Click Here
home features news forums classifieds faqs links search
6195 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
53 crawler(s) on-line.
 16 guest(s) on-line.
 0 member(s) on-line.



You are an anonymous user.
Register Now!
 AmigaSource:  46 mins ago
 matthey:  47 mins ago
 MEGA_RJ_MICAL:  1 hr 4 mins ago
 tygre:  1 hr 27 mins ago
 bison:  1 hr 34 mins ago
 Jasper:  1 hr 40 mins ago
 simplex:  2 hrs 21 mins ago
 ggw:  2 hrs 40 mins ago
 terminills:  3 hrs 18 mins ago
 DiscreetFX:  4 hrs 9 mins ago

/  Forum Index
   /  Amiga Emulation
      /  Amiga 1222
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 Next Page )
PosterThread
IridiumFX 
Re: Amiga 1222
Posted on 3-May-2021 13:15:50
#161 ]
Member
Joined: 7-Apr-2017
Posts: 41
From: London, UK

@NutsAboutAmiga
the VME bus of the 68K is different and incompatible with the 60x bus of the G2 PowerPCs. Expect such a CPLD or a functionally equivalent ASIC or FPGA to already exist on the accelerator cards and do the necessary adaption. It's already snooping the bus exactly like it would need to do in order to interface a CPU of any other ISA. Just not flipping bytes.

When it comes to cache, you surely don't expect the 68K and PPC cores to have cross cache coherency protocols in place right? I am afraid, there's none.
The 68040 only supports the MEI protocol (no 'S' shared bit on cache lines), just like the PPC 603. (The 604 implements the full MESI protocol, but the botleneck here is still the 68K side). So even if you added the necessary bus snooping logic both sides, you could still not share memory among the two CPU caches. You could only try to reduce the number of flushes.

That's why the PPC in the classic amiga world is more of an offload computation engine, but here we digress.

I hope this helps to dispel a few myths the Amiga community keeps repeating like a mantra.

Last edited by IridiumFX on 03-May-2021 at 01:32 PM.

 Status: Offline
Profile     Report this post  
IridiumFX 
Re: Amiga 1222
Posted on 3-May-2021 13:28:43
#162 ]
Member
Joined: 7-Apr-2017
Posts: 41
From: London, UK

@bennymee

I apologise, I should have clarified this.

Your PPC on the accelerator card in the classic Amiga, does not run any 68K Amiga application at all. It's still a 68K CPU that runs them, and the PPC sits idle all of the time.

Only WarpOS/PowerUp specific Apps can make use (note: make use) of the PowerPC. The Amiga OS Exec, ROM based libraries and resources never use nor know of the existance of the PowerPC. I am aware of "SetFunction" patches applied in order to let the Dos library recognize the EHF and ELF code sections. The point is still the same.

Take an OS Friendly app like Brilliance, Aegis Sonix, or any other app not specifically written with PUP/WOS support.

Run them with and without the PPC. Do you think having the PPC onboard would make them faster? I am afraid, it won't.

 Status: Offline
Profile     Report this post  
kolla 
Re: Amiga 1222
Posted on 3-May-2021 13:40:12
#163 ]
Super Member
Joined: 21-Aug-2003
Posts: 1636
From: Trondheim, Norway

@ppcamiga1

Quote:

Which means no changes in code.
C code works on ppc just after recompilation.

That's REALLY not the point!

The point is that with DATA being stored BIG ENDIAN, PowerPC and 68k (emulated or real) can SHARE DATA in memory! The implication being that a 68k MUI program can use PPC native MUI classes, for example. This has zero to do with C code, those MUI programs are written a range of different languages.

Last edited by kolla on 03-May-2021 at 01:40 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
kolla 
Re: Amiga 1222
Posted on 3-May-2021 13:42:32
#164 ]
Super Member
Joined: 21-Aug-2003
Posts: 1636
From: Trondheim, Norway

@IridiumFX

Quote:

Your PPC on the accelerator card in the classic Amiga, does not run any 68K Amiga application at all. It's still a 68K CPU that runs them, and the PPC sits idle all of the time


Wrong. Once you boot OS4 or MorphOS on a "classic" Amiga, all 68k code run only through emulation - the 68k is for all practical purposes not there. Just as if you run PowerPC Linux or BSD on a "classic" with PPC card.

