RE: Teraspace Storage Model and CEE APIs

["Schmidt, Mihael" <Mihael.Schmidt@xxxxxxxxxxx>]

Thanx but Scott was faster =) (seems that he doesn't sleep at all =) 

-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Vern Hamberg
Sent: Monday, March 14, 2011 12:51 PM
To: Midrange Systems Technical Discussion
Subject: Re: Teraspace Storage Model and CEE APIs

Mihael

You maybe know this already. It seems that RPG support for teraspace has 
been around since v4r4, as Scott said. You can create any module with 
the *INHERIT special value for the STGMD. Then you need to write your 
code to anticipate either model. All this, for the archives, is in the 
ILE Concepts manual.

Seems 7.1 has made some of this easier.

HTH
Vern

On 3/14/2011 4:19 AM, Schmidt, Mihael wrote:
> Thanks for the clarification. I think I'll stick to the bifs and then use H-specs keyword if needed as I don't need teraspace at the moment but would like to have it for the worst case scenario. (after giving it some more thought it probably won't do it any harm if I change it to teraspace now)
>
> It seems that an exchangeable memory management module wouldn't be that bad.
>
> Thanx
>
> Mihael
>
> -----Original Message-----
> From: midrange-l-bounces@xxxxxxxxxxxx [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Scott Klement
> Sent: Monday, March 14, 2011 9:31 AM
> To: Midrange Systems Technical Discussion
> Subject: Re: Teraspace Storage Model and CEE APIs
>
> Hi Mihael,
>
> I don't think the ILE CEE APIs support teraspace at all.  If you want to
> allocate memory in teraspace, why not call the teraspace APIs?  (This
> has been available in RPG since v4r4...)
>
> Here are the prototypes:
>
>     /if defined(*V5R2M0)
>    D TS_malloc64     PR              *   ExtProc('_C_TS_malloc64')
>    D   size                        20U 0 value
>     /endif
>    D TS_malloc       PR              *   ExtProc('_C_TS_malloc')
>    D   size                        10U 0 value
>    D TS_realloc      PR              *   ExtProc('_C_TS_realloc')
>    D   ptr                           *   value
>    D   size                        10U 0 value
>    D TS_free         PR                  ExtProc('_C_TS_free')
>    D   ptr                           *   value
>
> With these, RPG can allocate terabytes of memory to a single pointer.
> You don't need the teraspace storage model to use these...
>
> In 7.1, IBM added the teraspace storage model to RPG.  The storage model
> controls the memory used for automatic, static, and constant storage --
> in other words, it has no impact on the stuff you allocate yourself from
> the heap!
>
> IBM also added the ALLOC H-spec keyword in 7.1 that lets you control
> whether the ALLOC/REALLOC opcodes and %ALLOC/%REALLOC BIFs use teraspace
> or single-level storage.  However this has no impact on what happens
> when you call an API to allocate memory, it only affects the opcodes/bifs.
>
> So if you want to use teraspace for heap storage, and you don't want to
> wait for 7.1, just call the teraspace routines.  Maybe even wrap them
> with a subprocedure, so you can switch your code between the various
> alternatives (CEE APIs, RPG opcodes, teraspace APIs, etc) without having
> to change the logic of your main program.
>
>
> On 3/14/2011 2:17 AM, Schmidt, Mihael wrote:
> > I got a program which uses dynamic storage (CEE APIs) to store some
> > temporary data in memory. So far so good as everything is working
> > fine and there is no current business case which makes my program
> > exceeds the memory it can allocate. I am using the default heap.
> >
> > Now as I understand it IBM i 7.1 lets you compile RPG modules with
> > teraspace as the storage model. Does that also mean that the CEE API
> > uses memory from teraspace and does it mean that I can use 1TB with
> > the same api and just a different compile setting (
> > STGMDL(*TERASPACE) ) ? I have no chance to test this as we are on 6.1
> > still.