| Return-path: neth
Delivery-date: Thu, 15 May 97 10:48:27 -0400
Return-path: neth
Received: from quarry.zk3.dec.com by mailhub2.zk3.dec.com;
(5.65v3.2/1.1.8.2/18Sep95-0525AM)
id AA00195; Thu, 15 May 1997 10:48:25 -0400
Received: from granda.zk3.dec.com by quarry.zk3.dec.com;
(5.65v3.2/1.1.8.2/16Jan95-0946AM)
id AA28354; Thu, 15 May 1997 10:47:56 -0400
Received: from localhost by granda.zk3.dec.com
(5.65v3.2/1.1.10.5/11Nov96-0300PM)
id AA24485; Thu, 15 May 1997 10:49:34 -0400
Message-id: <9705151449.AA24485@granda.zk3.dec.com>
X-mailer: exmh version 1.6.7 5/3/96
Mime-version: 1.0
Content-type: text/plain; charset=us-ascii
X-mts: smtp
To: lowell
Cc: lionel@quark.zko.dec.com
Subject: Request to have STWG work on section alignment controls
Date: Thu, 15 May 97 10:49:34 -0400
From: "Craig Neth DTN 381-2174" <neth>
The following request is for an object file enhancement that would allow
different alignment of sections (and possibly symbols?) than the defaults.
We've heard this several times now - ev6 wants to be able to align .text
on stronger boundaries, and now this request.
I'd like to propose that STWG work on defining a mechanism for this
(my previous thinking was to reclaim the dead lnno (or lnnoptr?) field in the
section header for the section alignment). If you folks define it, we
can work on getting it into the linker.
Craig
------- Forwarded Message
Return-Path: lionel@quark.zko.dec.com
Delivery-Date: Thu, 15 May 97 10:38:25 -0400
Return-Path: lionel@quark.zko.dec.com
Received: from quarry.zk3.dec.com by mailhub2.zk3.dec.com;
(5.65v3.2/1.1.8.2/18Sep95-0525AM)
id AA32312; Thu, 15 May 1997 10:38:23 -0400
Received: from quark.zko.dec.com by quarry.zk3.dec.com;
(5.65v3.2/1.1.8.2/16Jan95-0946AM)
id AA27102; Thu, 15 May 1997 10:37:56 -0400
Date: Thu, 15 May 1997 10:38:12 -0400
Message-Id: <97051510381279@quark.zko.dec.com>
From: lionel@quark.zko.dec.com (Steve Lionel)
To: NETH@zk3.dec.com
Subject: Re: Forcing common blocks to page boundaries? - comp.lang.fortran
#51769
X-Vms-To: NETH@ZK3.DEC.COM
X-Vms-Cc: TLE::YOUNGS,TLE::DESPRES
In the following posting from Scott Stallcup of the Space Telescope Science
Institute, he explains how DIGITAL UNIX's lack of ability to specify
page-alignment for PSECT contributions prevents them from porting their
application to that platform. This has been a sore point for quite a while,
is there anything that could be done about this in the near term?
Steve
In article <3379C156.79D6@stsci.edu>, Scott Stallcup <stallcup@stsci.edu>
writes:
Steve Lionel wrote:
>
> ...
>
> > In VMS Fortran, how does one force a common block to start on a page
> > boundary?
>
> In the current versions of Digital Fortran, you can use:
>
> CDEC$ PSECT /FOO/ ALIGN=PAGE
>
> This will do the correct thing on VAX and Alpha. If you have an older
> version, use ALIGN=9 on VAX and ALIGN=16 on Alpha. (Caveat for
> completeness - PAGE isn't supported on UNIX and neither are alignment
> values greater than 3, due to lack of object language support.)
>
The lack of support for page alignment of common blocks on Digital Unix
has been a major problem for us. There are no reasonable workarounds
either (that don't require a major rewrite of out code).
Our system (million+ lines of code) doesn't run on Digital Unix because
of this one problem. We currently support only VMS and Sun/Solaris...
- --------------------------------------
Scott Stallcup (stallcup@stsci.edu)
Space Telescope Science Institute
------- End of Forwarded Message
[Posted by WWW Notes gateway]
|