"New" amiga hardware

[biggun]

http://www.apollo-core.com/bringup/

[biggun]

@agami

It's not really practical to do that level of detail, and I don't think it should be done.

For one thing, the blog would be the most boring ever. Imagine it being like this:
"Today I made no progress. Tracking a bug."
"Still tracking a bug."
"another day. where is it?"
"Found a missing equals sign. Hurrah!"
"Still doesn't work. D'oh!"
"Ripped out huge chunk of code. Too much like spaghetti."
"Bug gone. Now had to write data structures for input handler."
"Wrote input handler"
"Found bug and squashed it"

It'd be a proper yawn-fest, so nobody would read it anyway.

Also:


I don't know of anybody who blogs to this level, and even if someone does, it doesn't mean the rest of us have to.

Nearly all Amiga work is done part time now. You don't want to see lots of blogs saying "Nothing got done today", but that is exactly what you would see.

Also, I don't see why a developer "should" do anything. The developer is not employed by a user, nor is he behooven to him on any way. If Toni Willen does it then great, that's nice, but that's just his personal choice.

Put simply, a developer can do whatever he wants in the manner he wants, it is not for the user to decide what he must and must not do. Trying to force rules and regimes on people who are basically just doing it for a hobby and love of the platform more than anything else is just going to push what developers we have left away.

Very good write!

[biggun]

http://www.apollo-core.com/bringup/

Newsflash:
No real news.
We are still testing.
To make sure that all the games really work fine we have to play them to the end of course.
Therefore our testrate per day is limited.

[biggun]

Update:
new progress video uploaded
http://www.apollo-core.com/bringup/

Is this way of updating the progress what people want?

[biggun]

wery well, but couldnt you finally post screenshots of the current amiga benchmarks?
as it looks like the core runs amiga games just fine, im assuming this is 68000 software. what is the area to work on now, are there still missing instructions?

To answer your questions:

are there still missing instructions?

* All 68000 instructions are included
* Many good 68020 instructions are also included
* Some 68020 EA modes are included.
* Some usefull new instructions are included.


what is the area to work on now?

We test, test, and test again.

We want to make sure that all is working fine.
For this a lot of testing is needed to debug and verify that everything works perfectly.

Also we measure and benchmark.
Today we measured intesively the bus cycles on the chip memory bus.
We saw some occational bubbles there and work now on removing them.
Many AMIGA accelerators suck on their chip bus connection.
We aim to be better than the others.

We will publish benchmarks as soon as we have finished testing.

[biggun]

Is there any hardware that will be available -- as in ready and shipping -- in October or November?

Majsta has 2 cards:
* Vampire 600:
   useable in A600
   + 64 MB fastmem

* Vampire 500:
   useable in A500 (maybe also in A1000/A2000)
   + 64 MB fastmem

Vampire 600 is currently produced and sold by Kipper
I assume if people show interest then Vampire 500 could be produced and sold asap.


We have currently 1 card:

APPOLLO/PHOENIX 500
  useable in A500 /A1000/A2000
  + 128 MB
  + new 28nm FPGA
  + SDCard
  many more features ...

Of the Apollo/Phoenix card a bunch of beta test card was produced.
We test it right now.
As soon as testing it fully complete, card will be sold.

[biggun]

Update : Added Chip memory bus test benchmark results:

http://www.apollo-core.com/bringup/

[biggun]

Update:
Uploaded several Videos showing games running on CPU.

http://www.apollo-core.com/bringup/

[biggun]

chip ram access seems normal as expected

The result of 3.5 MB/sec is actually the best possible result on a 16bit AMIGA.
I'm not sure if this is normal to reach.
What do other cards reach on 16bit AMIGAs?

[biggun]

Hi Gunnar,  

yes, thanks for the update. I guess we talked about this already a bit. The problem with that is that there is no API to signal an "almost but not quite" 68020,

We use the GNU tools ourselves to write software.
We have added a new target "APOLLO" to them to be able to code for our Core.
This target matches our CPU and allows not onle to use the new instructions
but also to enables to new address mode options we have.

Phoenix support PC-relative also for updates.
Coding like this now works fine.
ADDQ.L #2,label(pc)

I - as a user - cannot install a 68020-only version of a software because I cannot be sure if the result works.

This is true.
For the first release of Phoenix we only promise 100% 68000 compatility.
But as most 68020 instructions work fine - many 68020 programs might run fine too.

At a later point we will add more here to reach full compatibility.

My idea would be that you provide some "unimplemented integer instruction" exception and an "unimplemented addressing mode" exception (whoops, there are already!)

This is of course another option and would work even today.
Good idea.

[biggun]

i think its self explanatory that it should be done as thomas proposes, the system should be usable at all times, even if non existent instructions lead to exceptions that are handled by a cpu lib instead crash.

That a CPU trows an excpetion for unimplemented instructions is normal.
Phoenix does this already.
And old 68000 cores did this also.


Nevertheless on AMIGA really no one cared about this.
In theory you could catch all FPU instructions  anfd run programs using the FPU this way on any AMIGA.
Did anyone do this?

[biggun]

Can't speak for 16-bit (I don't remember what my borrowed 040 did), but my CS060 MK1 was consistent 3.5/7 in my 4000. It was considered a good implementation AFAIR.
I think some of the 1200 accelerators were struggling...

Yes, according to this database no AMIGA card was able to saturate the bus.
http://amiga.resource.cx/perf/sysspeed.html

A4000/CyberStorm                   Chipread=4.4 of 7.0 possible
A1200/CyberStorm III               Chipread=3.3 of 7.0 possible
A2000/G-Force 030 (030/50)  Chipread=2.0 of 3.5 possible
A600/Apollo 630 (030/40          Chipread=1.9 of 3.5 possible

Looks like the Vampire from Majsta is the first card for the AMIGA which reaches 100% buspeed.

Does someone has some results for other cards that you can still buy today?

[biggun]

some more updates as promissed

http://www.apollo-core.com/bringup/

[biggun]

On the Sysinfo benchmark would it ever show up at a faster speed? Can it be optimised further?

Yes this is a debug/development build only.
I think in the end about 10% better is reasonable to assume.

This result is the same for the Vampire 600 and the Vampire 500.
The Apollo/Phoenix card is a lot faster though....

[biggun]

Why some developers choose the amiga 600? Is not popular like 500 or 1200.

Why not?

We have three card supporting Phoenix.

1) Vampire 600 (for Amiga 600)
64MB fast

2) Vampire 500 (for Amiga 500/ and others)
64MB fast
+ IDE Controller

3) Apollo/Phoenix (for Amiga 500/ and others)
128MB Fast
+ SDCard