Subject: Re: mac68k Packages for 1.6
To: None <port-mac68k@netbsd.org>
From: Bryan Vyhmeister <bsd@hub3.net>
List: port-mac68k
Date: 10/31/2002 13:26:10
As far as 68060 systems go, is this the sort of machine that works 
pretty well for this sort of thing?

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=2065023787

Thanks for all the info.

Bryan

On Thursday, Oct 31, 2002, at 13:11 US/Pacific, John Klos wrote:

> Hi,
>
>> I was wondering about how package builds would be done. I just 
>> compiled
>> standalone-tcsh on my SE/30 and it took around two to three hours. I
>> started a kernel compile last night and it is still going almost ten
>> hours later.
>
> Yeah, the 68030 is a nice chip, but the 68060 is easily an order of
> magnitude faster. It's even more superscalar (will dispatch more
> instructions per MHz) than the Pentium Pro (II, III). The P4, btw, is 
> even
> less superscalar; they just wanted faster clock rates for marketing
> purposes. The 68060 would've been a very nice chip had they ramped it 
> up
> to multi-hundred MHz.
>
>> I was also wondering if it is possible to cross-compile packages for
>> m68k? I have a a number of servers in my own rack here at home and I
>> would like to help out with some of the open source projects around. I
>> have a 1Mbit link to the internet so I thought maybe I would run a
>> mirror or something but I'm still not sure what would be most helpful
>> to the NetBSD project.
>
> Well, cross compiling is difficult. It's possible with the NetBSD 
> sources
> because, well, we're NetBSD, and that's one of our goals. To try to get
> everyone else out there to write clean enough code that we can cross
> compile it reasonably is not very realistic - most of the code out 
> there
> is written assuming that every processor in the world is some i386
> variant. A lot of that code won't compile without change on other
> processors, hence the need for pkgsrc patches and stuff.
>
> One way, although it's not really cross compiling (I suppose it could 
> be
> called that, but that'd be cheating) is to run a full m68k emulator on 
> one
> of your fast PCs. An example is Amithon, which is a heavily optimised
> Amiga emulator. While Basilisk and other emulators use C code for the 
> m68k
> emulation, Amithon had lots of assembly to make it fast. Many people
> running Amithon on 1.5 GHz and faster AMDs say that it's noticeably 
> faster
> than a 50 MHz 68060. So if one were to set up that with NetBSD and do 
> bulk
> package builds, that might be good.
>
>> What sort of bandwidth is needed for your 1U Amiga?
>
> Not much. I'd need a static IP (or at least one that doesn't change too
> often), download bandwidth to fetch the sources for the packages (not 
> too
> much needed), and upload bandwidth to upload approximately 1.5 to 2 
> gigs
> worth of packages every couple of months.
>
> More than that, it should be in a real rack. It's on shelves with 
> other PC
> and PC-like cases, and it's not an ideal situation. It should just be 
> more
> permanent.
>
>> Thanks for the info. That is quite interesting.
>
> Sure, no problem. I'm writing an article about the building of the 1U
> Amiga (it was a lot of work), and it will be in Total Amiga magazine. 
> I'll
> post the URL to the .pdf version when it's available. (Incidentally, I
> host Total Amiga's web site on another Amiga server that I run:
> http://www.totalamiga.org)
>
> John Klos
> Sixgirls Computing Labs
>
>