Last edited by kolla on 03-May-2021 at 01:45 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
IridiumFX 
Re: Amiga 1222
Posted on 3-May-2021 13:45:42
#165 ]
Member
Joined: 7-Apr-2017
Posts: 41
From: London, UK

@ppcamiga1

Quote:

trace bit will not help at finding null pointer access.
Especially in code write by others.


you surely are not telling me that you need an MMU to set the Vector #3 (misalinged memory access) or Vector #2 (Bus error, including null pointer dereferencing), isn't it ? Enforcer is a very nice tool, but not everything it does is strictly MMU related.

There are some legitimate usages for the MMU, like GigaMem and paged virtual memory systems in general, or marking areas of memory you do NOT want to cache, like the 68040.library does in order to allow an Amiga 4000 to properly function, or some old ShapeShifter custom video drivers which tried to detect modified memory regions (note: those never got faster than the standard MMU-less ones, but they were a nice experiment nonetheless)

 Status: Offline
Profile     Report this post  
IridiumFX 
Re: Amiga 1222
Posted on 3-May-2021 13:47:28
#166 ]
Member
Joined: 7-Apr-2017
Posts: 41
From: London, UK

@kolla

Quote:
all 68k code run only through emulation

Sorry, Kolla, what else should I add? If you like emulating stuff, you can just use WinUAE, isn't it ? I am missing your point probably

 Status: Offline
Profile     Report this post  
IridiumFX 
Re: Amiga 1222
Posted on 3-May-2021 14:05:44
#167 ]
Member
Joined: 7-Apr-2017
Posts: 41
From: London, UK

@kolla

having a shared address space is not the same as sharing data in memory. The PPC on the classic Amiga is nothing more than a computation engine. It's used like a slightly more intelligent device than a DSP, but exactly the same way.

 Status: Offline
Profile     Report this post  
bennymee 
Re: Amiga 1222
Posted on 3-May-2021 15:32:36
#168 ]
Cult Member
Joined: 19-Aug-2003
Posts: 633
From: Netherlands

@IridiumFX

You are wrong: the PPC does run every 68K application on OS4.x (or earlier versions of MorphOS). The 68K is disabled using these OS'es.

UAE emulates the chipset to some degree, the NG os'es only the 68K.


2nd: on AmigaOS with the WarpUP or PowerUP every Amiga programm which uses datatypes will be accelerated as the datatypes are PPC.
Especially OS friendly programms, there where even patches which make programma who did not use datatypes could use them.
And yes, the programm's it self where still using the 68K.

>do you think having the PPC onboard would make them faster?

No I did not think that, it was benchmarked many times in magazines some 20-25 years ago.
Besided that many programm's have PowerPC-plugin.

Does it mean the PowerPC addon's where the best ever made ?
Well, they where stunning with 2 different cpu's - which where not ment to use together - but also there was the lack of cache on the PPC side. And the context switching offcourse.
The WarpUP vs PowerUP did not help either.
On the other hand they where made as intermediate solutions.

Last edited by bennymee on 03-May-2021 at 03:34 PM.

 Status: Offline
Profile     Report this post  
IridiumFX 
Re: Amiga 1222
Posted on 3-May-2021 16:06:04
#169 ]
Member
Joined: 7-Apr-2017
Posts: 41
From: London, UK

@bennymee

here we agree, and we slowly get back on topic for the thread: these were intermediate solutions. They were meant to bridge towards something new. They were supported in a creative way, because we always had to cope with differences in each and every model, each and every hardware. This, to me, means embracing diversity in our ecosystem.

I would have loved to have an A1222 5 years ago, even if it was sold with FPU disabled, rather than reading threads full of mostly random opinions. And you know what? if tomorrow the A1222 went on sale for a decent price, still without FPU, I'd still buy it ... as a developer

 Status: Offline
Profile     Report this post  
ppcamiga1 
Re: Amiga 1222
Posted on 3-May-2021 17:19:00
#170 ]
Regular Member
Joined: 23-Aug-2015
Posts: 229
From: Unknown

@IridiumFX

On Amiga we have SysBase adress at location 4.
Very fragile and very easy to override.
Which almost always results in os hang.
Enforcer protect first page of memory.
On Amiga with Enforcer software with null pointer access do not hang os.
First page protection is also builtin in MOS and Amiga Os 4.
Vampire has nothing like this.



