Click Here
home features news forums classifieds faqs links search
6071 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
» Home
» Features
» News
» Forums
» Classifieds
» Links
» Downloads
Extras
» OS4 Zone
» IRC Network
» AmigaWorld Radio
» Newsfeed
» Top Members
» Amiga Dealers
Information
» About Us
» FAQs
» Advertise
» Polls
» Terms of Service
» Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
14 crawler(s) on-line.
 137 guest(s) on-line.
 1 member(s) on-line.


 kolla

You are an anonymous user.
Register Now!
 kolla:  52 secs ago
 OneTimer1:  11 mins ago
 BigD:  53 mins ago
 OlafS25:  55 mins ago
 amigakit:  1 hr 24 mins ago
 NutsAboutAmiga:  1 hr 27 mins ago
 Gunnar:  2 hrs 5 mins ago
 Frank:  2 hrs 11 mins ago
 MickJT:  2 hrs 50 mins ago
 A1200:  3 hrs 14 mins ago

/  Forum Index
   /  Amiga General Chat
      /  New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 Next Page )
PosterThread
cdimauro 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 13:41:18
#781 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@pavlor

Quote:

pavlor wrote:
@Vistaus

Quote:
You guys are talking about (valid) performance issues regarding SPE. IF AOS 4 ever gets multicore support, would the performance of Tabor be better then?


In some applications (Blender) yes, in other (Quake) no.

FPU in e500v2 is weak even without discussed compatibility issue (benchmarks show e500 1.2 GHz is comparable to G3 600 MHz in FPU depending tasks - that is with native SPE compiled code). We don´t know performance penalty for emulated FPU instructions, but don´t expect miracles.

Well, we have at least one case where FPU emulation was 93 (not 96, sorry) times slower. The trap handler didn't used the SPE unit, but it's not possible to get incredible speed-ups, due to the slowness of the trap mechanism itself (see e500v2 manual for this).

And I want to point out one thing here. Why Debian PowerPC was forked to directly (natively) support the SPE unit, if the FPU emulation is "without significant impact on performance" (cit. Hyperion; comment #466)? Wouldn't it have been enough to just provide a proper FPU handler which used the SPE unit instead of forking the entire distro?

Another 2 cents...

 Status: Offline
Profile     Report this post  
pavlor 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:06:33
#782 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9578
From: Unknown

@cdimauro

Isn´t much easier for Linux users to recompile source code for native applications, than at all cost retaining compatibility with older binaries?

Last edited by pavlor on 25-Oct-2015 at 02:06 PM.

 Status: Offline
Profile     Report this post  
iggy 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:08:33
#783 ]
Super Member
Joined: 20-Oct-2010
Posts: 1175
From: Bear, Delaware USA

@umisef

Quote:
you have to go through design and prototyping (i.e. the dreaded NRE), as well as component procurement. That procurement phase can involve very significant lead times. So there is a lot of money committed, both in NRE and in parts


Valid enough, although its been my experience that procurement isn't that difficult (just costly).
But the design phase, even when there is a development board to start with is really nasty.

Um, about the MPC8640/8641D and the MPC8610, I worked with the former (and still have chip samples left over) and I am familiar with the latter.
While its not 100% necessary, for our applications these chips require Southbridge ICs.
I had most of the work done to link an ATI SB600 to the MPC8640/8641D (Freescale's development board uses a ULi SB).

Both of these processors (while they have a nice core) are relatively slow.
I stopped all work on the 8641/8640 when the MorphOS team announced PowerMac G4 support.

Further, I don't think most people are aware of how difficult it is to design a layout for a board using chips that run at the frequencies we are looking at.

12 layers is not unusual, with a lot of attention to signal routing and power planes.
Designing with this stuff is a b*tch.

It was much easier with the 68000 (at least until that exceeded 10MHz).

