04 June 2020, 22:11 | #1 |
Registered User
Join Date: Jun 2018
Location: Calgary/Canada
Posts: 247
|
The Legality of The Vampire Itself
I know there's a thread on the legality of CoffinOS that was recently shut down and some innuendo's dropped about the general ethics of the Apollo Team and why one should boycott them. As a former Apollo Team member, I can say that there's more than the Kickstart and OS being ripped off here.
This may all be old news -- if it is, I apologize up front. Way back in 2003, the original NatAmi project was looking to be a 68030 accelerator for the C-One, an infamous FPGA based computer meant to replace the Commodore 64. Around 2008, the Natami was going to use the 68060 processor instead, the last and fastest 680x0 processor, but even then, these chips were in short-supply. There was some debate about using the ColdFire instead, but it's incompatibilities with the 680x0 and closed-source nature of AmigaOS made it an unrealistic proposition. And then IP Extreme released the ColdFire V1 "free". You could sign up, download the core and program it to you Altera Cyclone. And, if you happened to have the "key", you could use Quartus to reverse engineer the binary file and get something close to the source. Then, Apollo announces they're going to drop ColdFire AND 68060 and make a 68060-class clone in FPGA instead dubbed the 68050 and 68070. Almost overnight they apparently had something they could show some relative performance of, as if, by **MAGIC**. Now, whether he actually reverse engineered the free core or used that as an excuse to use source he acquired while working for IBM, I don't know -- there is evidence supporting both of these possibilities. Of course the ColdFire core, whether it be a soft or hard is still incompatible with the 680x0. Which brings me to Exhibit "A": https://community.nxp.com/thread/307326 "While creating a super scalar Coldfire ISA-C compatible FPGA Core implementation, I've noticed some possible "Enhancements" of the ISA." And proceeds to explain exactly what you do to a ColdFire core to make it much, much more compatible with the 680x0. He does go onto claim that it's not based on the "V4" (true) and his own design (suspect). This interestingly also coincides with Majsta dropping the TG68K core for his own Vampire 600 accelerator, or V1 as it's called now. Later cores that became "closed source" ran substantially faster than the TG68K and yet still fit on a rather small FPGA core. It wasn't long after that when we started seeing the first prototypes running that rotozoomer and we were all like "woah".... The second bit of evidence, Exhibit "B" comes from an interesting bit of source code in the Apollo Team FTP called pheonixXXX.s. This little bit of source code looks very much like a stripped-down version of MicroAPL's ColdFire 68K "emulator" (documented here: http://microapl.com/Porting/ColdFire...d/CF68KLib.pdf) which uses the processor exceptions to handle the unimplemented instructions and/or addressing modes. Specifically, you can see the addressing mode handling, bitfield instructions ... all put into exception handlers because these instructions did not exist on the ColdFire at all. https://pastebin.com/YnK0L7ta This file really evolved over time and I suspect that they added vectors to handle missing instructions faster (and was why their FPSP was pretty fast at FPU emulation). After a while it didn't bear much resemblance any more but the spirit was still there -- add into software what was missing between the ColdFire and 68K. Much later version: https://pastebin.com/bVG2AhAX They decided to move to the V2 and redo all that logic in the core. I doubt the V2 still uses any of this. Between V1 and V2, the core ballooned from around 6000 LEs to nearly 29000 by the end of its development arc. Aside from minor bug fixes, I would say the V2 development is dead now. The third interesting bit, Exhibit "C" comes from one of Apollo's most interesting features -- instruction fusing. The ColdFire V4 core claimed, erroneously, that it was superscalar by implementing a form of instruction fusing that could combine two instructions into one. This is the EXACT SAME feature present in the Apollo core and was present before true superscalar was born on the Vampire V2 and the never-publically-released ColdFire v5. This comes from the V4, of course, but I'll point out that the V1 came long after the V2, V3 and V4 -- it was a stripped down V4, not a V1 predecessor. It's possible that much of the logic was easy to re-add. https://i.imgur.com/bRxTrH2.png Lastly is the issue of the FPU. I need to preface this with the fact that in OpenCores.org there are no complete FPU implementations -- FPUs are hard and very complex, especially when you want to do things like deep pipelines for single-cycle division. When FPU was first demoed on the Vampire is was using a "lightly accelerated" version of FPSP and pulled off an impressive MFLOPS score (for emulation). This was DURING the development of GOLD2.7 way back in September 2017. http://www.apollo-core.com/knowledge...