 Status: Offline
Profile     Report this post  
kolla 
Re: Amiga 1222
Posted on 3-May-2021 23:44:31
#171 ]
Super Member
Joined: 21-Aug-2003
Posts: 1636
From: Trondheim, Norway

@IridiumFX

Quote:

having a shared address space is not the same as sharing data in memory.


One does not exclude the other - but that IS exactly why PPC was the obvious choice, data structures left by the 68k can be used directly by the PPC and vice versa.

Quote:
The PPC on the classic Amiga is nothing more than a computation engine. It's used like a slightly more intelligent device than a DSP, but exactly the same way.


I don’t know what rock you have been living under, it has been possible to use the 68k merely as a bootstrapper for PPC-only OSes since more than 20 years - I have a LinuxPPC CD here somewhere from 1999, it supports APUS (Amiga PowerUP Systems), I have MorphOS 1.4.5 or thereabouts as well as OS 4.1 on my A3000/CSPPC - trust me, the 68060 is NOT in use when running a PPC OS and 68k support is provided through emulation - OS4 has Petunia and MorphOS has Trance - and they are INTEGRATED to a level that you can mix 68k and PPC binaries (PPC programs using 68k libraries and vice versa) as much as you care precisely because shared memory space and shared big endian data formats.

Last edited by kolla on 03-May-2021 at 11:45 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
matthey 
Re: Amiga 1222
Posted on 3-May-2021 23:46:24
#172 ]
Cult Member
Joined: 14-Mar-2007
Posts: 888
From: Kansas

g01df1sh Quote:

hi that was just a guess as I cant come up with any other reason as to why HYperion has made very little progress on os4.1 in the last few years.


How about the obvious reason for Hyperion halting development?

amigakit Quote:

$34 for an Operating System in such a small market as ours is a poor commercial decision because it doesn’t generate enough revenue to pay for developers. This is partly the reason why core OS4.1 development has stagnated. If you recall 6 years ago the price of OS4.1 was 99 Euro.


Amigakit is likely correct that $34 U.S. (28 Euros) for AmigaOS 4 is not enough to fund development. Most likely, 99 Euros was *not* enough either. Judging from financial statements, Hyperion was likely losing money on AmigaOS 4. They needed to raise prices, increase volume of sales or cut costs until they returned to profitability. Raising prices decreases demand partially offsetting gains and the AmigaOS isn't good enough to support a premium price. Increasing sales volume is partially dependent on 3rd party hardware sales but Hyperion found a gold mine by expanding development to the larger 68k AmigaOS market. Developer costs for AmigaOS 4 were cut dramatically and 68k AmigaOS development costs were negligible. Hyperion did what they needed to survive although they may have been too desperate or greedy in not sharing some profits with the 68k developers to encourage sustainable upgrades and trying to register Amiga IP marks as their own for the AmigaOS 68k market.

When could Hyperion return to developing AmigaOS 4?

1. If sales of AmigaOS hardware like the AmigaOne 1222 picks up dramatically
2. If development costs of AmigaOS 68k and PPC can be partially shared
3. If development costs of AmigaOS and the Enhancer Software could be shared
4. If more developers could be convinced to work for free

The cost of an enhanced AmigaOS 4 was 99 Euros for the OS plus 57 Euros for the Enhancer Software which is 156 Euros ($188 U.S.). We can see already why we can't have cheap hardware. The original Amiga hardware was successful because it offered good value which the Amiga hardware today does not. The AmigaOS development budget at CBM was no doubt many times higher than the AmigaOS development budget post CBM, especially when adjusting for inflation. This was possible by spreading out development costs between hundreds of thousands of customers instead of hundreds of customers. We are stuck with a tiny niche market with expensive hardware and AmigaOS.

Most products in niche markets are good (specialized) at leading features desired in the market. The AmigaOS competes in the desktop market. I couldn't find a poll of top desktop features but it is not too difficult to come up with a top feature list.

