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
Performance and Benchmark Results!

Apollo Hardfpu Resultspage  1 2 3 4 5 6 7 8 

Mallagan Bellator

Posts 393
09 Oct 2017 17:17


Kolbjørn Barmen wrote:

Gunnar von Boehn wrote:

  The MMU is hidden and not accessible for you.
  Technically this is for today the only reasonable solution.
 

 
  Yes, exactly my understanding.
 
 
If we would "expose" the MMU then your old software would wrong program it - and you would crash or loose data.

 
  Does this also imply that the MMU is already used by the firmware that comes with Vampire cards?

"Is it being used", the troll under the bridge says...
If you have half a brain, you will understand that they wouldn't write an MMU just to not use it. Are you bored or something? Is this your idea of fun? Reading your posts are starting to get REALLY annoying...


Nixus Minimax

Posts 416
09 Oct 2017 18:14


Kolbjørn Barmen wrote:
Does this also imply that the MMU is already used by the firmware that comes with Vampire cards?

How do you think the Vampire maps Amiga chipmem to Vampire fastmem, hardware register addresses to either external (Amiga custom chips) or internal (FPGA functional units), decides on which addresses it may cache and which not, maps the ROM to fastmem? Open your eyes!



Mallagan Bellator

Posts 393
09 Oct 2017 18:43


The troll is hungry. Don't feed the troll. If we're lucky it will go somewhere else to feed


Thierry Atheist

Posts 644
09 Oct 2017 22:14


Kolbjørn Barmen wrote:
Why do you say so? Debian is not modern enough for you?

Who the heck would want to use an operating system that will use up 38 megabytes of your 128 megabytes of RAM?



Samuel Devulder

Posts 248
09 Oct 2017 22:37


Gunnar von Boehn wrote:

  As we protect the others from this - the MMU is today kept private.

Fair enough. Let's hope MMU spec will be eventually be made available for developpers to work with and not restricted to happy fews.


Mr Niding

Posts 459
09 Oct 2017 22:39


Thierry Atheist wrote:

Kolbjørn Barmen wrote:
Why do you say so? Debian is not modern enough for you?

  Who the heck would want to use an operating system that will use up 38 megabytes of your 128 megabytes of RAM?
 

I dont see anything inherenty bad about running Linux on any hardware, even if some other OS's are better suited for it.

The question at hand is wether or not the development should be changed to accomodate some userprefrences. We already know the answer to that, which is why I dont understand the need to harp on it.

But again, I dont see anything negative with Kollas desire to run any software he can get patched for example.


Kolbjørn Barmen
(Needs Verification)
Posts 219/ 2
09 Oct 2017 22:45


Nixus Minimax wrote:

How do you think the Vampire maps Amiga chipmem to Vampire fastmem, hardware register addresses to either external (Amiga custom chips) or internal (FPGA functional units), decides on which addresses it may cache and which not, maps the ROM to fastmem? Open your eyes!

Sorry for confusing you, I mean the Gold core 2 (and perhaps 2.5), and certainly not 2.7 or whatever will happen in the future.


Kolbjørn Barmen
(Needs Verification)
Posts 219/ 2
09 Oct 2017 23:10


Samuel Devulder wrote:

  Fair enough. Let's hope MMU spec will be eventually be made available for developpers to work with and not restricted to happy fews.

Thank you for sharing my view, and welcome to the troll club.


Nixus Minimax

Posts 416
10 Oct 2017 06:23


Kolbjørn Barmen wrote:

 
Nixus Minimax wrote:

  How do you think the Vampire maps Amiga chipmem to Vampire fastmem, hardware register addresses to either external (Amiga custom chips) or internal (FPGA functional units), decides on which addresses it may cache and which not, maps the ROM to fastmem? Open your eyes!
 

 
  Sorry for confusing you, I mean the Gold core 2 (and perhaps 2.5), and certainly not 2.7 or whatever will happen in the future.
 

 
  Most of what I listed above is a prerequisite to even make the Vampire boot any core version. You are the one that seems confused.
 
 


Nixus Minimax

Posts 416
10 Oct 2017 06:27


Kolbjørn Barmen wrote:

Samuel Devulder wrote:

  Fair enough. Let's hope MMU spec will be eventually be made available for developpers to work with and not restricted to happy fews.
 

 
  Thank you for sharing my view, and welcome to the troll club.

He's not in the troll club because he used the word "eventually". He does understand that time plays an important factor in all this.



Gunnar von Boehn
(Apollo Team Member)
Posts 6207
10 Oct 2017 11:48


Update




Samuel Devulder

Posts 248
10 Oct 2017 16:08


Do we have an idea on how these 73MFlops impact floating-points crunchers like ray-tracers programs?
   
I'm thinking here about the well known POVray ( EXTERNAL LINK ) program. For instance in 1992[*], it took 2hours for POVray1.0 on a SGI Iris 4D (presumably 4D/35, 6MFlops). On the vampire, the same scene might be rendered at least 13x faster, so overall in less than 8mins I guess (still a good time lapse).
   
Here is a page on how to use POVray for benchmarking CPUs on a standard scene: EXTERNAL LINK ( and a site collecting the results back from 1994: EXTERNAL LINK Better compare with older entries: EXTERNAL LINK )
 
___
[*] I wish I could get newer references, mines are really too old. Ok, found some: Bonzai EXTERNAL LINK , Pirate boat underwater EXTERNAL LINK     


Peter Heginbotham

Posts 214
10 Oct 2017 17:53


Gunnar von Boehn wrote:

Update
 
 
 
 

So if i read this correctly, A Vampire 2 with Gold 2.7 + Softfpu is out performing a 060 by a factor of 170%


Stefano Briccolani

Posts 586
10 Oct 2017 22:48


Phenomenal!!


Sean Sk

Posts 488
10 Oct 2017 22:55


Peter Heginbotham wrote:

So if i read this correctly, A Vampire 2 with Gold 2.7 + Softfpu is out performing a 060 by a factor of 170%

This is the hard FPU for Vampire V4 not soft FPU for V2.


Chris Dennett

Posts 67
11 Oct 2017 01:38


For doing float stuff (i.e., the game/utility does a lot of complex 3D graphics), is the FPU more efficient or more efficient to use the integer instructions? It seems hard to compare the MFlops and Mips and people may think they are directly comparable. We already know singular benchmark results are not the best way to directly probe how fast a processor will be under particular workloads..


Claudio Guglielmotti
(Apollo Team Member)
Posts 185
11 Oct 2017 07:22


the picture was taken by me, and it is my Vampire500.


Gregthe Canuck

Posts 274
11 Oct 2017 07:38


claudio guglielmotti wrote:

  the picture was taken by me, and it is my Vampire500.
 

 
Well that is definitely interesting. Some sort of hybrid FPU support? I believe that was mentioned elsewhere? :)


Mr Niding

Posts 459
11 Oct 2017 08:45


Yep Greg, on page 2 of this thread;

"While we have in the team builds with partial HARD-FPU for VAMP2 - please understand we can _not_ promise this for a customer release for the VAMP-2 as of today. "


Markus B

Posts 209
11 Oct 2017 09:51


OK, so it means even better results can be expected from the V4 FPU?

posts 158page  1 2 3 4 5 6 7 8