compileIt for revolution?

[Dennis Brown]

John,

I actually had a complete matrix algebra package as an external for  
Hypercard.  I know how to use it, because I cut my teeth on APL.  I  
implemented my complex algorithms and state machines in it.  The  
coding was cumbersome, and it was difficult to debug and handle edge  
conditions, sparse arrays, etc.  And to top it all off, it only ran  
25% faster than the simple HyperCard stack even though it could  
crunch numbers 10 times faster.  However, I have to admit that it  
spent a lot of time converting and passing arrays back and forth to  
HyperCard.  A native engine capability might make a big difference if  
the internal representation could be typed by making sure you only  
put all the same type numbers into the array, and if the indexes were  
truly integers.  The both matrix and element by element operations  
could be mixed and we would have the best of both worlds.

Dennis


On Jun 22, 2005, at 3:30 PM, Dr.John R.Vokey wrote:
> It seems to me that the issue of speed applies mostly to the  
> (mathematical) manipulation of numerical arrays; we need a complete  
> matrix algebra package as part of the engine.  If we had such, I  
> suspect that virtually all of the desire for CompileIt!, and  
> complaints about speed, would be vitiated.
>
> On 22-Jun-05, at 9:51 AM, use-revolution-request at lists.runrev.com  
> wrote:
>
>
>> Well, from what I see here is a business opportunity.
>>
>> If everyone that is interested in achieving this (and willing to  
>> pay).
>> Then I would suggest a group should be formed andcreate a well  
>> defined
>> set a specifications.
>>
>> I am sure that Chipp at Altuit or someone else would be able to
>> produce the much desired solution but for a price.
>>
>> This would be the quickest way of getting exactly what is being asked
>> for in the shortest period of time.
>>
>> I would be glad to coordinate (off the list ) creating the product
>> specifications and then putting it out to bid.
>>
>> ...any takers?
>>
>> Please contact me off the list.
>>
>> regards,
>>
>> Glen


On Jun 22, 2005, at 3:45 PM, Dr.John R.Vokey wrote:

> The simplest possibility for the external route would be to be able  
> to pass (and use) to and from the external, handles to large RAM- 
> based sets of floating-point numbers; that was the solution I used  
> in Hypercard back when the Earth was still cooling, and it worked  
> very well.
>
> On 22-Jun-05, at 10:35 AM, use-revolution-request at lists.runrev.com  
> wrote:
>
>
>> I think the issue here is an elegant and efficient way to interface
>> with externals.  If we had this, it would solve both problems.  When
>> I say efficient, I don't mean pass a 100MB array as a text string to
>> an external that converts it to numbers than does some processing and
>> converts back to a text string and passes the 100MB back, then the
>> 100MB is passed again to the external for the next processing
>> step...  NOT the solution!
>>
>> Dennis
>>
>>
> --
> John R. Vokey, PhD
> Professor
> B.E.R.G. - Behaviour and Evolution Research Group
> Micro-Cognition Laboratory
> Department of Psychology & Neuroscience
> University of Lethbridge
> Lethbridge, Alberta T1K 3M4
> CANADA
> (403) 329-2409 office/lab
> (403) 329-2555 FAX
>
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your  
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-revolution
>