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 

Krzysztof Wojteczek

Posts 29
07 Sep 2017 09:21


Gunnar von Boehn wrote:

 
  Why do you do this at all?
It was clearly stated that this version is not finished and not speed tuned at all.
You know yourself that FEMU will be magnitudes faster when fully speed tuned. So what intention do you have?


Because I can:). I will do test Gold 2.7 and a newer version of femu with joy, if I will put my hands on them. I have no bad intentions, just testing stuff available at the moment.

Gunnar von Boehn wrote:

  Second, your list does compare two different compiles of the same program. 

Nope I have done tests on LW3.5.fp version using Gold2+Femu and Blizzard 060 with oxypatcher.


Thierry Atheist

Posts 644
07 Sep 2017 09:36


I've lost track now....

I can't find the 68080 cache size. 68060 had cache sizes of 8K data and 8K instruction. Is the Vampire 2 using 32K/32K?

If so, for those wanting a hard FPU in the Vampire 2, how much of a performance hit would be taken (IF this is possible) if the cache sizes were reduced to 16K instruction/32K data OR 16K instruction and data each????


Vojin Vidanovic

Posts 770
07 Sep 2017 09:39


Krzysztof Wojteczek wrote:

  My score with vampire v2 gold 2.5 and nofpu LightWave 3.5  - 7h 12min 29 sec (25 949 sec)

So FPU-less we are ~7 times slower then 75Mhz 060 with full blown FPU and Oxypatcher. Waiting to see how that will decrease with 2.5+improved FEMu and then with 3+hardFPU on v4.

Rare beast that you have.


Krzysztof Wojteczek

Posts 29
07 Sep 2017 09:51


Vojin Vidanovic wrote:

Krzysztof Wojteczek wrote:

  My score with vampire v2 gold 2.5 and nofpu LightWave 3.5  - 7h 12min 29 sec (25 949 sec)
 

 
  So FPU-less we are ~7 times slower then 75Mhz 060 with full blown FPU and Oxypatcher. Waiting to see how that will decrease with 2.5+improved FEMu and then with 3+hardFPU on v4.
 
  Rare beast that you have.

Thank you. Im waiting also. I have put Vampire score with nofpu version of LightWave just for information. But after reading Gunnars post I will make test on 060 with the same math libraries as in VampireOs and nofpu version of LW.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
07 Sep 2017 10:00


Krzysztof Wojteczek wrote:

Gunnar von Boehn wrote:

 
Why do you do this at all?
It was clearly stated that this version is not finished and not speed tuned at all.
 

Because I can:).

Lets understand how what you did, might look from a different perspective.

People moan and complain about wanting a free FPU.
Someone (Jari) picks up the work and programs one for the benefit of all AMIGA users.

He gives our a pre-alpha - version clearly stating that its not done yet - and that its not performance tuned yet. And that people should please keep this in mind.

And first thing you do (because you can) is doing benchmarks on the known pre-alpha slow version to show in charts that its not fast an you run around and post these results.

What do you want to proof?

That its a failure to code for AMIGA people?

That giving out alpha version versions to users only causes trouble  and that programmers should better not do this?



Mr Niding

Posts 459
07 Sep 2017 10:10


@Gunnar

I think he is merely curious what he personally can achive with the publicly available cores and supporting files.

Nothing inherently wrong with that, but it would ofcourse be helpful if people add a note like "this test is done with older/outdated core, but atm newer higher performing cores are limited to betatesters".


Krzysztof Wojteczek

Posts 29
07 Sep 2017 10:19


Gunnar von Boehn wrote:

  What do you want to proof?
That its a failure to code for AMIGA people?
That giving out alpha version versions to users only causes trouble  and that programmers should better not do this?

Im really happy that Jari wrote FEMU, I like my Vampire Card too. I dont want to proof anything im just measuring current speed of this solution. Gunnar what torouble do I cause:)? You are publishing loads of screenshots from Sysinfo proveing that femu is getting faster.


