RE: [Dev-C++] new is a wrapper around malloc so...

[Per W. <pw...@ia...>]

64-bit isn't just a speed issue :-)

Anyway, while most motherboards may only support 2-4GB memory, but the
386 processor isn't limited to 4GB physical memory, and some server
motherboards supports a lot more. However, with 32-bit paged memory
the machine may then run a number of applications each allocating 2-3GB.
But each individual application is limited.

There is a way around it. It's the dreaded 48-bit far pointer (16:32)
where the processor is supporting up to 65536 memory segments of up to 4GB
each. However, all programmers and compiler vendors have already learned
to hate the old 16:16 far pointers from 8086 (1MB) and 80286 (16MB) so
nobody is rushing for 48-bit pointers.

Segmented memory doesn't support or work very well with a number of
concepts we have learned to use and love with our current paged memory
model:
- On-demand allocation. Automatically grow the stack in 4kB increments
whenever needed.
- Memory fragmentation. In a paged environment, the OS may never get
fragmented memory, since each memory page is identical. It just keeps
track of a list of available pages, and maps in the first available page
when a process needs more memory. With segmented memory, the OS will
normally have to block the task and move the memory segments around in the
physical memory to get a continuous free block large enough for a memory
allocation.

/Per W

On Wed, 16 Jul 2003, Daniel Glenfield wrote:

> Ahhhhhhhhhhh, so that's why everyone wants 64-bits. Games too are requiring
> more, probably because of the same memory-mapping issues.
>
>
> >From: Per Westermark <pw...@ia...>
> >To: Daniel Glenfield <sp...@ho...>
> >CC: mu...@gm...,  <dev...@li...>
> >Subject: RE: [Dev-C++] new is a wrapper around malloc so...
> >Date: Wed, 16 Jul 2003 12:26:07 +0200 (CEST)
> >
> >A 32-bit pionter is an unsigned entity, so it can span 2^32 bytes - 4GB.
> >From this address space, the different OS then reserves areas for OS code,
> >buffers, IPC etc. Depending on OS, you may normally not allocate more than
> >2 or 3GB for a single application.
> >
> >This is the big reason why the world is so desperately wanting 64-bit
> >machines. In database solutions it is quite common to memory-map the
> >database files. However, 2-3GB is a puny database.
> >
> >/Per W
> >
> >On Wed, 16 Jul 2003, Daniel Glenfield wrote:
> >
> > > All I know is that the MSDN documentation I read says 2GB. I'm not sure
> > > about how you calculate it all, so you're one up on me!
> > >
> > >
> > > >From: "Abhijit Shylanath" <mu...@gm...>
> > > >To: "Daniel Glenfield" <sp...@ho...>
> > > >CC: "Dev-Cpp mailing list" <dev...@li...>
> > > >Subject: RE: [Dev-C++] new is a wrapper around malloc so...
> > > >Date: Wed, 16 Jul 2003 00:40:19 +0530
> > > >
> > > >
> > > > > In 16 bit your code is separated into 'segments' each a maximum of
> >64k
> > > >in
> > > > > size. 32-bit can have up to 2GB segments (!!!).
> > > >
> > > >4G, right? (2^32).
> > > >
> > > >However, that's true only on a 32-bit system (as you've said)... what
> > > >platform are you building on? On 16-bit (`real-mode') DOS, even though
> > > >sizeof(farptr) will give you four, it cannot address 32-bits worth of
> > > >memory. Far pointers work like this:
> > > >
> > > >	32 bits of farptr  =
> > > >	16 bits of segment .
> > > >	16 bits of offset
> > > >
> > > >The effective address is actually calculated by multiplying the segment
> >by
> > > >16 decimal (0x10 (which is equivalent to left-shifting by four)) and
> >then
> > > >ADDING the offset... so that if your far-ptr has this value:
> > > >
> > > >	0x1234abcd
> > > >
> > > >Written as:
> > > >
> > > >	1234:abcd
> > > >
> > > >Its effective address, which is worth 20-bits (and can therefore
> >address
> > > >2^20 = 1M) is calculated  as:
> > > >
> > > >	0x12340
> > > >+	 0xabcd
> > > >-------------
> > > >=	0x1cf0d
> > > >
> > > >Whereas, in the flat memory model(protected-mode), the effective
> >address
> > > >would itself be
> > > >
> > > >	0x1234abcd
> > > >
> > > >--
> > > >Abhijit Shylanath
> > > >Homepage: http://mudeth.tripod.com/
> > > >
> > >
> > > _________________________________________________________________
> > > On the move? Get Hotmail on your mobile phone
> >http://www.msn.co.uk/msnmobile
> > >
> > >
> > >
> > > -------------------------------------------------------
> > > This SF.net email is sponsored by: VM Ware
> > > With VMware you can run multiple operating systems on a single machine.
> > > WITHOUT REBOOTING! Mix Linux / Windows / Novell virtual machines at the
> > > same time. Free trial click here: http://www.vmware.com/wl/offer/345/0
> > > _______________________________________________
> > > Dev-cpp-users mailing list
> > > Dev...@li...
> > > TO UNSUBSCRIBE: http://www23.brinkster.com/noicys/devcpp/ub.htm
> > > https://lists.sourceforge.net/lists/listinfo/dev-cpp-users
> > >
> >
> >
> >
> >-------------------------------------------------------
> >This SF.net email is sponsored by: VM Ware
> >With VMware you can run multiple operating systems on a single machine.
> >WITHOUT REBOOTING! Mix Linux / Windows / Novell virtual machines at the
> >same time. Free trial click here: http://www.vmware.com/wl/offer/345/0
> >_______________________________________________
> >Dev-cpp-users mailing list
> >Dev...@li...
> >TO UNSUBSCRIBE: http://www23.brinkster.com/noicys/devcpp/ub.htm
> >https://lists.sourceforge.net/lists/listinfo/dev-cpp-users
>
> _________________________________________________________________
> It's fast, it's easy and it's free. Get MSN Messenger today!
> http://www.msn.co.uk/messenger
>
>