Any information if the Coldfire runs with FPU and MMU?

The tool for converting 68k Assembly code to Coldfire maybe is from here:

MicroAPL Tools Overview

On this side is also a CF68kLib that can setup an 68k emulator on the coldfire that emulates an 68k without MMU and FPU for running 68k OS. It is also possible that the CF68k lib runs under a Coldfire OS and runs (slightly patched?) 68k programs there. At least that's what i read out of it.

Bye

_________________
<- Amiga 1260 / CD ->
Looking for:
A1200/CF CFV4/@200,256MB,eAGA,SATA,120GB,AROS

@Donar

Quote:

Any information if the Coldfire runs with FPU and MMU?

In one of the videos, Elbox said that they are using the MMU to do some PCI related things. There is a possibility to write a software that uses MMU but since it is not compatible with 68k MMU, the existing software won't work. Someone said that such software will be needed (for example for debugging purposes). Elbox answered that they are aware of it but most of the users won't need such thing (since user programs don't use MMU).

I haven't heared anything about FPU so I don't know.

I have made a small compilation of what was said on the first two videos when the Elbox guy explains some things. I'm sure that most of you don't userstand a thing so I wanted to help.

1. Dragon will be able to work with faster CPUs, also those which aren't produced yet but are planned.

2. What's on the main board.
2.1. CPU slot
2.2. AGP slot
2.3. 5 * full PCI 2.2 slots, 33/66 Mhz, 3 x 3V, 2 x 5V
2.4. The PCI Engine from the CPU is used, as opposite to the Mediator where Elbox's own PCI Engine was used (it was implemented in MACH chips there). As a side note: The CF PCI Engine is fully compatible with other Freescale CPUs, like PPC used in Efika.
2.5. Dragon has 2MB of flash memory. 1MB is needed to enable booting by the CF and enable Kickstart to run on the CF. The other 1MB is reserved for other purposes, for example booting from network.
2.6. The big Lattice chip. Its main function is to bridge the A1200 motherboard to the CF. The main problem Elbox come across when designing the Mediator was that in most of the A1200 accelerators there are only 24 CPU address lines available on the slot (which was an incompatibility with Commodore specification, there should be 32). This forced them to use "windows" technique there. On the Dragon, the CPU is on the same side with AGP and PCI so it can directly address these cards. The Bridge chip allows the CF to address the A1200 mainboard. Anything that is on the mainboard, can be accessed by the CF. And not only by the CF, also by the PCI cards. For example a PCI card can create its buffers in CHIP memory and exchange the data using CF's DMA-channels.
2.7. RTC clock.
2.8. The flash memory can be flashed by software. If something bad happends and one ends up with trashed/erased flash, the CF won't work (of course). In this case the 020 jumper can be used. It disables the Dragon and computer can be booted using the 020 CPU. In this case the flash memory is still visible and can be flashed too. So, the firmware flashing won't be dangerous anymore.
2.9. Other jumpers: flash write protection, RTC clock write protection, PCI Wake Up enable, PCI clock selection, LEDs, Power/Reset buttons.
2.10. The long on-board connector gives access to four fully programmable UARTs of the CF.
2.11. The small connector (which TBH I cannot find) is the JTAG connector for on-board chips.

3. What's on the CPU board.
3.1. MCF5475 CPU running at 266MHz (410 MIPS)
3.2. Two DDR-266 slots.
3.3. The small side-connector is a diagnostic connector. Using the CPU lines available on it the processor can be debugged and traced. For end users it's mostly useless.
3.4. The big side-connector gives access to additional functionality of the CPU. For current CPU this is two Fast Ethernet ports and USB 2.0. They plan to release an additional card which will connect to this slot and give access to the ports on the backplate. However, they said that most of the users will probably use PCI cards instead.

Again, this is just what I've understood from the videos. Some technical info like the CPU model was taken from elbox.com. The sound quality of the videos is poor and my knowledge is limited but I've done my best. I don't take any responsibility for any mistakes.

@yak
Thanks for the information.. was wondring what the connectors on the CF was.

@yak

SysInfo is useless for testing cpu speed, specially using emulators. I won't shoot the messenger, so thank you for your translations and information.

But I simply can't believe they plan to sell a cpu that runs 060 software slower than real 060. It's simply not acceptable for 2006.

_________________
The only spanish amiga news web page/club: CUAZ

@yak

Thanks a lot for compiling the information from the videos!

@Crumb

True. I forgot to mention that it was said the SysInfo results are useless.