1. Stability - AmigaOS lacks memory protection
2. Software availability and support - no
3. Hardware support - AmigaOS lacks SMP, 64 bit support, etc. but has some driver support
4. Security - no
5. Multi-user support - poor
6. Performance - good for the hardware used but not available on high performance hardware
7. Ease of use - good
8. Customization - good

The AmigaOS offers poor value in the desktop market. Even as a niche player it really isn't good at any of the most desired features. The following is a grid which includes desktop OSs and some niche OSs.

https://www.g2.com/categories/operating-system#grid

Here we see that the market is dominated by Windows and Unix like OSs which are mostly free. Most commercial OSs have been pushed out of the market as it is too expensive to keep up with technology, especially driver development. The AmigaOS is far behind practically every OS on the list feature wise yet is more expensive than most and only available on expensive low performance hardware. It is interesting that Raspbian made the list despite being on such cheap hardware.

Let's take a look at how the AmigaOS competes as an embedded OS. We will start by looking at top features desired which comes from the 2019 EE Times survey. Which factors most influenced your decision to use a commercial operating system?

1. Real-time capabilities
2. Good software tools
3. Ease of future maintenance
4. Technical support
5. Code size / memory usage
6. Processor or hardware compatibility
7. Support for my processor and drivers
8. Royalty free
9. Documentation
10. Network capabilities
11. Security
12. Overall cost
13. Suppliers reputation
14. Context switch time
15. Modularity
16. Scheduling efficiency
17. Customers desire
18. Multicore support
19. Safety certifications

The AmigaOS has real time capabilities (along with good context switch time and scheduling efficiency) and was on the 68k one of the best for code size and memory usage. It at least has potential as a niche embedded market player. Let's take a look at top embedded OS market leaders.

1. Embedded Linux
2. in house / custom
3. Free RTOS
4. Ubuntu
5. Android
6. Debian (Linux)
7. Microsoft (Windows 10)
8. Microsoft (Windows Embedded 7/standard)
9. Texas Instruments RTOS
10. Wind River (VxWorks)

Like on the desktop, the free Unix like OSs have taken market share while custom market OSs are losing market share. Most of these Unix like OSs have monolithic kernels which are not good for real time. Embedded Linux has multiple patches to reduce the amount of time executing code in the kernel (supervisor mode) which can't be interrupted for real time processing. The AmigaOS has a microkernel where very little time is spent in the kernel and supervisor mode. Embedded Linux was patched to allow a flat memory model with scatter loader for non-MMU using hardware which the AmigaOS had from the beginning. The AmigaOS may be a better embedded OS than many of the top Unix like embedded OSs. While Raspbian was not among the top 20 embedded OSs, the Raspberry Pi form factor was used in 16% of embedded hardware. The Raspberry Pi came out of nowhere to take impressive embedded and desktop market share. That's what cheap hardware does. Too bad the Amiga can't do that.

amigakit Quote:

You shouldn’t compare the price of a Raspberry Pi to OS4.1. They have little in common. OS4.1 is sold in fractions of the volume of the Pi. The economies of scale are completely different.

 Status: Offline
Profile     Report this post  
kolla 
Re: Amiga 1222
Posted on 3-May-2021 23:57:02
#173 ]
Super Member
Joined: 21-Aug-2003
Posts: 1636
From: Trondheim, Norway

@ppcamiga1

Quote:

Vampire has nothing like this.


Actually they have, but it is not accessible for “normal kids”, only for the core. So if you have specific needs, you have to go through Gunnar to get it implemented.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
matthey 
Re: Amiga 1222
Posted on 4-May-2021 0:27:43
#174 ]
Cult Member
Joined: 14-Mar-2007
Posts: 888
From: Kansas

ppcamiga1 Quote:

On Amiga we have SysBase adress at location 4.
Very fragile and very easy to override.
Which almost always results in os hang.
Enforcer protect first page of memory.
On Amiga with Enforcer software with null pointer access do not hang os.


It would be better for the program with a zero page write to be halted and ask if the program should be allowed to continue executing. Programs often crash or hang after a bad pointer access.

I always wondered why the SysBase pointer was not put in a register when programs starts so location 4 doesn't need to be read.

ppcamiga1 Quote:

First page protection is also builtin in MOS and Amiga Os 4.
Vampire has nothing like this.


The Vampire has a Memory Protection Unit (MPU). I don't know if a zero page protection program has been written but it should be possible. It certainly isn't a problem to protect the zero page in an FPGA.

AmigaOS 3 users can use ThoR's MMU library tools which protect the zero page. I believe this is done with the MMU but it may be possible without an MMU using the transparent translation registers. I have noticed increased stability from using this.

