Amithlon... HELP!

The AC97 mixer might be what (in combination with some other software) causes the recoverable errors... When I ran it while developing the debugger/monitor/development too for Umilator, it consistently produced "zeropage" accesses (i.e. almost certainly dereferencing a NULL pointer); IIRC, that included writes into that area.

In and by itself, that's relatively harmless (except for bytes 4-7, none of the bytes are used for anything by default, so *one* faulty piece of software often seems to work just fine). It's only in combination with something else which makes the same mistake that you run into almost-impossible-to-track-down trouble...

> AMIthlon is the 1st system to use P96 3.
Not really --- despite a lot of (continued) advertising to the contrary, Amithlon still uses P96 v2, just like "everybody" else. Check the versions on some of the P96 files
>68k JIT on x86 runs 80% speed of native. [ and PPC 75% ]
That's way way way too broad a statement (for either case). I'd estimate that *in general use, overall*, the JIT in Amithlon runs at 25-35% native speed for integer, and a bit less than that for floating point. Quoting absolute-best-case scenarios for JITs (like RC5 is just not really useful (on either host CPU), and AFAICT, there are no commonly available 68k-to-PPC JIT application benchmarks at this point. Potentially, the PPC should be able to do a bit better than the x86; Whether that potential was realized --- who knows?

A general comment --- is it really necessary that everytime someone mentions the word "Amithlon", the same old arguments need to be repeated?

All this crap has been discussed to death! A few more "It's killing the Amiga" "Did you know all this stuff was developed on emulators" "It stops people from buying the new hardware" "What new hardware? And it rather stops people from leaving altogether" won't change anyone's mind. If we were talking about a multi-million user market, someone might decide to do a scientifically sound survey. It's not, and so we will always be stuck with guesswork.

And in the end, it all comes down to personal freedom. Some people think emulators are great. Other people think they are evil. Yet other people cater to the first group and try to sell them stuff, and a fourth group tries to sell stuff to the second group. And those groups are by no means mutually exclusive.

I like Coke. Other people like Pepsi. Coca Cola is selling stuff to me. Pepsi is selling stuff to other people. If someone shows up online and asks what people think of the new PepsiTwist, or where to get PepsiBlue locally, does telling her that drinking Pepsi is evil, that she should be drinking Coke, and just think of all the good Coca Cola did for the Atlanta Olympics, what has Pepsi ever done for us --- is that really going to help *anyone*? Or does it just scare people away from the online forum (thus subverting the whole point of it!), and ultimately pushes them towards the (much healthier) Orange Juice? (Yes, I was overstretching the analogy at the end. So sue me

Sheesh, I must be getting old. I am writing "can't we all just get along" posts about a subject which I obviously feel passionately about

@umisef

Lombi's post Edited by Targhan: Personal Attack.  

For your information morphos on pegasos is the
1st currently available ppc/68k jit amiga os
available and its jit speed is 75% that of native.

And amithlon beats it.

As for the p96 version thingy, I think the
authors of it would know

I think umisef would know... he made Amithlon after all ;P who knows better then somethings creator.

[Kronos]

@mips
Ta-sching  :-P

@Lombi
1) the MOS-JIT is still only availablöe for those under NDA
(unlike me) and those aren't allowed to give any details.

2) Sure it gets beaten in CPU-intensive tasks, which is no
suprise when you look at the CPUs typical used for Amithlomn.

3) SW doing lots of GFX is faster due the direct access via CGX5
instead of going through the "mess" of CGX-emu,GFX/layers,P96
and linux-drivers.

4) SW calling lots of OS-functions (like for example GoldEd )
run faster because of the native OS, somthing Amithlon lacks.

And yes I got both systems (G3 and AthlonXP1600) so I DO know
what I'm talking about.

I had an xp1600+. I only just upgraded to a
xp2200+.  The xp1600+ is about 6x times faster
than a g3 pegy & it even out peformed it on the
overall gfx speed by far. I totally reject your
argument and comparison. G3 is slow old crap.
Why do you think IBM are investing in AMD now ?
Its just a case of a good OS selling ordinary h/w.
No reason why it cant be on a standard x86 m/b.
You could even plug a sonnet ppc card in an x86
m/b just for the hell of it & hav 2 cpu's.
You do realise tha morph is partialy derived from
AROS x86 ?