=8608&z=9BgCFO Then, faced with criticism, as if by magic, almost literally overnight, we suddenly had an almost perfect working (ignoring precision limitations) FPU core pushing almost 90 MFLOPS. A fully pipelined, 68040 compatible core that is cycle-to-cycle faster than even the 68060 was materialized overnight after Gunnar has fought tirelessly to not waste his time with FPU at all. Now, all of this is still, sadly, a little circumstantial and would hardly hold up in a court of law. And to play my own devil's advocate, if they had the sources, then why did we get that bastard son of AltiVec and MMX instead of the ColdFire EMAC. Why didn't we also get the ColdFire V4's impressive MMU that's mostly compatible with 68K while enhancing it greatly. For development, I'd much rather have an MMU than a "full" FPU, even if there's "no software for it." I don't know -- those could be ideological decisions. Maybe Gunnar didn't want to play all his cards on the table like that and make it too obvious. Anyway, that's my two-cents. I honestly have no problem with the V2 and I have no problem with it being allegedly based on the ColdFire core. It's a great board and the best bang-for-the-buck accelerator on the Amiga. That doesn't mean I approve of that group's ethics or heavy-handed behavior and that's why I left the team. IP Extreme has long since removed the ColdFire V1 core from it's website, although you can still find it through the Wayback Machine. I'm not sure why they reversed this decision -- I doubt a few thousand boards sold out of Eastern Europe had any affect. But it makes you wonder. I'd leave it up to everyone to make up their own minds on this one and maybe if we actually get to see the source, we can say for certain whether or not the Apollo Team owes Silvaco (nee IP Extreme) several thousand in back-royalties. |
04 June 2020, 22:17 | #2 |
Registered User
Join Date: Jan 2016
Location: North Bay, Ontario, Canada
Posts: 67
|
Awesome info .
|
04 June 2020, 22:50 | #3 |
Banned
Join Date: Aug 2005
Location: London / Sydney
Age: 47
Posts: 20,420
|
Heya nonarkitten,
Thank you for your post. Definitely implies even more than we thought; and that the Apollo team are most definitely a dodgy bunch at the very least. ...but I don't really want to kick this off again. However, I'll leave the thread open for now as people need to know the truth, but it has to remain civil. Last edited by DamienD; 04 June 2020 at 23:15. |
04 June 2020, 22:53 | #4 |
Total Chaos forever!
Join Date: Aug 2007
Location: Waterville, MN, USA
Age: 49
Posts: 2,186
|
@nonarkitten
The 68050 wasn't superscalar. The 68070 was though. @thread Gunnar could have licensed IP from his former employer at IBM. In fact there used to be a bar graph on the Apollo team website indicating performance between softcores running on a high-end Stratix FPGA devboard at IBM. Softcores that all had FPU code. IBM chose the PPC 440 softcore because their OS required an MMU. The 68050 was faster because it didn't have one. Let's not jump to conclusions. |
04 June 2020, 23:34 | #5 |
Registered User
Join Date: Jul 2018
Location: Gilbert
Posts: 16
|
Take it up with the Apollo team and keep the speculation out of the public until hard evidence or better, proof if something illegal, can be provided. All you are doing is smearing people. What is your agenda?
This is why we cannot have nice things. |
04 June 2020, 23:41 | #6 |
Banned
Join Date: Aug 2005
Location: London / Sydney
Age: 47
Posts: 20,420
|
|
04 June 2020, 23:48 | #7 |
Registered User
Join Date: Jul 2018
Location: Gilbert
Posts: 16
|
This line "Now, all of this is still, sadly, a little circumstantial", and this word "sadly", tells me all I need to know about the purpose of this thread. The hypocrisy of this site and so many of the people here is just baffling.