Yeah, again, the design phase is probably the worst part of development and its 100% sunk cost that is hard to recover.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:09:54
#784 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@pavlor: according to the problems that Spectre660 had 'til now, it's no easy. In fact, and if you take a look at the Debian PowerPCSPE wiki page, the port isn't yet finished, and it's quite incomplete/unstable.

Anyway, why so much effort if the emulation is OK? The problem isn't about keeping binary compatibility here: it's just for performance matters.

 Status: Offline
Profile     Report this post  
Rose 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:13:05
#785 ]
Cult Member
Joined: 5-Nov-2009
Posts: 982
From: Unknown

@cdimauro

Just wait and there will be stories how Hyperion haves secret info and can use it to make much faster emulation that those Linux coders.... (Plot twist, emulation code is written by Freescale employees on Linux)

 Status: Offline
Profile     Report this post  
pavlor 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:17:48
#786 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9578
From: Unknown

@cdimauro

Quote:
In fact, and if you take a look at the Debian PowerPCSPE wiki page, the port isn't yet finished, and it's quite incomplete/unstable.


e500v2 based Freescale SoCs have limited areas of use (IP application, printers etc.). I don´t think these boards are used for Blender rendering or video playback in MPlayer.

 Status: Offline
Profile     Report this post  
pavlor 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:21:26
#787 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9578
From: Unknown

@Rose

Current emulation doesn´t use "FPU" of e500 based SoC at all. That is why Hyperion can claim, they have better solution. As always, we need real application benchmarks of ready product to make conclusion.

 Status: Offline
Profile     Report this post  
Spectre660 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:37:52
#788 ]
Elite Member
Joined: 4-Jun-2005
Posts: 3918
From: Unknown

@pavlor

http://www.freescale.com/video/qoriq-p1022-processor-plus-quick-boot-plus-h.264-software-codec-demo:FTF11ENA_665_VID?uc=true&lang_cd=en

Quote:

pavlor wrote:
@cdimauro

Quote:
In fact, and if you take a look at the Debian PowerPCSPE wiki page, the port isn't yet finished, and it's quite incomplete/unstable.


e500v2 based Freescale SoCs have limited areas of use (IP application, printers etc.). I don´t think these boards are used for Blender rendering or video playback in MPlayer.

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card

 Status: Offline
Profile     Report this post  
iggy 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 14:39:38
#789 ]
Super Member
Joined: 20-Oct-2010
Posts: 1175
From: Bear, Delaware USA

@pavlor

Quote:
Current emulation doesn´t use "FPU" of e500 based SoC at all. That is why Hyperion can claim, they have better solution. As always, we need real application benchmarks of ready product to make conclusion.


@cdimauro

Quote:
Anyway, why so much effort if the emulation is OK? The problem isn't about keeping binary compatibility here: it's just for performance matters.



So we basically have to wait until the 'secret sauce' work around is introduced by Hyperion?
My guess is the the first solution will be entirely integer based like the Linux work arounds.

Then we will be told that the solution needs to be 'optimized'.

 Status: Offline
Profile     Report this post  
pavlor 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 15:01:25
#790 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9578
From: Unknown

@iggy

Quote:
My guess is the the first solution will be entirely integer based like the Linux work arounds.


Good guess.

Quote:
Then we will be told that the solution needs to be 'optimized'.


Exactly this was written by Costel Mincea...

 Status: Offline
Profile     Report this post  
iggy 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 15:06:28
#791 ]
Super Member
Joined: 20-Oct-2010
Posts: 1175
From: Bear, Delaware USA

@pavlor

Quote:
Exactly this was written by Costel Mincea...


Thanks for pointing that out, Pavlor.
I will pay more attention to Costel's postings in the future.

But then, shouldn't have repeated what should be obvious to everyone now, this is going to drag out.

 Status: Offline
Profile     Report this post  
wawa 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 15:58:39
#792 ]
Elite Member
Joined: 21-Jan-2008
Posts: 6259
From: Unknown

@umisef

Quote:

umisef wrote:
@wawa

Quote:
if these boards were old stock by acube, bought by aeon, why would they carry aeon logo. also as i understand trevor said the production run has just been commissioned.


