case allow clients to cheat).
For that I wanted exactly a RandomFunction like this which would simply
map each integer to an independent random value.  This is essentially
the same interface Mathias was asking for elsewhere in the thread.

What I actually ended up using was a linear congruential generator with
O(log(n)) stepping, which didn't satisfactorily achieve any of my design
goals.  A good random function in Boost.Random would have been perfect.

John