On 8/3/2010 9:43 PM, Christian Soeller wrote:
> Is it possible to change things so that 64 bit sizes
 > can be passed in the two places you identified and see
 > if that works?
>
> I appreciate that things could still fall over in various
 > PP autogenerated code pieces if ints are used for offset
 > calculations in slice and other vaffine operations.

It could work.  The problem is it needs someone with
a 64bit OS, lots of memory, and a willingness to
debug the issue.  I don't have any *large* memory
systems at the moment.  Not that I wouldn't like to
have one.  :-)

--Chris


> On 4/08/2010, at 12:46 PM, Chris Marshall wrote:
>
>> I took a further look at the SvGROW calls in
>> PDL/Basic/Core routines and the two that I found
>> both use int type for their sizes.  That would
>> limit a piddle size to<2**31 or about 2GB.
>>
>> It looks like 64bit support for PDL may need
>> to be added to the list for the future.  I don't
>> know the scope of the changes that would be
>> required to support larger PDL data objects.
>>
>> --Chris
>>
>> On 8/3/2010 8:39 PM, Chris Marshall wrote:
>>> On 8/3/2010 8:30 PM, P Kishor wrote:
>>>> On Tue, Aug 3, 2010 at 7:19 PM, Chris Marshall<[email protected]>    wrote:
>>>>> On 8/3/2010 8:01 PM, P Kishor wrote:
>>>>>>
>>>>>> also on 64-bit Snow Leopard (Mac OS X 10.6.4)
>>>>>>
>>>>>> punk...@lucknow ~$ perl -MPDL -e '$PDL::BIGPDL=1; $x = sequence(float,
>>>>>> 23171, 23171); print $x->info("%M")."\n"'
>>>>>> perl(85899) malloc: *** mmap(size=18446744071562166272) failed (error
>>>>>> code=12)
>>>>>> *** error: can't allocate region
>>>>>> *** set a breakpoint in malloc_error_break to debug
>>>>>> Out of memory!
>>>>>> punk...@lucknow ~$
>>>>>
>>>>> What is perl -V?
>>>
>>> I looked at the PDL/Basic/Core stuff and it looks like
>>> if SvGROW can handle a>2GB string then, in principle,
>>> PDL should be able to handle piddles of that size.
>>>
>>> Could you see if you can create a string more than 2GB
>>> long?  It might take a while but it will tell us if the
>>> limit is perl or PDL.  Since the PDL routines for growing
>>> a new piddle use 4byte ints for their sizes (rather than
>>> size_t objects), it is pretty clear that there is a bug
>>> in the PDL allocation stuff if perl can handle the longer
>>> strings.
>>>
>>> --Chris
>>
>> _______________________________________________
>> Perldl mailing list
>> [email protected]
>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
>
>
>
>
> No virus found in this incoming message.
> Checked by AVG - www.avg.com
> Version: 9.0.851 / Virus Database: 271.1.1/3049 - Release Date: 08/03/10 
> 10:22:00
>


_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

Reply via email to