Making PCBs and loading them with components is a relatively quick and easy part of making a motherboard. Prior to that, you have to go through design and prototyping (i.e. the dreaded NRE), as well as component procurement. That procurement phase can involve very significant lead times. So there is a lot of money committed, both in NRE and in parts, long before the first production PCB is being printed, and before the first production board is ever produced.


Sure. It still doesnt explain why the developer batch of boards being in hands of testers since one or two years would carry aeon logo without aeon involvement at the point of production. Not that it is so important whose failure was the design.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 16:12:33
#793 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@Rose

Quote:

Rose wrote:
@cdimauro

Just wait and there will be stories how Hyperion haves secret info and can use it to make much faster emulation that those Linux coders.... (Plot twist, emulation code is written by Freescale employees on Linux)




@pavlor

Quote:

pavlor wrote:
@Rose

Current emulation doesn´t use "FPU" of e500 based SoC at all. That is why Hyperion can claim, they have better solution.

You have to put some resources to do it, and it'll take time. Especially if the plan is to write it in assembly (Petunia is written in such language).
Quote:
As always, we need real application benchmarks of ready product to make conclusion.

The new trap & emulate FPU layer, WHEN it'll be finished, will not change the picture: the FPU emulation performance of the Tabor board will be ridiculous low.

To quote someone else:

"FPU in e500v2 is weak even without discussed compatibility issue (benchmarks show e500 1.2 GHz is comparable to G3 600 MHz in FPU depending tasks"

and that's with NATIVE (not trapped) code. Imagine what can happen with the trap & emulate mechanism, whatever optimization they can bring out...

I know that dreaming is cheap, but real life is a different thing. And will be different. No doubt about it.

Again, IF we're talking of trap & emulate. For native / SPE-based code, as I already stated, performance will be decent and comparable to the already existing low-end PowerPC machines. But I think that you agree here.

Quote:

pavlor wrote:
@iggy

Quote:
My guess is the the first solution will be entirely integer based like the Linux work arounds.


Good guess.

Not so good, if the plan is to recycle the existing Linux's FPU emulation layer, since the code is GPL, and that license is very well know to be viral.

It means that there's an high risk that the OS4 code could be "contaminated" using that layer, and Hyperion will be forced to release her code (for the happiness of Olaf ).

The best thing is to rewrite it from scratch (with a clean-room approach), in order to avoid any possibility. Hyperion have some lawyers which should understand this.

And the average Joe should understand that it takes quite some time in order to get a good enough solution to this problem...

 Status: Offline
Profile     Report this post  
pavlor 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 16:22:50
#794 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9578
From: Unknown

@cdimauro

Quote:
and that's with NATIVE (not trapped) code. Imagine what can happen with the trap & emulate mechanism, whatever optimization they can bring out...


If Hyperion achieves 1/10 of native performance, I would call it success...

Quote:
The best thing is to rewrite it from scratch (with a clean-room approach)


This is exactly what I had in mind and this is - according to published statements of Hyperion management - also plan for AmigaOS.

Quote:
And the average Joe should understand that it takes quite some time in order to get a good enough solution to this problem...


Time... again such precious resource for Amiga developers.

 Status: Offline
Profile     Report this post  
iggy 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 16:40:51
#795 ]
Super Member
Joined: 20-Oct-2010
Posts: 1175
From: Bear, Delaware USA

@cdimauro

Quote:
The best thing is to rewrite it from scratch


I never suggested any other way of doing it, I said 'like' Linux.

And I doubt an spe based translation layer will be that much faster than the integer route.
Its the trapping and conversion, not the execution of the converted code, that will carry the most overhead.
AND, anyway, the spe code generated probably will be slower than a standard fpu.

It isn't really a 'wait and see' scenario, its a 'how bad' scenario.

