🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.sys.dec
Article #2836

Re: alpha memory requirements (vs vax) ?

#2836
From: Petzi Schweda
Date: Tue, 10 Nov 1992 09:17
44 lines
2794 bytes
In article <1992Nov9.172318.22005@unislc.uucp>, dold@unislc.uucp (Clarence Dold) writes:
|> From article <JOHN.92Nov9101505@sekrit.WPI.EDU>, by john@sekrit.WPI.EDU (John Stoffel):
|> 
|> > (Clarence Dold) writes:
|> >> ... experience in moving code from a Motorola
|> >> 68040 to the Motorloa 88100 RISC set shows a substantial change in the
|> >> requirement for disk space for executables, as well as run-time memory.
|> >> I recall a comparison of ~20% for like code.
|> 
|> > It's been pretty good here (we're a Beta site) for compiled binary
|> > sizes.  You really don't notice shared libraries for most code.  Where
|> > it really shines is for very grahics intensive code.  We had one guy
|> > who went from a 1.2 Mb on disk program to 90K on the Alpha.  
|> 
|> I think we're talking about two different animals at once.  Going from
|> non-shared libraries to shared can be a major win, especially if your
|> X11/Motif libraries are shared (which they are not on Unisys Motorola boxes).
|> 
|> The other issue is the size of the executables (perhaps just .o should be
|> compared) for a CISC machine verses a RISC machine.  
|> 
|> In joining this thread, I am assuming that the VAX, non-alpha that we are
|> talking about is a CISC machine.  It appears that OSF verses VMS might also
|> be part of the conversation, but on a different thread.
|> 

Assuming that the organization of your code (shared VAX/VMS-libraries,
shared Alpha/OpenVMS libraries) keeps the same, i was told that the space 
needed on disk (as well as runtime-memory) should aproximately be DOUBLED.
This is because complex (CISC-) instructions like movc3, cmpc3 simply don't
exist on a RISC-architecture machine, so the compiler has to generate extra 
code for it.
Even worse, when adepting your c-sources (structures !) to 
long- or quadword-alignment, the amount of dynamically memory needed can 
increase realy dramatic! If you don't believe that longword alignment
is neccessary, take a look at what you (or your compiler) has to do to
manipulate bytes or words on a ALPHA ...

Hard times ahead ...
regards, Petzi
-- 
----------------------------------------------------
Petzi Schweda (sca@gese.ge14.mdadv.gv.at)
MD-ADV, Municipality of the City of Vienna, Austria
Technical Details
Message-ID: <1992Nov10.101757@gese0n.ge14.mdadv.gv.at>
Path: arc1-ca.pugleaf.net!sparky!uunet!mcsun!hp4at!gese0n.ge14.mdadv.gv.at!sca
References: <JOHN.92Nov9101505@sekrit.WPI.EDU> <1992Nov9.172318.22005@unislc.uucp>

Show Full Headers
Complete Article Headers
Newsgroups: comp.sys.dec
Path: arc1-ca.pugleaf.net!sparky!uunet!mcsun!hp4at!gese0n.ge14.mdadv.gv.at!sca
From: sca@gese0n.ge14.mdadv.gv.at (Petzi Schweda)
Subject: Re: alpha memory requirements (vs vax) ?
Message-ID: <1992Nov10.101757@gese0n.ge14.mdadv.gv.at>
Sender: usenet@hp4at.eunet.co.at (USENET - Admin account)
Nntp-Posting-Host: gese.ge14.mdadv.gv.at
Organization: MD-ADV
References: <JOHN.92Nov9101505@sekrit.WPI.EDU> <1992Nov9.172318.22005@unislc.uucp>
Date: Tue, 10 Nov 1992 09:17:57 GMT
Lines: 44