There are also some talks about using AROS on CF. Marcik on PPA.pl said Elbox told them that they are working on it for some time now. For example on the possibility to use the A1200 chips from it.

@yak

AROS for Dragon would be a great idea. If they improved it enough they could make that Dragon could work without an A1200 connected to it...

Anyway... I can't stop thinking that it's slower than my current 060 so I wouldn't be able to watch scene demos that require 060 smoothly...

Last edited by Crumb on 13-Nov-2006 at 01:58 PM.

_________________
The only spanish amiga news web page/club: CUAZ

How terrible: One of the very few companies that can still produce hardware, wasting their resources in some STUPID product

@Crumb


Quote:

AROS for Dragon would be a great idea. If they improved it enough they could make that Dragon could work without an A1200 connected to it...

One step after the other, at least it seems to work, which some people just could not believe.

Quote:

Anyway... I can't stop thinking that it's slower than my current 060 so I wouldn't be able to watch scene demos that require 060 smoothly...

It should be way faster than your 060, at least for programs compiled for Coldfire. As for Demo watching... the ColdFire Roadmap has not reached it's end, IF it is not possible to "emulate" a 68060 at full speed now, maybe later.

_________________
<- Amiga 1260 / CD ->
Looking for:
A1200/CF CFV4/@200,256MB,eAGA,SATA,120GB,AROS

@Donar

Quote:

One step after the other, at least it seems to work, which some people just could not believe.

It seems to work more slowly than a 040@25 for 68k applications, according to people that were there. So no, it doesn't work as it was supposed to, to run unchanged 68k applications at fast speeds. *WHY* would you want ANOTHER emulator?

_________________

@Donar

Quote:

It should be way faster than your 060, at least for programs compiled for Coldfire.

I don't care much about coldfire specific apps. I care about my current software that requires 060 and won't be recompiled.

_________________
The only spanish amiga news web page/club: CUAZ

@Crumb

"I can't stop thinking that it's slower than my current 060"


You can't base that on any scientific fact, atleast until it's been finally released and properly benchmarked.

I agree with you that if the end result is something alot slower than a 060, then it really defeats the point downgrading to 030 like speeds.

Hopefully that isn't the case and the final result will demonstrate this. I think also the dragon is upgradeable with future coldfire CPUs, so things will only get better.

_________________
4000T/BFG9060
CD32/Elsat ProModule, TF360
CD32/ Edu's CD32 <> A1200 Adapter, Vampire V2
CD32/ FMV Module

@Akiko

Quote:

You can't base that on any scientific fact

My observation is empirical: ATM Dragon shows the icons much slower than my current 040/30 and my 060/50

They should patch OS3.x with AROS functions to give the user the illusion of Coldfire running 680x0 software faster than real 060s

Quote:

Hopefully that isn't the case and the final result will demonstrate this

can you base that on any scientific fact?

Why should future 68k emulation be faster than the one shown?

What's the real advantage in using a ColdFire if emulation is slower than the real thing? Couldn't they simply use a PowerPC with a 680x0 emulator?

A G3/600 using interpretive emulation runs faster than a 060... and software could also be recompiled from 680x0 to PPC easily using CoyoteFlux's PPC680X0 software.

_________________
The only spanish amiga news web page/club: CUAZ

@Colin_Camper

The videos show way doing presentation on a CRT monitor is a bad idea, they should have added a scandoubler to the package, did not understand what they where taking about, they way did they not add more power full PPC CPU on it, after all they need to run the 68k throw a JIT emulator anyway.

Guess this is yet, one more fork in the Amiga binary format, more to support; way can’t companies cooperate on projects?

Quoted from: yak

Quote:

One of the attenders said on PPA.pl that elbox claims that most of the unmodified 68k software runs at speed between 040 and 060 (because of the emulation).

Way oh way not a faster PowerPC chip? that supports OS4?

Quote:

There is a PC program that can convert 68k assembly sources to Coldfire ones. There are probably some talks behind the scenes with some Amiga software developers to recompile their software for CF.

I beat you can do the same whit a PowerPC chip, convert 68k to PPC binary files,

Quote:

It was also said that SysInfo reports CF as 68030 clocked at 1700MHz.

I don’t think SysInfo is where precise if you normally get speed between a 040 and 060 on normal 68k binary files.

Quote:

old software like NDOS games

I guess you can boot whit out Dragon.


