| From: | Jesse McClusky |
| Date: | 11 Sep 2000 at 19:32:53 |
| Subject: | Re: Subject: Re: [RE: AMIOPEN: has anyone run SPEC benchmark using VP?] |
> From: "James Russell" <amiga401@netscape.net>
>
> > > The fact is that we do not know how well VP works compared to natively
> > > compiled programs.
>
> I DO know that VP code looks for a native tool FIRST (they are marked as 0,
> where non-native is 1 and above).
Um, actually, VP code is marked as 0. 1+ is specific processor-native code.
> I don't know how far this goes, so concievably you could use
> entirely native toolsets. If you were only ever going to use one
> processor, it would then make sense to do this(?).
>
> Is this the case, anyone? Can you go entirely native with VP code? Should
> you?
Can you? Yes.
It's probably preferable in embedded systems to use as much native
code as possible, except in what you'd call the "automatic device driver"
portion -- the chunks that get sent to nearby devices that want to use it.
The only other reason to use native code is if you have a very
processor-intensive task, and the VP translator isn't optimizing the
code enough (such as with certain 3D tasks, currently, on the P3).
Jesse
Subscribe/Unsubscribe: open-request@amiga.com
Amiga FAQ: http://www.amiga.com/faq.html