Overview Features Coding ApolloOS Performance Forum Downloads Products Order Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
Please visit our Apollo-Discord Server for support.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Information about the Apollo CPU and FPU.

Status of the FPUpage  1 2 3 4 5 6 7 8 9 10 11 12 13 

Samuel Devulder

Posts 248
30 Aug 2017 07:46


I do not understand the matter about the MMU. The 68080 do have a MMU. It isn't compatible to the other 68k MMUs because the legacy MMUs don't support modern architectures, but it definitely already have a memory management unit.


Gregthe Canuck

Posts 274
30 Aug 2017 08:55


It breaks some developer tools as far as I can tell. But that is no different than the 68030 to 68040 jump where the MMU was changed. I know all about that since Amiga UNIX broke when I tried using an A3640 accelerator. It only supported 68030 and the MMU logic was never upgraded.
 
So whichever developer tools are important will need updating to support the new 68080 MMU... once the functionality is exposed.
 


Rollef 2000

Posts 29
30 Aug 2017 12:17


What is the difference between testing a fully working HardFPU and testing a SoftFPU?


Samuel Crow

Posts 424
30 Aug 2017 14:43


rollef 2000 wrote:

What is the difference between testing a fully working HardFPU and testing a SoftFPU?

The HardFPU requires a bigger FPGA than what current Vampire models supply so the SoftFPU will succeed while the HardFPU is impossible to use.


Vojin Vidanovic

Posts 770
30 Aug 2017 14:48


Samuel Crow wrote:

  The HardFPU requires a bigger FPGA than what current Vampire models supply so the SoftFPU will succeed while the HardFPU is impossible to use.

That is, for V1 and V2.

Note that Gunnar just promised hardFPU for v4.


Samuel Crow

Posts 424
30 Aug 2017 14:59


Vojin Vidanovic wrote:

  That is, for V1 and V2.
 
  Note that Gunnar just promised hardFPU for v4.

Since V4 isn't quite released yet, I don't consider it to be a current model.  It is a future model.


Rollef 2000

Posts 29
30 Aug 2017 15:39


OK, i understand. The hardFPU does not fit to the current Vampire models.

Gunnar von Boehn wrote:
 
  While the APOLLO FPU was successfully used in some customer project, for usage in 68K AMIGA OS some more testing was found required.
  The Apollo-team did ask for AMIGA coders community to help testing 12month ago - in this time no coders applied for this.

Gunnar von Boehn wrote:
 
  We look for someone interested in helping to write a new high end FPU application - which shows that FPU performance can be pushed to new limits.

But how fit these statements from Gunnar to your answer.
Was there are a time the hardFPU fit to current Vampire?
How should the coders write and test programms?



David Wright

Posts 373
30 Aug 2017 16:08


Aren't they manufacturing them now?


Vojin Vidanovic

Posts 770
30 Aug 2017 17:40


David Wright wrote:

  Aren't they manufacturing them now?
 

 
  And should be testing them. But its stil considered as future ...
 
  Well, v4 as supposed to be out Q4 17 and Q1 18 depending on models!

Note that some amiga.org users try to make joke out of it.
EXTERNAL LINK


Rod March

Posts 119
31 Aug 2017 00:53


Ok, so there's a zero % chance my V2 will ever get an FPU. :(
 
At least I know now...

Sorry to sound negative, just want to have realistic expectations.


Vojin Vidanovic

Posts 770
31 Aug 2017 07:03


Rod March wrote:

Ok, so there's a zero % chance my V2 will ever get an FPU. :(
  At least I know now...
  Sorry to sound negative, just want to have realistic expectations.

FPGA limits, one day we will have something that wont fit V4.

There is FEMU as soft version.



Thierry Atheist

Posts 644
31 Aug 2017 08:38


Rod March wrote:
Ok, so there's a zero % chance my V2 will ever get an FPU. :(

At least I know now...

Sorry to sound negative, just want to have realistic expectations.


