home *** CD-ROM | disk | FTP | other *** search
- Path: soap.news.pipex.net!pipex!usenet
- From: m.hendry@dial.pipex.com (Mathew Hendry)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: 680X0 -> PPC translator?
- Date: Wed, 13 Mar 96 01:51:37
- Organization: Private node.
- Distribution: world
- Message-ID: <19960313.45C640.2166@am059.du.pipex.com>
- References: <19960307.41C900.103A8@an168.du.pipex.com> <Dny169.BJH@cix.compulink.co.uk> <19960308.41E5A8.1098C@an157.du.pipex.com> <3145556F.2839@sapiens.com>
- NNTP-Posting-Host: am059.du.pipex.com
- X-Newsreader: TIN [AMIGA 1.3 950726BETA PL0]
-
- Avi Lev (avil@sapiens.com) wrote:
- : Mathew Hendry wrote:
- :
- : > In other words, the _user_ would effectively be guiding the static translation
- : > of an executable. With each run of the program, hopefully more code would be
- : > translated, and the program would run progressively faster. This completely
- : > removes the problem of misplaced code / data, because only code which would
- : > really be executed ends up being translated (you do have the problem of how
- : > to deal with programs which go haywire and jump into data segments, though -
- : > almost impossible to handle whichever method you use)...
- : >
- : > Note: DEC's FX!64 system uses this method, and claims 70% of native
- : > performance for fully translated programs.
- :
- : let me give you a simple reason why you're wrong and why dynamic translation
- : is NOT the way to go, well you're assuming that the translated PPC code will
- : be of the same size as the original code, well that is simply not
- : necessarally true
-
- Um, it is almost certainly not true. And it's obviously not a major problem -
- FX!64 seems to be able to handle it...
-
- : you can't change the
- : segment size of the hunk in memory during run-time and even if you could
- : that would have a great performance impact cuz you would have to rewrite all
- : the code in the newly allocated segment each time you need more space,
-
- You call 70% of native speed a "great performance impact"? I have already said
- that emulation will be slow initially, and become faster as more code is
- translated...
-
- : the best way to go is by doing static translation of all the code and there
- : are algorithms to do that, they're complexity is what prevents them from
- : being implemented that's all.
-
- It's pretty much impossible in the general case. But go ahead, try it.
-
- -- Mat.
-