🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.arch
Article #20574

Re: Unaligned Accesses (was Re: How to use silicon)

#20574
From: Mark Lord
Date: Mon, 27 Mar 1989 14:25
38 lines
2648 bytes
In article <1989Mar25.013342.22214@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes:
>In article <362@bnr-fos.UUCP> mlord@bnr-public.UUCP (Mark Lord) writes:
>>The debate for devoting more silicon space to support efficient handling
>>of misaligned accesses seems to hinge around supporting the huge amounts
>>of perfectly good software that already exist, written before programmers
>>in general became keenly aware of alignment/efficiency tradeoffs.
>
 [stuff I agree with deleted]
>...  I continue to think that the quantity
>of good software which disregards alignment is grossly overestimated.
>(It is not zero, and attention to it is necessary, but the idea that
>a random software package has a 90% chance of suffering from such
>carelessness is unjustified.)

I partially agree with Henry here, but would like to point out that over
the past six years or so, a lot of systems, including a very large (many
millions of lines) system with which I am intimately familiar, have taken
the migration path from 16-bit to true 32-bit processors.  Such software
tends to be quite well aligned on 16-bit boundaries, but not always well
aligned on 32-bit boundaries.  Again, this thread is dealing with very
large (10's of megabytes) of allocated data, tightly packed into structures
of related fields.  Yes, such software could be revisited and aligned
properly for 32-bit or 64-bit (or whatever-bit) processors, but only at
great expense in both manpower and risk to a perfectly good working package.

From my experience, I do agree that alignment on 16-bit boundaries is 
probably not at issue here.  But a lot of software is still being moved
into the 32-bit age from 16-bit systems.  This will probably not continue
beyond another ten years or so, but it would be nice to have the hardware
support for it while it is still important.

Cheers

-Mark

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
If I were writing this for the Company, I'd demand a lot more money for it.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Technical Details
Message-ID: <843@bnr-rsc.UUCP>
Path: arc1-ca.pugleaf.net!utzoo!utgpu!bnr-vpa!bnr-rsc!mlord
References: <37196@bbn.COM> <1989Mar16.190043.23227@utzoo.uucp> <24889@amdcad.AMD.COM> <355@bnr-fos.UUCP> <13@microsoft.UUCP> <362@bnr-fos.UUCP> <1989Mar25.013342.22214@utzoo.uucp>

Show Full Headers
Complete Article Headers
Path: arc1-ca.pugleaf.net!utzoo!utgpu!bnr-vpa!bnr-rsc!mlord
From: mlord@bnr-rsc.UUCP (Mark Lord)
Newsgroups: comp.arch
Subject: Re: Unaligned Accesses (was Re: How to use silicon)
Message-ID: <843@bnr-rsc.UUCP>
Date: 27 Mar 89 14:25:21 GMT
References: <37196@bbn.COM> <1989Mar16.190043.23227@utzoo.uucp> <24889@amdcad.AMD.COM> <355@bnr-fos.UUCP> <13@microsoft.UUCP> <362@bnr-fos.UUCP> <1989Mar25.013342.22214@utzoo.uucp>
Reply-To: mlord@bnr-rsc.UUCP (Mark Lord)
Organization: BNR -- Birthplace of The Digital Whatchamicallit
Lines: 38