home *** CD-ROM | disk | FTP | other *** search
- From: Kay Roemer <roemer@informatik.uni-frankfurt.de>
- Posted-Date: Wed, 8 Jun 94 9:03:58 MESZ
- Received-Date: Wed, 8 Jun 94 09:03:58 +0200
- Message-Id: <9406080703.AA15701@hera.rbi.informatik.uni-frankfurt.de>
- Subject: Re: MiNT 1.10 re-sync
- To: howard@harry.lloyd.com (Howard Chu)
- Date: Wed, 8 Jun 94 9:03:58 MESZ
- In-Reply-To: <m0qBBLb-000ESDC@lloyd.com>; from "Howard Chu" at Jun 07, 94 5:18 pm
- Mailer: Elm [revision: 70.85]
-
- > Upon reflection, it would probably be easier to load shared text regions at the
- > top of memory, and continue creating the TPA from lower memory. That way you
- > don't have to worry about artificially limiting the amount of memory you give
- > to a newly spawned process. (I wonder how topdown solved this problem...)
-
- This sounds like a possible solution. TPA doesn't matter in this case, because
- it is freed when the process (running a shared text) exits. Fragmentation seems
- to occur, when
-
- 1) running shared text
- 2) last process running shared text exits, shared text still hangs around.
- 3) allocating memory in eg, a device driver or fs or running TSR or daemon.
- 4) goto 1
-
- After a few iterations the stuff allocated in 3) prevents the free shared
- text regions from beeing joined together to one large hunk.
-
- Sigh! Thats a real problem when physical address space == virtual address
- space.
-
- Kay.
-