I also noticed this http://www.microapl.co.uk/Porting/ColdFire/index.html
Linked to by Donar
Quote:

Other porting options As well as these options for ColdFire, MicroAPL offers a range of solutions and services for porting your legacy 680x0 code to other architectures or to C. We have versions of PortAsm which translate 68K code to run on the PowerPC, 80x86 and M*Core architectures, as well as emulators allowing 68K code to run with minimal porting effort on the PowerPC, M*Core and MIPS architectures.

Last edited by NutsAboutAmiga on 13-Nov-2006 at 04:46 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

The biggest problem with emulating unsupported opcodes is that exceptions are generated which takes a long time, so to get better performance you have to patch the executable during loading, like Oxypatcher/Cyberpatcher did, replacing the unsupported opcodes with JSR's to the emulation routines. This made a significant speedup on 060 CPUs. Do they already do something like that on the Dragon?

@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote: .............. Way oh way not faster PowerPC chip?

Maybe because those chips cost more.

Maybe because those chips are harder to intregrate into the Amiga-infrastructure.

Maybe because SW-support for PPC is still lacking comapred to plain 68000.

Maybe because a non-JITed 68k emu would be even slower on a comparable PPC.

Maybe because Elbox's changes of getting support from one of the PPC-OSes are on the same scale as those of a snowman surviving in the middle of the Sahara.

Maybe because Elbox tries to make market out of those who have a different opinion on what "a real Amiga" is.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

Quote:

Maybe because those chips cost more.

Maybe, but maybe your just making useless excuses.
Quote:

Maybe because those chips are harder to intregrate into the Amiga-infrastructure.

Don’t think so.

Quote:

Maybe because SW-support for PPC is still lacking comapred to plain 68000.

But ColdFire CPU can’t run 68000 code whit out emulation, and ColdFire CPU have less support then PPC chips anyway, at this point in time.

Quote:

Maybe because a non-JITed 68k emu would be even slower on a comparable PPC.

If the JIT compiler is put in side the firmware, there are absolutely no need for full scale PPC OS.

Quote:

Maybe because Elbox's changes of getting support from one of the PPC-OSes are on the same scale as those of a snowman surviving in the middle of the Sahara.

So way not drop NG OS support and just put a JIT compiler in to the flash, if they Hyperion or MorphOS reconsider they can port there OS to it.

Quote:

Maybe because Elbox tries to make market out of those who have a different opinion on what "a real Amiga" is.

Look I’m not flaming the products for ever thing, there are many things that great whit it, but non of that has any thing to do whit clockports or zorro slots, what great about the products is in fact the PC standards.

This quoted from Amigaweb / DRAGON Technical Specification:
Quote:

Supported AGP and PCI cards: * Graphic cards (Radeon 9700, Radeon 9500, Radeon 9200, Radeon 9000, Radeon 7500, Radeon 7000, Voodoo 5, Voodoo 4, Voodoo 3) * Sound cards (Sound Blaster 4.1 Digital, Sound Blaster 128 and cards based on the ForteMedia FM801 chipset--e.g. Terratec 512i Digital) * Serial ATA controllers (SiI 3114 support in development) * Hi-Speed USB 2.0 controllers (Spider and Spider II) * TV tuner cards (based on Bt848/849/878 and FUSION 878/879) * Fast Ethernet 100Mbps cards (based on RTL8139 chipset) Ethernet 10Mbps cards (based on RTL8029 chipset)

This product is designed for the classic Amiga computers just to get access to the custom chips for legacy, whit out it the custom chips it will just be an other product that do not directly support Amiga Games whit out UAE, it’s quite similar idea as the ACK design idea, and some original AmigaOne design concepts, as I no loner own an classic Amiga system, I have hard time convening my self that buying an a computer designed in 1990 (or even older), and buy a ColfFire on top is going to be worth by pocket money, when basically I can run the same games on E-UAE whit almost the same game play.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

Quote:

It seems to work more slowly than a 040@25 for 68k applications, according to people that were there. So no, it doesn't work as it was supposed to, to run unchanged 68k applications at fast speeds. *WHY* would you want ANOTHER emulator?

I somewhere read that speeds were between 40 and 060 speeds depending on application, whatever that means (060@5MHz? ) . The Coldfire is no 68k, so i used the term "emulate" for it running 68k code, don't know if there is some better term for it, but would like to know if there is a better "explanation".

Bye

_________________
<- Amiga 1260 / CD ->
Looking for:
A1200/CF CFV4/@200,256MB,eAGA,SATA,120GB,AROS

@COBRA

The problem is, exceptions will not handle incompatible (yet implemented) opcodes. Those will have to be replaced and unless if they are replaced at run-time and NOT at load-time, self-modifying code will fail.

_________________