NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/programmer/84

Internet Message Format | 1996-08-05 | 3.3 KB

Path: mail2news.demon.co.uk!sisyphus.demon.co.uk
From: [e-mail]
Newsgroups: comp.sys.amiga.programmer
Subject: Re: PPC compilers
Date: Tue, 2 Jan 96 19:32:22 +0000 (GMT)
Organization: to be supplied
Distribution: world
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]> <4bqdk8$[e-mail]> <[e-mail]> <4c9h7d$[e-mail]>
X-NNTP-Posting-Host: sisyphus.demon.co.uk
In-reply-to: [e-mail]'s message of 1 Jan 1996 20:46:37 GMT
X-Attribution: DaveS
Content-Length: 2600
X-Lines: 66
X-Mail2News-Path: relay-4.mail.demon.net!post.demon.co.uk!sisyphus.demon.co.uk

>>>>> "JF" == Juergen "Rally" Fischer <[e-mail]> writes:

  JF> [e-mail] wrote:
  >> >>>>> "JF" == Juergen "Rally" Fischer
  >> <[e-mail]> writes:

  >> The issue isn't whether or not it does it wrong, but whether what it
  >> does is defined (for all 680xx processors).  If you know what
  JF> agree.
  >> this instruction is _defined_ to do, quote the definition.  Motorola's
  >> definition, not yours.

  JF> mhm, is Hitatchi doku also ok ?

  JF> "adress register indirect with postincrement: The adress of the
  JF> operand is in the adress register specified by the register field.
  JF> AFTER THE OPERAND IS USED, IT IS INCREMENTED by one, two or four
  JF> [...]"

My original post used PREDECREMENT, not postincrement.

  JF> I would interpret "after operand is used" that way:

  JF> when data from (a0) is read, a0 was used (used to build an adress).
  JF> so a0 is incremented. then data is written at (a0)+.

"after the operand is used" does not necessarily mean "before
the other operand is used".  A good definition leaves nothing
to "interpretation".

  JF> As there are AFAIK no warnings in the docu, it's IMHO legal to assume
  JF> the logic way (if my test code isn't wrong, the EC020 behaves exactly
  JF> that way. 060 users tell your experiences ;)

You were criticising C because its definition very carefully
states that the order of evaluation of expressions between control
points is not fully defined, so expressions like

   *--A0 = *--A0;

do not give uniformly predictable results.  You then claim that
assembly code is better because the assembly code definition gives
no such warning for

   MOVE.L -(A0),-(A0)

leaving you free to assume that _all_ processors implementing this
instruction use the interpretation you choose to use.

(BTW, if anyone wants to test this on an '060, remember that the
effect may depend on whether A0 is used or modified in the few
instructions executed before or after this one, or on the state of
the instruction and data caches, or on ...)

  >> -- [e-mail] (Staffordshire, England)
  >> ... details are more implementation-dependant than defined.

The line in my .sig is a direct quotation from another
manufacturer's architecture manual.  It looks horrifying, but at
least it's honest (and in context applies only to privileged
instructions which would be used only in the operating system).

--
        [e-mail]        (Staffordshire, England)

 ... details are more implementation-dependant than defined.