05 September 2016, 20:16 | #1 |
Posts: n/a
|
GOLD core released for Vampire 600 V2
The GOLD core has been released for the vampire 600 V2, and it brings new cool features !
Full changelog : * Added fusing for MOVE/EXT * added MPU for finding illegal memory access * added TRACE support * fixed in CHK2, CMP2, MOVEM and CAS * fixed some special LOOP condition * fixed TRAPcc * optimized MOVEQ/MOVE.l # and depending instructions Download it : http://www.apollo-accelerators.com/wiki/#latest_cores Apollo team would be happy to get feedbacks from users using this new GOLD core. Be free to join #Apollo-Team IRC channel on Freenode to discuss it. |
06 September 2016, 03:52 | #2 |
Registered User
Join Date: Sep 2007
Location: Melbourne/Australia
Posts: 4,441
|
Very impressive but what happened to the FPU support, wasn't that meant to part of the gold core?
|
06 September 2016, 07:19 | #3 | |
Registered User
Join Date: Feb 2015
Location: Sweden
Age: 50
Posts: 2,988
|
Quote:
It seems implementing a bug free, properly functioning FPU is harder than they initially thought. Skickat från min HTC One via Tapatalk |
|
06 September 2016, 07:35 | #4 |
Registered User
Join Date: Jul 2014
Location: Finland
Posts: 1,187
|
is any of that 'AMMX' etc stuff documented by now?
|
06 September 2016, 09:20 | #5 |
Unregistered User
Join Date: Sep 2012
Location: Copenhagen / DK
Age: 44
Posts: 4,190
|
Cool, will try it out. Is there any noticable performance improvements? I think the fusing and optimizations of MOVE could make a difference?
And how about compatibility with e.g. FUSION/Shapeshifter? Does these fixes relate to those? |
06 September 2016, 14:41 | #6 |
Total Chaos forever!
Join Date: Aug 2007
Location: Waterville, MN, USA
Age: 49
Posts: 2,200
|
It's pretty much identical to the Intel MMX instruction set but at least it's finalized so it can be documented.
http://www.apollo-accelerators.com/w...s_instructions is where the documentation will appear when it's done. Last edited by Samurai_Crow; 06 September 2016 at 16:30. Reason: link added |
06 September 2016, 16:53 | #7 |
Registered User
Join Date: Jul 2014
Location: Finland
Posts: 1,187
|
No opcode encoding details, no real functional descriptions, no code examples, pretty much nothing.
And similar to Intel MMX.... yes the mnemonics are very similar but the register handling is undefined (what registers do this map to?) and these seem to be 3 operand. If you want someone to use your fancy stuff you need to provide them a means to. this isn't how you do it... |
06 September 2016, 18:47 | #8 |
Registered User
Join Date: Sep 2011
Location: Paris/France
Posts: 278
|
I agree but see here :-)
http://www.apollo-core.com/index.htm?page=instructions especially the Parallel MMX like instructions PPERM others are not renseigned yet |
06 September 2016, 19:43 | #9 |
Registered User
Join Date: Jul 2014
Location: Finland
Posts: 1,187
|
still very little...
Its also weird that documentation is a 'after thought' if you design something like a CPU you spec first and then implement. So how can you not have a document? |
06 September 2016, 19:54 | #10 |
Natteravn
Join Date: Nov 2009
Location: Herford / Germany
Posts: 2,552
|
There are still no opcode encodings.