If recompilation isn't part of the solution, my guess is pretty bad.

@pavlor

Quote:
If Hyperion achieves 1/10 of native performance, I would call it success...


Yes, if that is the hit, and its only on fpu instructions, it would be a success.

Last edited by iggy on 25-Oct-2015 at 04:42 PM.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 17:21:31
#796 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@pavlor

Quote:

pavlor wrote:
@cdimauro

Quote:
and that's with NATIVE (not trapped) code. Imagine what can happen with the trap & emulate mechanism, whatever optimization they can bring out...


If Hyperion achieves 1/10 of native performance, I would call it success...

10 cycles only to execute the fist instruction of the trap handler, as I already reported, is equal to 20 instructions executed by the SPE unit. So, 1 FPU instruction = 20 SPE instructions. Which isn't realistic, because it's just the first instruction executed.

However the real code isn't made only of FPU instructions. Usually there's a mix of integer and FPU instructions.

So the impact of the emulation depends on this mixture.

Anyway 1/10 isn't realistic.

BTW, with such performance even WinUAE becomes more interesting.


@iggy

Quote:

iggy wrote:
@cdimauro

Quote:
The best thing is to rewrite it from scratch


I never suggested any other way of doing it, I said 'like' Linux.

And I doubt an spe based translation layer will be that much faster than the integer route.
Its the trapping and conversion, not the execution of the converted code, that will carry the most overhead.
AND, anyway, the spe code generated probably will be slower than a standard fpu.

It isn't really a 'wait and see' scenario, its a 'how bad' scenario.

If recompilation isn't part of the solution, my guess is pretty bad.

I have the same feeling.

 Status: Offline
Profile     Report this post  
pavlor 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 17:29:40
#797 ]
Elite Member
Joined: 10-Jul-2005
Posts: 9578
From: Unknown

@cdimauro

Quote:
Anyway 1/10 isn't realistic.


This could be nice bet.

Quote:
BTW, with such performance even WinUAE becomes more interesting.


Not without improvements in other fields (2D, 3D, compositing, network, ...).

 Status: Offline
Profile     Report this post  
iggy 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 19:21:56
#798 ]
Super Member
Joined: 20-Oct-2010
Posts: 1175
From: Bear, Delaware USA

@Spectre660

Quote:
http://www.freescale.com/video/qoriq-p1022-processor-plus-quick-boot-plus-h.264-software-codec-demo:FTF11ENA_665_VID?uc=true&lang_cd=en


800x600?
With HD requiring two cores.
Hm, I have been able to do 720p for some time with more powerful (but not significantly faster) cpus (without SMP).
And 1080p resolutions are possible on faster cpus.

Not a good idea to post videos that make it clear that higher resolution video will require SMP.

Last edited by iggy on 25-Oct-2015 at 07:22 PM.

 Status: Offline
Profile     Report this post  
Spectre660 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 19:56:28
#799 ]
Elite Member
Joined: 4-Jun-2005
Posts: 3918
From: Unknown

@iggy

Single core can playback 720p.

Quote:

iggy wrote:
@Spectre660

Quote:
http://www.freescale.com/video/qoriq-p1022-processor-plus-quick-boot-plus-h.264-software-codec-demo:FTF11ENA_665_VID?uc=true&lang_cd=en


800x600?
With HD requiring two cores.
Hm, I have been able to do 720p for some time with more powerful (but not significantly faster) cpus (without SMP).
And 1080p resolutions are possible on faster cpus.

Not a good idea to post videos that make it clear that higher resolution video will require SMP.

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card

 Status: Offline
Profile     Report this post  
iggy 
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon
Posted on 25-Oct-2015 21:10:38
#800 ]
Super Member
Joined: 20-Oct-2010
Posts: 1175
From: Bear, Delaware USA

@Spectre660

Quote:
Single core can playback 720p.


OK...I'll keep that in mind when we get support for this board...

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 Next Page )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle