Vampire 600 more cores..

[wawa]

Originally Posted by cgugl

At the moment, only Christoph Hoehne has a Cyclone 5 board and is trying to compile Phoenix... he told me that he need to write the chipmem controller (seems that the ram controller is a bit different from the controller on the vampire600)...

well, my understanding was, that few apollo board prototypes (http://www.apollo-core.com/fpgacards.png), that means the adapter boards for the sockit boards, were commissioned few months ago and should be in hands of developers since long?

[cgugl]

Originally Posted by wawa

well, my understanding was, that few apollo board prototypes (http://www.apollo-core.com/fpgacards.png), that means the adapter boards for the sockit boards, were commissioned few months ago and should be in hands of developers since long?

yes, few PCB have been commissioned one or more months ago.
Now, only one of them is working, because the pcb requires the soldering of some components and that Christopher writes a good memory controller (there is no need of new apollo cards until Chris finishes his work)


So Gunnar is improving the core Phoenix on the small vampire 600 (this is not a waste of time, because the same core that works on the vampire 600 can be (we hope) ported on the new C5 board, and run faster. All the improvements we have on the vampire 600 will give benefits also to the C5 apollo board)

[Reido]

Exciting stuff!

[jbenam]

Still no answer on ISA compatibility though

Will the ISA be based on 68000 or 020? (I seem to remember reading it was a franken-ISA based on 000 with some additional 020 instructions?)

Is it 100% complete? No missing/unimplemented instructions?

[cgugl]

Originally Posted by jbenam

Still no answer on ISA compatibility though

Will the ISA be based on 68000 or 020? (I seem to remember reading it was a franken-ISA based on 000 with some additional 020 instructions?)

Is it 100% complete? No missing/unimplemented instructions?

Now Phoenix is compatible with the 68000, and has many instructions from the 68020.

We lack 1 instruction in the 68000 set (moveP). It is a speed choice, because this is a rare instruction that requires 2 cycles of clock (all the other instructions needs only one cycle).
The missing instruction Movep is emulated through an external library (like on the 68060)... the compatibility is very good (i can't say total but...).

We have many (most) instructions from the 68020 set, the missing instructions can be emulated through the external library (yes, like on the 68060).
Phoenix has plus native instructions than the TG68_020, but we can't say to have the 020 compatibility because an important decoding is missing in phoenix (but Gunnar is working on this)...

to know what is missing, i need to ask to Gunnar !

[alexh]

Is this enough to be detected as and run the majority of 68020+ software?

You guys know much much more but I seem to remember the stack frame being different. Lots of new registers such as VBR etc.

[cgugl]

[QUOTE=alexh;997387]Is this enough to be detected as and run the majority of 68020+ software?

yes, i know something because i have a vampire 600, i was on the irc channel in the last 6 months, and i tested all the gunnar's core !

About the new registers, you can use them only if you write your code in ASM ... if not, phoenix is a plain 68000

I don't know if phoenix is capable to run 68020 programs, many of them do a check and refuse to work...
I asked a patch to the kickstart to recognise phoenix as 68020... we will see what happens

[alexh]

Originally Posted by cgugl

About the new [instructions/]registers, you can use them only if you write your code in ASM ... if not, phoenix is a plain 68000

I think I understood.

You currently have use plain vanilla 68000, or write specific "pheonix" ASM code to use the 020 extensions because pheonix is currently too different to a real 020+ at the moment?

[cgugl]

Originally Posted by alexh

I think I understood.

You currently have use plain vanilla 68000, or write specific "pheonix" ASM code to use the 020 extensions because pheonix is currently too different to a real 020+ at the moment?

- If you compile the code for 68000 (ASM or GCC). It must work.

-If you compile the code for the 68020 (ASM or GCC), maybe it works (very difficult if Gunnar doesn't add the missing EA mode).

- If you want to use the new registers and some specific instruction to perform chunky to planar and planar to chunky in two cpu cycle, You must write your code in ASM.