[Oberon] Project Oberon running from LPDDR memory on Pipistrello

Joerg joerg.straube at iaeth.ch
Mon May 11 21:18:48 CEST 2020


Thx for the explanation.
Jörg

> Am 11.05.2020 um 20:34 schrieb Andreas Pirklbauer <andreas_pirklbauer at yahoo.com>:
> 
> 
>> 
>> I guess it‘s a typo. Must be ORB iso ORS as the scanner doesn’t need heap space.
>> If I rember correctly, some objects unintentionally got allocated more than once.
> 
> No, the issue was that a change made earlier replaced some BYTE with INTEGER in
> ORB.ObjDesc, causing heap space consumption to double (from 64 to 128 byte for each
> allocated object). I reverted that change, so that the fields now occupy 56 bytes again. If
> you add 8 bytes (for the mark field and the type tag), you get 64 bytes per allocated obj.
> 
> PS: This is also the reason why ModDesc is an extension of ObjDesc.. so that 128 bytes
> of heap space is allocated only for modules (typically there are only a few), whereas for
> all other objects in the symbol table 64 bytes suffice.
> 
> 
> --
> Oberon at lists.inf.ethz.ch mailing list for ETH Oberon and related systems
> https://lists.inf.ethz.ch/mailman/listinfo/oberon



More information about the Oberon mailing list