Hi Rod,

I really REALLY hope that this doesn't rub too many people the wrong way, so that they want to abandon what's been achieved thus far....

This is precisely why I was pushing so hard for a jump to the Arria 10, rather than having too many stages with different levels of performance.

It would have been Vampire 1, V2, Arria 10 ----> ASIC.


Michal Warzecha

Posts 209
31 Aug 2017 11:23


Rod March wrote:

Ok, so there's a zero % chance my V2 will ever get an FPU. :(
 
  At least I know now...
 
  Sorry to sound negative, just want to have realistic expectations.

I think You souldn't worry about this too much. Too many ppl buy V2 vampires and it's still under development. There are smart people in Vampire team and I think when FPU will be fully working under V4, they will port it to V2, mayby without all extra features.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
31 Aug 2017 13:14


Thierry Atheist wrote:

This is precisely why I was pushing so hard for a jump to the Arria 10,

We try to do the best for all users.

The ARRIA 10 chip alone cost over $300.
You can not produce and sell a product for 300 when one of the chips alone does cost more.

The team aims to create a very good but also affordable products.

The situation with the V2 was explained before.
The full FPU is there but its not easy fit in the V2 FPGA atm.
Saying this you have to mind that the V2 offer the best and biggest FPGA of all AMIGA FPGA products. The FPGA of the V2 is far bigger than what MIST/ARCADE/etc offer.

With the V4 the team got even a step further.
The FPGA on the V4 is about twice as big as the V2 FPGA.
So fitting the FPU in the V4 is no problem.

There are also FPU solution for the V2.
1st solution is FEMU. FEMU allows to run programs making light FPU use.
2nd option is size tweaking the FPU.
This is complex work but can be done.
But the key people needed for this are currently busy testing the V1200. After the V4 and V1200 are out on the market we will have time to look into size tuning.

So if you buy the VAMPIRE - then buy it for the features it has now.



Rod March

Posts 119
31 Aug 2017 13:33


Thanks Gunnar for such a thorough reply.

You guys are doing great work and I can appreciate that you have to balance the competing limitations of time/cost/performance.

I think I'm in the unfortunate position wanting mostly to do FPU intensive tasks and had just assumed it was part of the package when I ordered. This is no one's fault but my own however.

It's nice to know there are still FPU possibilities for the V2. I guess we'll just have to be patient and see what's possible.

Best of luck to all!



Thierry Atheist

Posts 644
31 Aug 2017 14:47


Thank you for the reply Gunnar.

I certainly hope that everyone else is as happy with that clear explanation as Rod March is.

The Vampire 2 is GREAT, and the Vampire 4 will just be more of EVERYTHING that Vampire has to offer.


Sebastian Blanco

Posts 148
31 Aug 2017 21:15


The apollo MMU and FPU will be implemented in future vampire core releases ?, is enought room left on the fpga for it ?. :D



E Penguin

Posts 46
31 Aug 2017 21:41


Gunnar von Boehn wrote:

size tweaking the FPU.

Purely out of curiosity, could some kind of hybrid hard/soft FPU be made to work? Maybe the most common instructions in hardware, anything esoteric in software? Or, is it that if you implement some floating point calculations it's trivial to do the rest?




Samuel Devulder

Posts 248
31 Aug 2017 22:25


Purely out of curiosity, could some kind of hybrid hard/soft FPU be made to work?

Once you have addition/multiplication/reciprocal in HW, you can probably code the remaining function (trigonometric, exponential, logarithm, square root, etc.) in SW and still get a reasonable spead using some kind of Padé approximant I suppose.
 
Another way to gain size is possibly to drop the HW support for rarely used values such as denormalized and NaNs. These would be handled in SW after a specific TRAP exception occurs.
 
Just my two cents.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
31 Aug 2017 22:37


Daniel Lakey wrote:

could some kind of hybrid hard/soft FPU be made to work?

 
As a matter of fact, the original 68K hard FPUs also did their instructions in software too.(micro code rom)



posts 254page  1 2 3 4 5 6 7 8 9 10 11 12 13