[Kronos]

6x faster doing what ? Cracking RC5, or some other stupid benchmark ?
 :lol:  :lol:  :lol:

This was supose to be an amithlon help topic
so I will not respond to any more arguing or
system comparisons from bias people.

Lombi I agree G3 and PPC in general suck... but  it's not nessisarily true that a 68K JIT performs slower clock for clock on a G3... I would bet anything the X86 XP class Athlons stomp any G3 in emulation... heck I'd bet the P4 does to...  but thats not the point... I think umisef and the rest of us are thinking of a clock-for-clock ratio here...

But after all he made Amithlon for X86 for the speed and genericness of it anyway so it's all a null-argument.

[bhoggett]

@Lombi

You seem to be suffering from a serious case of foot-in-mouth syndrome. All the way past the ankle, I'd say.

Your case isn't going to be boosted by calling Bernie an "arrogant moron", and your claims about its performance are not going to have more credibility than those of the author himself.

You have made a number of incorrect statements in this thread, so my suggestion is that you try to read and learn so as to improve your knowledge rather than lash out at anyone correcting you.

>You are an arrogant moron.
Thank you. And you are a very polite and well-mannered individual with expertise that I can't possibly hope to match....
> For your information morphos on pegasos is the
> 1st currently available ppc/68k jit amiga os
> available and its jit speed is 75% that of native.    And amithlon beats it.
Doing WHAT? Running the distributed.net client for RC5? RC5 is *the* ideal case for a 68k-to-whatever JIT compiler --- its inner loop is extremely long. It doesn't contain *any* conditional instructions. *All* instructions in that loop operate on 32-bit sized operands. Of the many dozens of instructions, only *one* produces flags that ever get looked at (the comparison at the end).
Real life code is *very* different. Typical length of code before a branch instruction (or some other conditional instruction) is reached is 4-6 instructions. Unless you do *very* clever things (which will cost you in JIT compile time!), you cannot optimize away flag generation across such instructions. Furthermore, typical 68k code will happily use 8 and 16 bit values, write small sizes to a register and then read the same register at bigger sizes, and do lots of other things that will slow down JIT-ed code. So you will have to excuse me if I want to see some real-world application benchmarks, not some vague and unspecific claim in a PDF file. Oh, and as for Amithlon beating it --- I happen to know a bit about Amithlon, and I can tell you that a PPC should get closer to native for a number of reasons in at least non-flag-heavy code.
>As for the p96 version thingy, I think the authors of it would know
I agree. I apologize, I failed to introduce myself --- Name's Umisef, pleasure to meet you. In real life, I go by the name of "Bernd 'Bernie' Meyer", live in Melbourne, and have spent a lot of 2001 programming a small program you might be familiar with....

Hello Bernie (beginer)
                      Yes you should know alot
more about amithlon than me
But the 75% figure came from the morph people
Thats would have to the best response I got
from you since all the email i sent you.
Alot of your terminology is meaningless to me.
I made my conclusions by using AMiGOD to make my
speed comparisons.

                     

My appologies

I think I recall reading p96 3 on your previous
AMIthlon web site I have that backed up some
where but there is no point in taking this
discussion on any further.
Once again my appologies.

You didn't see "p96 v3" mentioned on *my* website. It was (and still is) on H&P's www.amithlon.com (go to English->Products->Amithlon), but hey, if they want to add false advertising to the long list of things they have screwed up while handling Amithlon, that's their choice.

Of course, a certain individual often referred to as "Lord Voldemort" has ideas of his own --- according to him, he and he alone defines what "P96 v3" is or isn't, and he and he alone decides who does or doesn't get it. See http://www.ann.lu/comments2.cgi?show=1014129779&category=files&number=67 for some rather amusing reading.

But even given Lord Voldemort's view, you will notice that he talks about "doing" v3 work, not about "having done" that work --- in February 2002, fully 4 months *after* Amithlon was released. So --- no v3 in Amithlon, even if one agreed with Voldemort's weird views (which, I think I am at liberty to say, the _authors_ of P96 don't).