Krzysztof Wojteczek

Posts 29
07 Sep 2017 10:39


@Gunnar
 
  I have question. Have you got any idea why LightWave shows wrong rendering time on Gold 2.5? In my case it was less then half time it actually took to render the scene.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
07 Sep 2017 14:18


Krzysztof Wojteczek wrote:

  @Gunnar
   
I have question. Have you got any idea why LightWave shows wrong rendering time on Gold 2.5? In my case it was less then half time it actually took to render the scene.

 
I would never try to run this now.
 

1) FEMU is not designed for this usecase.
2) The FEMU version that you use is an untuned test/debug release - which is known to be 10 times slower than it can be.
3) The demo FEMU version that you use does depend of lots of external conditions. Is VBR moved in fast? What math libs are installed. All these conditions influence the performance again.
With VBR in chip - your FEMU test-release will run halve speed.
 
 
As you see your measurements are today, flawed by design.
 
I would in your position wait for the "final" release of FEMU.


Mallagan Bellator

Posts 393
10 Sep 2017 19:40


Gunnar von Boehn wrote:

Mr Niding wrote:

  I guess what Vojin is saying; is that current high level compilers doesnt support the new features of the 68080 core, and as such wont be able to take advantage of the performance potential.
 

 
  Do you assume you need to use new 68080 instructions?
  This is NOT the case.
 
  The FPU of the 68080 is 100% compatible to the previous 68K FPUs.
  You can continue to write using the normal FPU instructions.
 
  The point of the FPU coding will be to "hand control" the instruction scheduling.
  In other words the order in which the instructions are placed in the program. And you can only control this in ASM.
  Therefore this can only be sensible be coded in ASM. (as of today)
 
 
 

Would you be able to program for, let's say, ammx, this way Gunnar?


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
10 Sep 2017 21:57


Mallagan Bellator wrote:

  Would you be able to program for, let's say, ammx, this way Gunnar?

Sorry I do not understand the question.


Mallagan Bellator

Posts 393
11 Sep 2017 19:58


Gunnar von Boehn wrote:

Mallagan Bellator wrote:

  Would you be able to program for, let's say, ammx, this way Gunnar?
 

  Sorry I do not understand the question.

Sorry, I should be more specific.
So, any C/Cpp/C# compiler / env that I know of have the instructions of 040/060 in them, but can features like ammx and other stuff inside the core be programmed for using a compiler that is available somewhere? Like from direct coding, or headers (or what the name was... it has been a while since I looked into C programming) or Cpp classes? Or is that yet so far only accessable through assembler?


Mallagan Bellator

Posts 393
11 Sep 2017 20:13


Gunnar von Boehn wrote:

 
  As you see your measurements are today, flawed by design.
   
  I would in your position wait for the "final" release of FEMU.

My interpretation is that he was probably really eager to try this out with what he had, and patience was not his strongest side
Some people can wait a long time for stuff, while some try to get the most out of every hour of every day, and want said stuff asap.

I can understand both sides, but sometimes, misunderstandings may be had. Especially since there is such a great divide in the Amiga comunity, which is really sad. I really wish people could just get along... (sigh...) comunication would have been so much easier, and motivation for those who WANT to do something for the greater lot, would be higher.

I know it's no fun trying to do something awesome (FEMU / 68080 core / Vampire) for some who appear ungreatful.

Let me clarify one thing. I think you're all doing great jobs, any of you guys who develop those things!

I've seen the videos on YouTube, and I can't wait to plug a vampire into my A1200

:)


Remi Jakobsen

Posts 9
12 Sep 2017 21:42


Hello Gunnar. Do you have any FEMU version I can "play with" on my A600 vampire, or Gold 2.7 Beta maybe? I've got the USB flasher of course, and loves to test the software I used professionally in the 90's wile using Amiga in the multimedia/infochannel industry.

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