home *** CD-ROM | disk | FTP | other *** search
- From: bernard@labri.u-bordeaux.fr (BERNARD Sebastien [93-94])
- Subject: Questions about memory and library
- Date: Tue, 11 Jan 1994 14:41:02 +0100 (MET)
- Mime-Version: 1.0
-
- Hello from france,
- I'am a mint user since a few years now (from the 0.6 version precisely) and
- now I'm a bit disappointed with MiNT for the following reasons :
- - I started working on it with a 4MB STE but it was slow and lacking of memory.
- so I went on a Falcon (8 months ago). It was faster but with still 4MB of
- memory. But it's still short of memory. Unable to compile anything with
- GNU and miniwin on desktop but little toys (files less than 60Kb) without
- running anything with. Using OUTSIDE make the Falcon crash after the memory
- has gone short.
-
- - So I ask myself, about three points that could be done :
- 1- virtual memory :
- Looking the source of MiNT 1.09 allow me to think that everything is rea-
- dy for this but the code ( there is even an empty function - init_swap -
- for initialising the swap device ). It seems that virtual memory has been
- voluntary stopped at the memory protection. Is it interesting for anybody
- to develop virtual memory ? I am insteresting in it. But, maybe is there
- some development running ? Please, let me know.
-
- 2- shared memory :
- For solving probkems due to lack of memory, a good solution could be the
- sharing of memory for libraries so we don't have to duplicate the code
- in the memory just the instance of variables. For this - it joins the
- first point - we have to extend the MiNT kernel : add the possibility of
- sharing pages of memory for root's process - or others' - .
- This could creates some problems with executables that could absolutely
- not be run under normal TOS ( but who wants downard compatibility ).
- This will be very helpfull for make compilations in severals directory
- where you can find up to 10 shells and 5 make running. We can make this
- by improving the memory protection. Every process could start to the
- same adresse ( relocation is no more needed ). The problem is that it
- could not be done for the 68000 machines ( only MMU machines ).
- This leads me to the third points :
-
- 3- a new executable format :
- For differenciating shared executables and normal executables, it could
- interesting to use a different binary format. I wish to use a standard
- format to simplify the porting of GNU software and getting fully funtionnal
- ports ( dump of emacs directly usable - straight port of GDB ). For this
- should we modify the Pexec call to allow the execution of new binary format
- and on the fly the working shell hack "#!/bin/sh" directly implemented.
-
- This is what I'd loved to see inside MiNT and the MiNTLIB. I wish I do it but
- I don't know 680x0'asm. I'm not fluent in MiNT but I could try. If anybody is
- interested, let me know.
-
- S. Bernard
-