Seriously. This is all a hobby for a few people left in the world. Reverse engineering is not illegal, and there is no proof that anything other than this has occurred. What is your agenda DamienD? Why the hatred for all things Vampire? What is the threat? |
04 June 2020, 23:51 | #8 | ||
Banned
Join Date: Aug 2005
Location: London / Sydney
Age: 47
Posts: 20,420
|
Quote:
More than you, that's for sure. Quote:
Anyway, this is not about me... back to the topic please. Last edited by DamienD; 05 June 2020 at 10:00. Reason: Corrected my post to reflect that nonarkitten is a lady; sorry about that nonarkitten. |
||
04 June 2020, 23:52 | #9 |
Total Chaos forever!
Join Date: Aug 2007
Location: Waterville, MN, USA
Age: 49
Posts: 2,186
|
Reverse engineering might not have even been necessary. I was told in no uncertain terms by an IBM engineer that Coldfire was just an attempt to slap 680x0 instruction sets onto a second generation PowerPC. Gunnar used to work on PPC CPUs for a living.
|
04 June 2020, 23:59 | #10 | |
Registered User
Join Date: Jul 2018
Location: Gilbert
Posts: 16
|
Quote:
I think "possibly" pretty much says all I need to know. What is the threat? Why the hate for all things Vampire? Last edited by DesertBlizzard; 05 June 2020 at 00:07. |
|
05 June 2020, 00:05 | #11 |
Total Chaos forever!
Join Date: Aug 2007
Location: Waterville, MN, USA
Age: 49
Posts: 2,186
|
Nonarkitten has different reasons than I do for leaving the Apollo team. Her evidence is not very compelling in that her case is based on the premise that Gunnar used a Coldfire superscalar design as the basis for his 68050 core. The fact that the 68050 was NOT superscalar pokes a massive hole in that hypothesis.
|
05 June 2020, 00:05 | #12 |
Banana
Join Date: Jul 2016
Location: Darmstadt
Posts: 1,213
|
If you don't like it, don't buy it. Endless vitriol isn't helping anyone.
|
05 June 2020, 00:09 | #13 |
Global Moderator
Join Date: May 2001
Location: Derby, UK
Age: 46
Posts: 2,287
|
I'm nipping this in the bud before it gets going....
Move along one and all. |
05 June 2020, 00:10 | #14 |
Banned
Join Date: Aug 2005
Location: London / Sydney
Age: 47
Posts: 20,420
|
|
05 June 2020, 00:16 | #15 | |
Global Moderator
Join Date: May 2001
Location: Derby, UK
Age: 46
Posts: 2,287
|
Quote:
I can sense the incoming poopstorm, and frankly if it needs to happen, then it better be based on infallable evidence, as I don't have the time, knowledge or motivation to police such a topic. As it stands it's a nice work of fiction, but it's already triggered some so it's done with. Any problems with this, then please shout at me in my left ear, I'll make sure to take my hearing aid out. |
|
05 June 2020, 11:03 | #16 | |
Banned
Join Date: Aug 2005
Location: London / Sydney
Age: 47
Posts: 20,420
|
I received a PM from Thomas Richter.
For the sake of clarity; I'm posting in this thread: Quote:
|
|
05 June 2020, 12:07 | #17 | |
Banned
Join Date: Aug 2005
Location: London / Sydney
Age: 47
Posts: 20,420
|
Again, for the sake of transparency... here's a PM from Samurai_Crow to Ian:
Quote:
|
|
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Legality of CoffinOS and Vampire Team | Hewitson | Amiga scene | 268 | 05 October 2020 12:45 |
Vampire V4 SA - UK | MartinW | MarketPlace | 11 | 17 May 2020 15:42 |
vampire V2 - what should i be using it for...? | RobA1200 | Amiga scene | 238 | 17 July 2017 21:36 |
Vampire II - Who is first? | JackLeather | support.Hardware | 2 | 26 January 2016 13:56 |
EAB and legality... | TCD | project.EAB | 41 | 08 September 2008 10:45 |
|
|