Gene's article only mentions the space needed for the initial table used in the
selection of the blists by BLT.
(32 bit)
y=.8
xx=.(y#y)#: i.y^y
7!:5 <'xx'
1073741824
So my technique generates an initial table twice the size of Gene's. This does
not include the space used for subsequent processing by bl and ~. .
--
David Mitchell
David Mitchell wrote:
> You are welcome, Ian.
>
> My reconstruction must be a bit more brute-force than Gene's version.
>
> I ran this on 64 bit J on a system with 5G of memory:
>
> 7!:2 'BLT 8'
> 7247762688
>
> This is with 8 byte integers, but still exceeds the 1073741824 bytes Gene's
> formula would estimate.
>
> --
> David Mitchell
>
>
> Ian Clark wrote:
>> Thank you, David, for your reconstruction of verb BLT. It works fine
>> and gives the expected results. I have inserted it in
>> http://www.jsoftware.com/jwiki/Doc/Articles/Play171 after the first
>> mention of BLT.
>>
>> I will also include you in the list of those acknowledged in the
>> production of APWJ Edn 2.
>>
>> What follows is not needed for APWJ Edn 2 but interests me nonetheless...
>>
>> #BLT 7
>> 47293 (...which accords with Gene's figures)
>> #BLT 8
>> |out of memory: BLT
>> | ~. bl"1(y#y)#:i.y^y
>>
>> Gene too says in the paper in question that he can't use BLT beyond 7.
>> He calculates the space needed for BLT 8 (admittedly by his version)
>> as 536,870,912 bytes -- say half a GB. But my iMac has 4GB of
>> installed memory. I would have thought off-hand it would have managed
>> BLT 8 (though certainly not BLT 9, which would need roughly 14GB by
>> Gene's calculations).
>>
>> What's going on? How much of my 4GB is available for J? ...to malloc
>> in one go? Or are my integers taking up two/four/eight times the bytes
>> of Gene's?
>>
>> Ian
>>
>>
>> On Wed, Oct 21, 2009 at 11:11 AM, David Mitchell <[email protected]>
>> wrote:
>>> Oleg Kobchenko wrote:
>>>> In the article 24 "Blists in OLEIS"
>>>>
>>>> http://www.jsoftware.com/jwiki/Doc/Articles/Play171
>>>>
>>>> the definition of the BLT verb is missing.
>>>>
>>>> It is needed to complete the publication of
>>>> At Play with J.
>>>>
>>>> I guess if we should come up with a substitute,
>>>> it might as well be a best shot.
>>>>
>>>>
>>>>
>>>> ----------------------------------------------------------------------
>>>> For information about J forums see http://www.jsoftware.com/forums.htm
>>>>
>>> In the keeping of brute force:
>>>
>>> bl =: ] i.~ [: /:~ ~.
>>>
>>> BLT=: 3 : 0
>>> ~. bl"1 (y#y)#: i.y^y
>>> )
>>>
>>> --
>>> David Mitchell
>>> ----------------------------------------------------------------------
>>> For information about J forums see http://www.jsoftware.com/forums.htm
>>>
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
