Le Wed, 11 Aug 2004 09:59:41 +0200 (CEST)
Frank Naumann <fnaumann@cs.uni-magdeburg.de> a écrit:
Hello!
I think this is the thing we should go for here.
That don't help at all to save memory. E.g. the binary is smaller on
harddisc but need the same memroy on runtime.
Without MMU, this is the only solution we can use.
Example for people wanting an explanation:
Process 1: (mintlib for process 1)
(at 0x100000) (loaded at 0x200000)
/----+----+---\/----+----+---\
|TEXT|DATA|BSS||TEXT|DATA|BSS|
\----+----+---/\----+----+---/
Process 2: (mintlib for process 2)
(at 0x300000) (loaded at 0x400000)
/----+----+---\/----+----+---\
|TEXT|DATA|BSS||TEXT|DATA|BSS|
\----+----+---/\----+----+---/
As you say, we don't reduce memory usage, so in memory it is not worse
than current situation. However, we reduce disk space, and avoid
recompilation when a library change: Do you recompile all the binaries RPM
when the mintlib is updated ? So from my point of view, this would be a
progress over current situation.
Plus the fact we could make libraries specifically for each CPU (68000,
68020, 68020+fpu) and the needed ones would be loaded.