| From: | Martin Baute |
| Date: | 31 Aug 2000 at 13:28:45 |
| Subject: | Re: AMIOPEN: Objective C on Amiga? |
Hello Zoltan
On 29-Aug-00, you wrote:
> Every now and then I seem to find web pages devoted to ObjectiveC,
> apparently it has some advantages over C++ and I know its being used on
> MacOSX (thanks to the NeXT connection). What's more, apparently it is a
> "pre-processor" for a regualr C compiler, so that it be run before
> something like GCC. Anybody had a look into this language?
I had a rather heated discussion with Ralph Schmidt ("laire", of Phase 5
and MorphOS "fame", excuse the quotation marks, they are IMHO) on IRC
#amigager. He also mentioned Objective C.
He claimed he would use Objective C if he "would need OO at all".
One of the points I remember: If creating a (shared) library, adding a
method to a (C++) class would break the library code for existing
binaries, which would not be the case for Objective C. (I cannot
confirm or deny this myself.) Objective C seems to be used extensively
e.g. in GNOME code.
BTW, there were several other points made in this discussion worth
being mentioned here.
I will quote him in full. Note: He said he must not be quoted, because
he has "no intention to be pursued by Amiga Inc. fans", and he wants to
avoid "quotes out of context". Quite frankly, I think his views of the
people on this ML are quite twisted, because I believe we are not as
fanatic as some in the community. Second, I think talking this kind of
talk on IRC and then denying permission to quote is... well... stupid.
Please do not prove me wrong by attacking him because of what I am
quoting here.
So here�s the transcript, in full. (My nick is "Krise".)
*** start of transcript ***
Laire: Let�s have a look at the pcode... practically useless for efficient
driver
development... desktop computers are sold by CPU performance nowadays...
I doubt you can win costumers with slower pcode emulation / transcription.
Pcode takes the descission over the target market away from the game/app
manufacturers... this is not acceptable for these companies. Moreover, for
the desktop this will only work with a host OS, because it lacks the whole
basis. What do you want with a Linux/Bsd/Win** with a pcode """candy"""
layer... Elate is a system developed for embedded... small RAM footprint
etc.,
and is simply not suited for the desktop.
Krise: I�m not convinced.
Laire: So you see a market for a pcode system on top of a host OS xyz,
which is slower and without benefit for the desktop user? (Java has it�s
uses, because it has secure access, thus employable in a network... this
is not the case with Elate)
Krise: I consider it being possible if not likely to remove the need for a
host OS, in the medium run. (more)
Laire: And where is the network layer, drivers and apps coming from?
Modern driver development with Elate is a joke, because it doesn�t know
about cache.
Krise: Who said that low-level drivers have to be in Elate-VP? "takes
the descission over the target market away from the game/app
manufacturers... not acceptable..." <- cannot see the logic behind that
Laire: You are company X... this company X gets a license for... say,
Heretic2. This licence is for target system x only. Now, the layer is
suddenly available for system y, too (e.g. Windows). And you have a
*serious* potential for licencing conflicts.
Krise: Where�s the problem, if x = Elate?
Laire: What if target system X allows for higher prices than target system
Y.
The irony with the whole thing is that Elate allows to compile software
native,
which is probably what all these companies will be doing to avoid these
troubles and get real speed; this would bring down the basic philosophy...
Krise: Fine, your opinion. However your arguments aren�t that striking,
and partially can be solved without major technical effort.
Laire: Yes? Then counter them, specifically.
Krise: E.g. driver performance can be achieved by native compiled drivers
with according interfaces. Do you agree?
Laire: Sure... but *Elate* advertises pcode driver development
Krise: So why not? If the translator already offers decent cache usage? Or
the driver is not performance critical?
Laire: How do you *virtualize* e.g. vector engines to a sufficient degree...
therefore there will be no true usage of SIMD engines... essential today.
Laire: sigh... read about e.g. writecombining, partialbursts in current 3d
drivers. Such things are extremly CPU relevant.
Krise: As if _any_ driver as of today makes use of e.g. Altivec.
Laire: This was about applications. Moreover, the Altivec streaming
extensions are also useful for drivers.
Krise: We were talking about drivers, please don�t jump. I never said that
Altivec wouldn�t be great for drivers, but it�s not (yet) made use of.
Laire: Altivec, SS2... they are so much different you cannot virtualize it
sufficiently... as in defining in pcode. So pcode is completely useless for
this.
Krise: Correct, and already recognized by both Tao and Amiga.
Laire: They want to solve it by *library*... this however solves only
certain
standard problems.
*** end of transcript ***
Regards
Subscribe/Unsubscribe: open-request@amiga.com
Amiga FAQ: http://www.amiga.com/faq.html