 Status: Offline
Profile     Report this post  
ppcamiga1 
Re: Amiga 1222
Posted on 4-May-2021 5:56:14
#175 ]
Regular Member
Joined: 23-Aug-2015
Posts: 229
From: Unknown

@kolla

I try to convice GvB to release it. Communication is very hard and its end in nothing.

 Status: Offline
Profile     Report this post  
ppcamiga1 
Re: Amiga 1222
Posted on 4-May-2021 6:03:35
#176 ]
Regular Member
Joined: 23-Aug-2015
Posts: 229
From: Unknown

@matthey

Quote:

It would be better for the program with a zero page write to be halted and ask if the program should be allowed to continue executing.

Thats how it works under Amiga Os 4.x.
On MOS or Amiga Os 3.x with MMU and Enforcer bad access is logged and program continue executing.

Quote:

I always wondered why the SysBase pointer was not put in a register when programs starts so location 4 doesn't need to be read.

They do this that way 35 years ago, it cannot be changed, so we need MMU.

Quote:

It certainly isn't a problem to protect the zero page in an FPGA.

It is not a problem, but after so many years it is still not done.
Vampire acts like 68000. It is really annoying.

 Status: Offline
Profile     Report this post  
kolla 
Re: Amiga 1222
Posted on 6-May-2021 1:41:41
#177 ]
Super Member
Joined: 21-Aug-2003
Posts: 1636
From: Trondheim, Norway

@ppcamiga1

Quote:

I try to convice GvB to release it.

Release what? The AC68080 MMU? It is already released, it is in the AC68080, used by the Vampire cores... but not readily available from AmigaOS or the user, as that would most certainly break everything. As I wrote earlier, you have to go through Gunnar if you have particual specific needs that may be solved using the MMU.

Quote:
Communication is very hard and its end in nothing.

Maybe your needs weren't specific enough.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
ppcamiga1 
Re: Amiga 1222
Posted on 6-May-2021 6:07:51
#178 ]
Regular Member
Joined: 23-Aug-2015
Posts: 229
From: Unknown

@kolla

I need something that protect first memory page, log access to it, and write where it is in sources.
It is in Amiga Os with Enforcer for more than 30 year.
It is in Amiga Os 4 .
It is in MOS.
But not on Vampire.



 Status: Offline
Profile     Report this post  
Hammer 
Re: Amiga 1222
Posted on 6-May-2021 16:05:00
#179 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4116
From: Australia

@amigakit

Quote:

amigakit wrote:
@matthey

$34 for an Operating System in such a small market as ours is a poor commercial decision because it doesn’t generate enough revenue to pay for developers. This is partly the reason why core OS4.1 development has stagnated. If you recall 6 years ago the price of OS4.1 was 99 Euro.

You shouldn’t compare the price of a Raspberry Pi to OS4.1. They have little in common. OS4.1 is sold in fractions of the volume of the Pi. The economies of scale are completely different.


Raspberry Pi also has UK government funding.

https://magpi.raspberrypi.org/articles/national-centre-computing
The Raspberry Pi Foundation is one of the organisations that, jointly, have secured £78 million in UK government funding.

_________________
Intel Core i9-9900K, DDR4-3800 32 GB RAM, MSI GeForce RTX 2080 Ti GX Trio
AMD Ryzen 9 3900X, DDR4-3200 32 GB RAM, ASUS GeForce RTX 2080 EVO OC
Amiga 1200 (Rev 1D4, KS 3.1.4, 882 40Mhz, 10 MB RAM), .
Amiga 500 (Rev 6A, KS 3.1.4, 68K 50Mhz , 12 MB RAM)

 Status: Offline
Profile     Report this post  
matthey 
Re: Amiga 1222
Posted on 6-May-2021 22:33:35
#180 ]
Cult Member
Joined: 14-Mar-2007
Posts: 888
From: Kansas

Hammer Quote:

Raspberry Pi also has UK government funding.

https://magpi.raspberrypi.org/articles/national-centre-computing
The Raspberry Pi Foundation is one of the organisations that, jointly, have secured £78 million in UK government funding.


I knew the Raspberry Pi Foundation was goverment subsidized and received special pricing in some cases. I believe it is possible to come close to the performance/price ratio of Raspberry Pi hardware with mass production and no subsidies. Most big companies don't try because they enjoy high margins.

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 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