I would be willing to support the Apollo core with the vasm assembler some day. But then I need better documentation. And the Apollo ISA must reach a defined state, where no more instructions are added or removed every few weeks. |
06 September 2016, 21:23 | #11 | |
Registered User
Join Date: Jun 2015
Location: Germany
Posts: 1,927
|
Quote:
A lot of work went into the Gold release. As you probably know the last 10% are 90% of the entire work. All 68k integer instructions are now implemented and thoroughly tested, even such instructions that do not exist in Amiga software but only in MacOS. |
|
06 September 2016, 23:29 | #12 | |
Registered User
Join Date: Feb 2015
Location: Sweden
Age: 50
Posts: 2,988
|
Quote:
Frankly, they've done amazing things not many of us expected to ever see in Amigaland again.. ..but I'm not so sure the FPU implementation is going exactly as expected. If it is, it's hard to explain Gunnar's post from May 24th in this thread: http://www.apollo-core.com/knowledge.php?b=1¬e=1108 Spoiler: "#For the next Release we are currently working on the integration of 1) fully pipelined = fast and compatible 80Bit FPU 2) 128 bit SSE compatible SIMD instructions" And now we know the AMMX is already implemented while the FPU is not... I'm sure its a tough nut to crack even for the magicians in the Apollo team, but even if its still far off, IMO I'd love to hear more about it. Can we help beta test it etc? |
|
08 September 2016, 00:32 | #13 |
Registered User
Join Date: Dec 2013
Location: Lake Havasu City, AZ
Posts: 741
|
I am not aware of any issues with FUSION. I can run everything.
|
08 September 2016, 06:09 | #14 |
Unregistered User
Join Date: Sep 2012
Location: Copenhagen / DK
Age: 44
Posts: 4,190
|
Ok, I just remember that there were some issues previously which were addressed, but I wasn't sure if it was fixed 100% back then. But good to hear. I guess you will still have trouble running Mac software which expects an FPU?
|
08 September 2016, 06:58 | #15 |
Registered User
Join Date: Dec 2013
Location: Lake Havasu City, AZ
Posts: 741
|
The core had some of the FPU instruction hooks in it that were causing FPU errors (the Mac thought there was a real FPU available). Gunnar removed the hooks and that fixed the issues that did exist. Programs that "require" a FPU can be ran using the SoftwareFPU extension.
When Gunnar puts in the final FPU core, the emulation is going to be insane. It's already testing faster than 8100/80 PowerMac in native PowerPC mode! |
08 September 2016, 07:38 | #16 | |
Registered User
Join Date: Jun 2015
Location: Germany
Posts: 1,927
|
Quote:
A lot of work was done on the FPU years ago, all 040-FPU instructions are implemented and also tested. What is missing is testing program flow and flags set by the instructions. This will also be a lot of work but I'm pretty sure we will see it progress quite quickly. |
|
08 September 2016, 08:29 | #17 | |
Registered User
Join Date: Feb 2015
Location: Sweden
Age: 50
Posts: 2,988
|
Quote:
Skickat från min HTC One via Tapatalk |
|
08 September 2016, 12:35 | #18 |
Registered User
Join Date: Jan 2015
Location: Melbourne, Australia
Posts: 548
|
My question is, has anyone with a Vampire II installed it yet? I'm curious to see if there are improvements, and if so, what sort of a difference the end user would notice. I'll probably be ordering one of these fantastic critters soon, just have some other more important expenses to take care of first.
|
08 September 2016, 12:50 | #19 | |
Registered User
Join Date: Feb 2015
Location: Sweden
Age: 50
Posts: 2,988
|
Quote:
(Meaning, my Vamp works fine with v6, but updating with v7, 8, 9 will brick it. Simply naming v10 " Gold" doesn't necessarily mean it's any different in that respect. But I have a USB Blaster to fix bricked Vamps, so I'll give Gold a go soon I guess. Skickat från min HTC One via Tapatalk |
|
08 September 2016, 15:43 | #20 | |
Registered User
Join Date: Feb 2016
Location: London / UK
Posts: 166
|
Quote:
|
|
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
SILVER9 core released for Vampire 600 V2 | TuKo | News | 1 | 04 August 2016 13:13 |
SILVER8 core released for Vampire 600 V2 | TuKo | News | 3 | 02 August 2016 04:13 |
SILVER7 core released for Vampire 600 V2 | TuKo | News | 0 | 10 July 2016 23:12 |
SILVER5 core released for Vampire 600 V2 | TuKo | News | 12 | 21 May 2016 00:25 |
SILVER3 core released for Vampire 600 V2 | TuKo | News | 4 | 03 April 2016 14:01 |
|
|